------------ From: /usr/tody/ Thu 18:35:41 29-Dec-83 Package: system Title: software release notices (revisions) Task 'revisions' added to the system package. Please use to record all revisions and modifications to a released package. It is not necessary to record revisions to a package which is still under development and has not yet been released. Usage is very similar to 'bugmail'. Type 'rev' to read all revision notices added since your last 'rev'. Type 'rev package_name' to add a revision notice for the named package. Note that '~e' can be typed while entering a revision notice to edit the text that has been entered, just as '~v' is used in the UNIX mail utility. ------------ From: /usr/tody/ Thu 18:41:47 29-Dec-83 Package: cl Title: config change Stack size increased from 1000 to 4000. Note that scripts are compiled onto the stack, hence lots of stack space is needed to run large, nested scripts. If you see a message something like "pc exceeds topcs", that means that the control stack has overflowed; report it with bugmail system. ------------ From: /usr/tody/ Thu 19:14:32 29-Dec-83 Package: system Title: tail The task 'tail' has been modified to permit negative 'nlines'. If nlines is positive (default), the last nlines lines of the file are printed. If nlines is negative, the first nlines lines are skipped, then the remainder of the file is printed. ------------ From: /usr/tody/ Mon 16:33:01 02-Jan-84 Package: softools Title: syntax bug in spp fixed The SPP would not recognize BEGIN and END statements unless they were all alone on a line, i.e., comments were not permitted on the same line. Edited lexical definition of XPP (xpp.l) and remade XPP. ------------ From: /usr/tody/ Mon 19:06:20 02-Jan-84 Package: system Title: full sysgen I did a full sysgen of the system today; the system libraries, the CL, the SPP compiler, and the SYSTEM package were all recompiled. The value of SZ_FNAME has been increased from 33 to 63, the value of SZ_PATHNAME from 81 to 127. Several minor bugs were discovered in the process (some of this stuff has not been recompiled for months and months). ------------ From: /usr/tody/ Mon 19:09:48 02-Jan-84 Package: system Title: system errors The system error handling code has been extensively revised. The error code is now stored in the system error message file (lib$syserrmsg) on the same line as the message. For example, 572 Out of memory whereas before we had Out of memory and where the error code is defined in <syserr.h> as define SYS_MFULL 572 This increases the reliability of message fetch significantly, and makes it easy to leave lots of space for new messages, and to add new messages. A companion routine to "SYSERR (errcode)" was also added. The new routine "SYSERRS (errcode, string)" looks up the error message string in the system error message file and concatenates the user supplied string, producing an error message of the form "errmsg (string)". The FILERR routine is similar but I have never felt that it should be used outside the FIO package, and I have replaced all non file i/o references to FILERR with calls to SYSERRS. Additional routines, i.e., SYSERRI, may be added in the future if needed. Any changes to the error codes in <syserr.h> require a full sysgen, which is why I waited until now to make this revision. ------------ From: /usr/valdes/iraf/ Thu 10:17:56 05-Jan-84 Package: xtools Title: Bug fix in ranges The routine decode_ranges in ranges.x had a bug which did not allow proper decoding. This bug has been fixed. If any further problems are noted please check with me. The old version of ranges has been (temporarily) saved as ranges.old.x. ------------ From: /usr/valdes/iraf/ Fri 11:06:02 06-Jan-84 Package: xtools Title: Bug fix in ranges decode_ranges Additional info about bug fix in procedure decode_ranges in file ranges.x. The old version had incorrect braces. It turned everything into a range. Thus, "1, 4, 7" became "1 - 4, 7 - MAXINT" It also did not skip arbitrary delimiters after a '-'. Thus, "1 -, 8" would be wrong. Old version: # Get last limit of current range. # Skip delimiters while (IS_WHITE(str[ip]) || str[ip] == ',') ip = ip + 1 if (str[ip] == '-') { ip = ip + 1 if (ctoi (str, ip, last) == 0) last = MAX_INT ---> } else if (IS_DIGIT(str[ip])) { # ,-n, if (ctoi (str, ip, last) == 0) # can't happen return (ERR) } else last = first The new version recognizes all the following cases: "1, 3 5" = 1, 3, and 5 "- 6" = 1 through 6 "-" = 1 through MAX_INT "1 ,, -, 7 , 9 -" = 1 through 7 and 9 through MAX_INT. "1 - END" = 1 through MAX_INT The most useful form is "-" to denote all values. New version: # Get last limit of current range. # Skip delimiters while (IS_WHITE(str[ip]) || str[ip] == ',') ip = ip + 1 if (str[ip] == '-') { ---> while (IS_WHITE(str[ip]) || str[ip] == ',' || str[ip] == '-') ip = ip + 1 if (str[ip] == EOS) last = MAX_INT ---> else if (IS_DIGIT(str[ip])) { if (ctoi (str, ip, last) == 0) ---> last = MAX_INT ---> } else ---> return (ERR) } else last = first ------------ From: /usr/tody/ Fri 17:43:07 06-Jan-84 Package: cl Title: logout procedure The system logout script no longer attempts to delete the graphics descriptor files if they do not exist, avoiding the warning message issued by delete in verify mode. ------------ From: /usr/tody/ Fri 18:37:48 06-Jan-84 Package: system Title: new software development tools The softools package is now available. Currently, the package contains only two programs, MAKE and XCOMPILE. Make talks directly to the UNIX make, sending zero, one, or two arguments on to the UNIX make without modification. This version of make will someday be replaced by a completely IRAF version. XCOMPILE is more sophisticated, attempting to provide a machine independent compiler function at the CL level. Usage is as follows: xcompile (files) The following hidden parameters are provided: files = file1,file2 list of files to be compiled or linked (xcompiler = xxc) name of OS compiler program (output_file = first) name of executable output file (optimize = yes) run the object code optimizer (compile_only = no) produce only object modules (fortran = no) keep fortran source (libraries = none) list of user libraries to be referenced (osflags = none) OS compiler flags (debug = no) print but do not execute OS command (verbose = no) print OS command for verification (fname = ) (syscmd = xxc -F -O file1 file2 -ldeboor -lxtools) (file_list = /usr/irafx/tmp/xcr027809) (mode = ql) Usage is straightforward and should be obvious to the experienced CL user, given the above param declarations. There is one restriction; if a file template is used, it should reference only files in the current directory. Files in other directories should be referenced only by an OS pathname. Libraries are referenced as a simple quoted list, i.e., xc file.x, lib='deboor,xtools' Special OS compiler flags may be passed via the osflags parameter, as in xc "file1,file2", os = '-S' If no output file is named and linking is to occur, the name of the process file is taken from the name of the first file referenced by the files template. ------------ From: /usr/tody/ Sat 12:45:03 07-Jan-84 Package: system Title: 'bad mode spec' bug fixed I finally tracked down the bug that was causing the 'bad mode spec' warning message on exit from a package. After a lengthy and abstruse examination of the CL code, guess what it turned out to be -- the mode parameter in the parameter file was actually bad!! The moral of the story is, try not to overlook the obvious when tracking down a bug. Mode params are treated a bit differently than other params by the CL, and I was assuming that there was a bug having to do with that difference. ------------ From: /usr/tody/ Wed 09:19:51 11-Jan-84 Package: system Title: stdimage modified The device "iis70" has been renamed "iism70" throughout the system. In particular, the environment definition stdimage has been changed to iism70. ------------ From: /usr/tody/ Wed 09:38:22 11-Jan-84 Package: system Title: magtape deallocation at logout time Bug fix; the device was being deallocated within IRAF but not within UNIX, when automatically deallocated at logout time. Replaced task "_mtdeallocate" by "deallocate" in lib$cllogout.cl. ------------ From: /usr/tody/ Wed 22:00:17 11-Jan-84 Package: imio Title: imdelete added The "imdelete(imagefile)" procedure has been added to the IMIO interface. Use this to delete imagefiles, rather than "delete(imagefile)", which currently leaves headerless pixel storage files lying about. ------------ From: /usr/valdes/iraf/ Fri 10:49:39 20-Jan-84 Package: xtools Title: New ranges The range parsing procedures have been rewritten, improved, and documented. The main change is that the syntax allows both ranges and steps. Thus, to select every third number in the range 1 to 15 the range string is 1-15x3 or -15x3. The documentation is in the source text and also in the subdirectory doc. In addition to the original procedures-- decode_range and get_next_number -- a new procedure -- is_in_range -- returns a boolean value for whether the specified number is in the range array. ------------ From: /usr/tody/ Wed 18:13:02 25-Jan-84 Package: system Title: llsq library Lawson's and Hanson's linear least squares routines have been compiled and installed in the system library as the library "llsq". Reference the library on the xcompile command line as "lib=llsq". See me for documentation. Minor modifications to the software were necessary to remove i/o, and the only record of the modifications is in the book describing the routines. ------------ From: /usr/tody/ Thu 13:53:38 26-Jan-84 Package: system Title: exception handlers The call given for a user exception handler in the crib sheet was in error and has been fixed. A handler should return the entry point address of the next handler to be executed, or the magic value X_IGNORE, as the second argument rather than as the function value. The correct call is thus procedure user_exception_handler (exception, next_handler) begin (handle exception) next_handler = (old_handler or X_IGNORE) end This is posted with a call of the form shown below. The value of "old_handler" must be passed to the user handler to be returned as shown above when the exception is handled. call xwhen (exception, user_exception_handler, old_handler) In general, handlers should be chained by saving the old_handler address and passing it back to the system in each handler. This is the only way that all handlers can be called when an exception occurs. In addition to the error in the crib sheet, there was a bug in ZXWHEN which prevented chaining; this has been fixed and a new sysgen performed. ------------ From: /usr/tody/ Thu 14:24:08 26-Jan-84 Package: imio Title: bug in immap fixed The variable "mode" (the file access mode) was being used in place of "acmode" (the image access mode) in several places, preventing special image access mode NEW_COPY from being recognized. Sections of the code for new copy images were therefore unreachable. Mode was changed to acmode in several places, but the code has not yet been tested in NEW_COPY mode. ------------ From: /usr/tody/ Sun 12:58:32 29-Jan-84 Package: fio,system Title: file deletion It is now an error to attempt to delete a nonexistent file (previously, such an attempt was silently ignored). ------------ From: /usr/tody/ Mon 11:14:00 30-Jan-84 Package: system Title: bug in errcode() The old errcode() procedure would return status OK if an error condition was not actually in effect, making it unusable in error handlers. Errcode now returns the integer error code of the last error posted. Whenever an IFERR block is entered, the error code is cleared, i.e., set to OK. The system error codes are defined in <syserr.h>. ------------ From: /usr/tody/ Mon 20:58:46 30-Jan-84 Package: system,clio,softools Title: bug in "xcompile" Several mods were made to fix a bug in xcompile. The bug was caused by "files", called by xcompile to expand the list of files to be compiled or linked. Files would always sort the file list; this was improper since the first file has a special significance (it provides the default root name of the output executable file). Revs: (1) FILES and PATHNAMES were modified by the addition of the hidden parameter "sort", type boolean, default value "yes". (2) CLIO was modified by the addition of the procedure CLPOPNU (open list unsorted) to the usual CLPOPN[IS]. (3) XCOMPILE no longer sorts either the files list or the libraries list. ------------ From: /usr/tody/ Tue 22:19:50 31-Jan-84 Package: system Title: error recovery: file cleanup The file cleanup routine, called during error recovery, did not know the difference between a "string" (sprintf or stropen) file and a normal file. When error recovery occurred while a string was opened as a file, the message "No write permission on file (StringFile)" would be given, misleading one as to the the cause of the error. Fio$fio_cleanup.x was modified so that it merely closes a string file, rather than trying to flush the output, close, and sometimes delete the file. ------------ From: /usr/tody/ Fri 19:42:28 03-Feb-84 Package: fio Title: byte-oriented i/o The following procedures have been added to the FIO interface: areadb (fd, buffer, maxbytes, byte_offset) awriteb (fd, buffer, nbytes, byte_offset) nbytes = awaitb (fd) The old AREAD, AWRITE, and AWAIT are still available, are unchanged, and should continue to be used for programs which can work in units of chars. The new routines are identical except that they are byte-oriented, like the device z-routines. Using these routines, byte-oriented asynchronous block i/o can now be done in a device independent manner (i.e., as opposed to calling the z-routines directly), with full error checking. ------------ From: /usr/tody/ Sun 12:06:41 05-Feb-84 Package: system, tty Title: New TTY package installed A completely new version of the TTY package has been installed, supplanting the old interface. The new version interfaces directly to the UNIX termcap file, which will be exported with IRAF. The termcap file concisely describes the characteristics of many terminals, and is in wide use. This interface change is not upward compatible; all programs which used the old TTY interface must be modified. The system is configured for a system defined default terminal when you log onto the CL (i.e., the VT100 on our systems). Default values are provided for the following variables in the CL environment: terminal termcap name of terminal in use (i.e., vt100) ttybaud baud rate for the terminal (9600) ttynlines number of lines on the screen (24) ttyncols number of chars per line (80) termcap filename of termcap file (dev$termcap) printer termcap name of default printer (i.e., imagen) PAGE, CLEAR, LPRINT, DELETE, etc. all use the new device independent TTY interface. All programs which print tabular data on the terminal, i.e., ?, lparam, etc. in the CL, use the environment variable ttyncols to format the table. The line printer facilities are described in a separate revision notice. If you are using a terminal other than the system default, system.stty should be used to select the terminal. The command stty tek4012 for example, would change the values of the TTY environment variables as required for the Tektronix 4012 series terminals. The command "stty baud=1200" would tell the system that the baud rate is 1200, for example when working over a modem (the system must know the baud rate to be able to generate delays). STTY without any arguments prints the terminal status. The TTY interface is described in tty$TTY.hlp, where tty="sys$tty/". ------------ From: /usr/tody/ Sun 12:36:20 05-Feb-84 Package: system Title: line printer interface The LPRINT task is now installed. LPRINT provides a device independent line-printer interface. usage: lprint (files) params: files = *.x list of files to be printed (device = printer) name of output device (map_cc = yes) make unknown control chars printable (header = yes) enable printing of header on each page (mode = ql) examples: program | lprint # direct output of program to printer lprint "*.x,*.cl" # make package source listing If called with a file list, Lprint by default prints each file starting on a new page and breaks pages, putting a one-line header at the top of each page (like the UNIX pr). If used in a pipe, Lprint copies lines of text to the printer, but does not break pages or print headers. Escape sequences are still mapped, however. Page breaks and headers may be disabled with the hidden parameter lprint.header. The following control characters have a special significance in text input to Lprint (also to TYPE): FF=formfeed advance to new page SO=<ctrl/n> begin standout mode SI=<ctrl/o> end standout mode The TTY interface is used to translate such directives as required for the device. Programs like HELP, MANPAGE, and LROFF insert these three directives into output text to break pages and embolden text in a device independent fashion. TTY is also used to determine whether tabs must be expanded, to define the number of lines per printer page, the number of chars per line (long lines are broken), and so on. The parameter lprint.device determines the printer device to be used. If the value is the string "printer", the system defined default printer device is used (defined in the environment). Some other printer may be explicitly named if desired. For a printer to be accessible, there must be a termcap entry for the device, and the device must be installed in the ZOPNLP printer device table. The devices currently installed are "imagen" and "versatec". Output for both devices is currently spooled and printed asynchronously. Low level interface routines installed in this release include the line printer z-routines (os$zfiolp.c), and the device FIO open procedure LPOPEN, source in "sys$system/lpopen.x". ------------ From: /usr/tody/ Sun 13:02:54 05-Feb-84 Package: fio Title: bug fixes The following bugs|restrictions were found and fixed in the process of installing and testing the new TTY interface and LPRINT program: (1) If a binary file is opened in APPEND mode, and the file does not exist, one will be created and opened for WRITE_ONLY access. Previously, the file had to exist be be opened for appending. This bug did not pertain to text files, which have a different interface. (2) AREADB/AWRITEB: if file is a streaming (sequential only) file, the offset parameter passed to the z-routine is now guaranteed to be zero. The old routines were passing the offset argument from FIO (which is nonzero) on the z-routine unchanged, and the z-routine would perform an illegal seek (ignored on a streaming device, but would cause a write error for the printer). ------------ From: /usr/tody/ Thu 19:35:24 09-Feb-84 Package: os Title: debug routine added to memory allocator A routine ZMEMCK has been added to the OS package to aid in debugging programs which use dynamic memory. The operation of ZMEMCK is highly system dependent. In the current implementation, the UNIX memory allocator keeps a list of memory regions (malloc'ed buffers and other areas). ZMEMCK traverses the list and prints the address and size in bytes (both in octal) of each region, and tells whether or not the region is "busy" (in use) or not. Only regions which have been released with "mfree" will show up as not busy. If memory is corrupted, the list will be trashed near the area of memory where the buffer overflow or underflow occurred. SALLOC stack segments show up in the list as regions with a length of 4004 (octal) bytes. ------------ From: /usr/tody/ Sun 20:33:43 12-Feb-84 Package: system,memio,os Title: bug fixes: malloc,realloc The MALLOC and REALLOC routines were checking for an ERR return, while the OS routine ZMALOC and ZRALOC were returning the NULL pointer in the event of an error. I added a third argument "status" to each z-routine, returning OK or ERR, and changed malloc and realloc to properly check the status parameter. ------------ From: /usr/tody/ Mon 13:17:56 13-Feb-84 Package: fio Title: asynchronous i/o Awrite/await bug: file size was being incremented by await when called after a write. Ok, but the FIO buffer offset was being used by await to determine the final offset as boffset+nbytes_written. This meant that FIO could not keep track of the file size when writing a file using only awrite/await. I moved the updating of the file size into awrite, deleting it from await. This is ok since it is an error if fewer bytes are written than requested. A couple new test programs can be found in debug$afio.x. ------------ From: /usr/tody/ Tue 09:52:36 14-Feb-84 Package: cl Title: radix intrinsic function RADIX has been changed to format the operand as an unsigned integer for non decimal output bases. Thus, "radix(-1,10)" produces "-1" as output, whereas "radix(-1,8)" produces something like "37777777777". ------------ From: /usr/tody/ Wed 15:48:27 15-Feb-84 Package: fio Title: stropen bug fix Stropen had a bug when opening a string with read access. There was a loop in the routine which searched for the EOS; the loop was infinite because the input pointer was not being bumped. Fixed and did a new sysgen. ------------ From: /usr/tody/ Thu 16:28:44 16-Feb-84 Package: softools Title: lroff utility installed A preliminary version of the lroff utility has been installed in the Softools package. This version does not break pages and print page headers, but it is useful nonetheless. This is the new version of Lroff, including support for standout mode, and numbered section headers. The Troff font change escapes \fB, \fI, and \fR are recognized in help text. The .nh or .NH directive is used for numbered sections. Output must be passed through type, page, or lprint to make the font changes "visible". The control chars SO (control-N) and SI (control-O) may be inserted in text to be processed by one of the above output programs to turn standout mode on and off. Processing of control directives is under control of the new TTY interface and is quite device independent. ------------ From: /usr/tody/ Fri 00:05:03 17-Feb-84 Package: system Title: lprint modified Several hidden params were added to lprint. By default, pages are broken only if the standard input is not redirected, i.e., if one or named files are to be printed. It is now possible to explicitly specify that pages are to be broken or not broken. If reading from the standard input, a file name label string may be given, else the file name will appear as "STDIN". ------------ From: /usr/tody/ Tue 19:55:57 21-Feb-84 Package: fmtio Title: gctol, lexnum mods The following modifications were made to the FMTIO routines gctol and lexnum: (1) GCTOL would only recognize the B and X suffixes (for octal and hex numbers) if the suffix chars were in lower case. The procedure now recognizes the suffixes in either case. (2) LEXNUM is a lexical analysis routine which figures out whether a string is an IRAF number, and if so, what kind (octal, decimal, real, etc). The calling sequence is "lexnum (numstr, ip, nchars)". Formerly, the IP arg was input only. Lexnum was changed to always advance IP over any leading whitespace, so that the routine can be used to find the start of a token as well as decide whether or not it is a number. ------------ From: /usr/tody/ Tue 20:03:01 21-Feb-84 Package: softools Title: YACC compiler compiler now available for SPP The Yacc compiler has been modified to produce parsers in the SPP language, and is now installed in softools (this was done because the prototype image calculator needs a parser, as will the database tools, etc.) This is a full implementation, and an SPP parser can be every bit as sophisticated as a C parser, with a parser token value stack of arbitrary structures, the usual error recovery tools, and so on. In combination with the Lexnum lexical analysis routine (used to build lexical analysers), it is straightforward to build compilers for sophisticated applications programs. The standard Yacc documentation remains valid; a list of differences in the SPP version are given in "softools$yacc/README". A complete working example of an SPP/Yacc algebraic desk calculator is given in "yacc/debug/dc.y". This example resembles that given in the Yacc reference manual. ------------ From: /usr/davis/ Wed 13:15:40 22-Feb-84 Package: dataio Title: mtexamine The task mtexamine has been installed in the dataio package. Documentation can be found in mtexamine.hlp. ------------ From: /usr/davis/ Wed 13:24:33 22-Feb-84 Package: dataio Title: rfits, pdsread, rcamera A bug was found and fixed in the error handling code of rfits, pdsread and rcamera. When an error occurred in reading a tape file the routines returned control to the main program without closing the file in which the error was encountered. Any attempt to process the next tape file in the list resulted in a "doubly_opened magtape" error and an abort. ------------ From: /usr/tody/ Sun 13:12:12 26-Feb-84 Package: system Title: hidden param added to type, page The hidden parameter "device", default value "terminal", has been added to the parameter files for the TYPE and PAGE utilities. This parameter may be used to specify that output be processed according to a termcap entry other than that for the terminal. I also added a termcap entry for the pseudodevice "text", which is a device with no capabilities at all (i.e., the capabilities of the final device are unknown). This is useful for removing standout mode escapes from Lroff output, producing a pure text file as output. When producing output for a device which cannot overstrike, embolden, etc., TTY renders standout mode as upper case. Thus, to remove the standout mode escapes from Lroff output, saving the result in a text file which can be printed on any device or system: lroff file | type device=text, > textfile ------------ From: /usr/tody/ Wed 22:26:43 29-Feb-84 Package: system Title: name of LOC function changed to ZLOC The names of the LOC and LOCC functions (OS pkg) have been changed to ZLOC and ZLOCC, to eliminate a library conflict with the NCAR function LOC. Hopefully all references in the sys and pkg directory systems have been found and modified. As far as I know, these functions have only been used in system code thus far, so I doubt if anyone's code will be affected. ------------ From: /usr/tody/ Fri 22:14:43 02-Mar-84 Package: softools Title: xcompile modification The XCOMPILE utility has been modified to permit compilation of Fortran source files without checking for undeclared variables and functions. The old version would complain about undeclared variables, and stop the compilation, whether the source file was SPP or F77. Now the checking is only done for SPP sources. ------------ From: /usr/tody/ Mon 17:41:11 05-Mar-84 Package: memio Title: bug discovered and fixed Woops.. I missed a few occurrences of the function LOC in MEMIO (recall the recent name change to ZLOC). Did not cause any problems because the old object module for LOC was still in the system library, and would get linked in in any case. Fixed memio routines; will leave old loc in library until I am sure all occurences of the old function have been changed. ------------ From: /usr/tody/ Mon 18:44:02 05-Mar-84 Package: softools Title: bug fix in XCOMPILE XCOMPILE had the following bug: if asked to compile a nonexistent file "file.x" when there was a file "file.f" in the directory, it would proceed to add the file "file.f" to its list of intermediate Fortran files to be deleted, and then delete the user's Fortran source! The program has been modified to check for the existence of files. If a file does not exist, a warning is issued, and the file name is not added to any of the internal lists. ------------ From: /usr/tody/ Wed 18:58:58 07-Mar-84 Package: system Title: magtape allocation bug fix The allocate task has been smartened up to determine when the drive has been allocated outside of IRAF. If the OS error message "mt? already allocated to..." appears, the IRAF system will no longer think that it has successfully allocated the drive. Related mods: rewinding has been made an option under control of a hidden param to deallocate. The argument "rewind_tape" has been added to the argument list of the "mtdeallocate" procedure in MTIO, and a sysgen performed. ------------ From: /usr/tody/ Wed 21:16:28 07-Mar-84 Package: system Title: match speeded up The pattern matching utility 'match' was noticeably slow when searching large files or lists of files. I have speeded it up some by treating searching for an alphanumeric string as a special case. A new pattern matching utility, STRSEARCH, calling sequence identical to STRMATCH and PATMATCH, has been added to FMTIO for searching for substrings that do not use metacharacters. MATCH examines the pattern you give it and automatically determines the most efficient searching procedure to use. A new hidden param 'metacharacters', default =yes, was added to permit disabling pattern matching metacharacters entirely. Comparative timings of the three procedures, searching for a simple string in a list of 5000 full length lines: strsearch 9.4 sec, cpu strmatch 43.5 sec patmatch 66.1 sec Real timings, searching for a simple string in a 3711 line Fortran program: old 'match' 20.5 sec, cpu new 'match' 7.8 sec unix 'grep' 4.3 sec unix 'egrep' 2.8 sec Hmm... Looks like we still have a ways to go, but a significant improvement nonetheless. ------------ From: /usr/tody/ Sun 17:28:25 11-Mar-84 Package: system Title: LPRINT no longer spools versatec output LPRINT has been modified to write directly to the versatec (like the UNIX VPR), rather than writing a spool file and disposing of it to the printer when closed. This is desirable for making listings, in which case it is inefficient to spool the output. If spooling of versatec output is desired, run LPRINT with device=lpr; this is the versatec with spooling. Note that in IRAF, pipes are a spooling mechanism (because they are temp files), so spooling with device lpr is generally not necessary. ------------ From: /usr/tody/ Tue 15:22:34 13-Mar-84 Package: fio Title: 'access' modified The behavior of 'access' for the null string and for the standard files has been modified slightly. The behavior for ordinary files is unchanged. The new behavior is as follows: yes/no = access (fname, mode, type) If fname is the null string or a string containing only whitespace, access returns NO (the file is not accessible). If fname is one of STDIN, STDOUT, etc., and the mode and type are given as zero, the function returns YES. If the mode and type arguments are given and a standard stream is named, the mode and type must agree with those given by FSTATI or the function will return NO. Thus, "STDIN" always exists, is accessible with mode READ_ONLY, is of type BINARY_FILE when a task is run from the CL, and TEXT_FILE when run in debug mode. ------------ From: /usr/tody/ Wed 19:58:26 14-Mar-84 Package: fio Title: tempfiles Previously if a tempfile were opened with mode TEMP_FILE and deleted explicitly before task termination, the system would print a warning message when it found (during the file cleanup which occurs at task termination time) that it could not delete the file. File "fio$fsvtfn.x" was modified to silently ignore files in the tempfile list which do not exist at fio cleanup time. ------------ From: /usr/tody/ Wed 20:44:30 14-Mar-84 Package: clio Title: clcmd now flushes STDOUT The clcmd procedure, used by system utilities to send commands to the CL, was modified to flush the standard output before sending the command. ------------ From: /usr/davis/ Thu 14:51:22 15-Mar-84 Package: dataio Title: rcardimage, wcardimage mods Rcardimage and wcardimage can now read and write EBCDIC cardimage tapes as well as ASCII tapes. ------------ From: /usr/tody/ Thu 22:55:09 15-Mar-84 Package: vops Title: bug fix in acht_b.c, efficiency enhancement in achtb_.c. Fixed a bug in the ".. to unsigned byte" type conversion vector primitives. The npix argument, used as a loop counter, is a pointer variable and the pointer was being used as the loop counter rather than the array length, causing the output array to be overrun. While I was at it, I also modified achtb_ to put the loop counter in a register for greater efficiency. ------------ From: /usr/davis/ Fri 15:32:29 16-Mar-84 Package: dataio Title: task reblock The task reblock has been added to the dataio package. Reblock can be used to copy tape or disk files, optionally changing the blocking factor, padding blocks and/or records or changing the density in the case of a tape file. See directory /usr/irafx/pkg/dataio/reblock for documentation. ------------ From: /usr/tody/ Fri 17:46:57 16-Mar-84 Package: imio Title: imunmap bug fix When unmapping a new image which had never been written into, IMUNMAP would attempt to close the pixel storage file even though it had never been written into. The close is no longer attempted unless the pixel storage file is open. ------------ From: /usr/davis/ Mon 14:56:31 19-Mar-84 Package: dataio Title: rfits bugfix A bug in reading FITS files with long headers has been fixed. FITS header parameters with unknown keywords were being added to the user area of the IRAF image header. However insufficient space was being allocated to accomodate the EOS on the character string and the buffer would occasionally overflow. ------------ From: /usr/davis/ Mon 15:07:44 19-Mar-84 Package: dataio Title: rfits mods Rfits now applies bscale and bzero to the integers on tape. If no output data type is specified rfits selects one based on the bits per pixel on tape and bscale and bzero. Rfits now initializes the coordinate trans- formation parameters. ------------ From: /usr/tody/ Tue 19:40:19 20-Mar-84 Package: system Title: eprintf now flushes the standard output A call to flush STDOUT has been added to EPRINTF, so that warning messages will be synchronized with interactive output. Note that writing to STDERR, however (i.e. with putline), does not flush STDOUT. ------------ From: /usr/tody/ Sat 11:14:54 24-Mar-84 Package: fmtio Title: lexnum did not recognize indefs The LEXNUM procedure, used to lexically analyse a string to determine if it is a number, did not recognize the special number "INDEF". It now returns the token LEX_REAL when given INDEF as input. This bug was discovered when trying to read a list containing indefinites using SCANR or SCAND. The scan routines (and GCTOD) will now interpret INDEF as a legal number with the value INDEFR, INDEFD, etc. depending on the datatype requested. ------------ From: /usr/tody/ Sat 21:08:47 24-Mar-84 Package: fmtio Title: printf was not autoscaling properly The printf routines, when presented with a number which will not fit within the given field width, are supposed to automatically decrease the precision until the number fits, and if that fails, print out the number anyway in a larger field (as opposed to printing "******" which is silly in a language with free format i/o). This was not working when printing reals, i.e., "%10g" would result in ********** when printing a very large or very small number, because a real in E format would not fit in a field width of 10 using the default precision (7 digits). The bug has been fixed and a format of %Ng should work without overflowing the field for any N >= 7. ------------ From: /usr/tody/ Sun 22:55:37 25-Mar-84 Package: system Title: package reorganization The ultimate decomposition of the IMAGES package into subpackages has not yet been defined. As a first step, the following changes have been made: [1] The graphics utilities have been moved out of images into a new system package called PLOT. [2] A new images subpackage called TV has been created to hold all programs having to do with the image display. The package name TV is reasonably descriptive and has an unambiguous abbreviation. The utilities in the TV package have been reworked somewhat to more closely resemble their expected final form. Type "help tv" for a brief description of the contents of the package. ------------ From: /usr/tody/ Sun 23:06:25 25-Mar-84 Package: plot Title: new package PLOT Testing of the new plotting utilities introduced last week has now been completed, and the three utilities are fully up. Note that new parameters have been added to CONTOUR and SURFACE to give more control over the plot (floor and ceiling levels can be specified, the number of contours can be specified, and so on). GRAPH can now overplot several curves (lists or image sections), compute projections, draw dashed lines, log scale the axes, annotate plots, and lots more. Manual pages for all these utilities will be available soon. ------------ From: /usr/tody/ Mon 00:08:33 26-Mar-84 Package: plot Title: handy-dandy routines for plotting lines and columns of images Versions of the ever popular Forth/Camera routines PROW, PROWS, PCOL, and PCOLS have been added to the PLOT package. These are implemented as scripts calling GRAPH. GRAPH itself should be called if you want to do anything fancy. ------------ From: /usr/tody/ Mon 10:08:34 26-Mar-84 Package: utilities Title: utilities package online The UTILITIES package has been brought online. UCASE and LCASE were moved from the LISTS package into utilities. We are in the process of adding TRANSLIT, DETAB, ENTAB, PRECESS, and AIRMASS. ------------ From: /usr/davis/ Mon 17:10:45 26-Mar-84 Package: dataio Title: tasks txtbin, bintxt The tasks txtbin and bintxt have been added to the dataio package. Bintxt converts binary files containing only text to text files and txtbin does the inverse. ------------ From: /usr/tody/ Mon 23:20:16 26-Mar-84 Package: softools Title: new task MKMANPAGE A new utility MKMANPAGE is available in softools. Use as follows: mkman taskname This copies the manual page template into the new file "taskname.hlp", leaving you in the editor ready to fill in the blanks. ------------ From: /usr/davis/ Tue 13:15:17 27-Mar-84 Package: utilities Title: task translit The task translit has been added to the utilities package. Translit can be used to delete or replace characters in text file(s). Documentation in translit.hlp in the utilities directory. ------------ From: /usr/davis/ Tue 16:31:16 27-Mar-84 Package: utilities Title: tasks detab, entab Tasks detab and entab have been added to the utilities package. See detab.hlp and entab.hlp for documentation. ------------ From: /usr/tody/ Wed 19:25:49 28-Mar-84 Package: system Title: HELP moved to system; package LOCAL added to root The HELP program has been moved into the system package, with the new package LOCAL taking its place. LOCAL provides a place for users or remote sites to add their own software, without having to modify the system itself. It also provides a way to get software into the system so that people can use it, without having to meet the strict standards required for mainline software. ------------ From: /usr/tody/ Wed 19:29:25 28-Mar-84 Package: local Title: new utility URAND A new utility program URAND (uniform random number generator) has been added to the UTILITIES package. URAND is particularly useful for generating lists to test software. Try the example given in the manual page. ------------ From: /usr/tody/ Thu 01:27:13 29-Mar-84 Package: os Title: zfpath modified The calling sequence for the ZFPATH primitive has been changed from call zfpath (fname, pathname, maxch, status) to call zfpath (fname, pathname, maxch, isdir, status) If the "isdir" argument is XYES, the pathname returned by the primitive will be a directory name, suitable for prepending to a file name to generate a pathame for the file (i.e., on UNIX, the directory pathname will have a '/' as the last character). ------------ From: /usr/tody/ Thu 01:30:38 29-Mar-84 Package: system Title: copy, rename modified; movefiles added The COPY and RENAME tasks in the system package have long been tricky to use, because it was hard to predict when there had to be a $ in a directory name. These programs have been smartened up so that they will always know when a file is a directory. For example, copy files, '.' will suffice to copy the named files to the current directory on our UNIX systems. copy file, lib would make a copy "lib" of the file "file", while copy files, lib would copy the named files to the lib$ directory (provided there is more than one file matching the template). RENAME can no longer be used to move a list of files to another directory; use MOVEFILES for that function instead. If RENAME is called with a list of files, the default action is to change the root name of each file: rename 'task.*', newtask might be used to change 'task.cl' to 'newtask.cl', and 'task.par' to 'newtask.par', all in the same command. A hidden param makes it possible to change the extension instead. If a single file is named, it is simply renamed. MOVEFILES moves the named files to the destination directory. The destination must be a directory or it will abort. A command like move '*.x', name will move the file the subdirectory 'name' if there is one; otherwise the program looks for a logical directory 'name$' and if it finds one, moves the files there (but not across file systems at present). ------------ From: /usr/davis/ Thu 15:33:40 29-Mar-84 Package: xtools Title: gstrsettab bugfix Gstrsettab has been fixed so that it will accept an arbitrary first tabstop and tab size. ------------ From: /usr/tody/ Thu 22:26:21 29-Mar-84 Package: cl Title: fix to avoid deadlock when killing a nonexistent process The CL would sometimes get into a strange state where, in response to a ctrl/c interrupt, it would try to kill a process which had already been killed. Deadlock would occur waiting for the nonexistent process to die; additional interrupts would only result in more attempts to kill the nonexistent process. To be honest, I do not understand why this happens in the first place, but nonetheless the OSKILL primitive has been modified to detect and avoid this kind of deadlock. ------------ From: /usr/tody/ Wed 10:03:36 04-Apr-84 Package: fmtio Title: LEXNUM did not recognize INDEF at the end of a line LEXNUM was using streq to compare its input with "INDEF", thus the test would return false if the input string was longer than 5 characters. Replaced the call to streq with one to strncmp. Indefinites should now be properly recognized in formatted input. ------------ From: /usr/tody/ Fri 11:25:31 06-Apr-84 Package: plot Title: bug in graphics output to batch devices We have been experiencing intermittent failures of the graphics programs when writing to the batch devices (versatec, imagen). The problem has been traced to the system metacode translators (Jan's Fortran programs), which evidently create temporary files in the current directory. I modified the IRAF interface to these routines to call them from /tmp, so that the graphics software will not abort when used in someone else's directory. ------------ From: /usr/tody/ Thu 16:15:00 12-Apr-84 Package: fio Title: bug fix: stderr was being directed to stdgraph output The recent addition of the pseudofile STDGRAPH introduced a bug into file i/o causing the standard error output to be directed to STDGRAPH instead of STDERR. The current CL does not differentiate between STDOUT and STDGRAPH for output, so the ultimate effect was to direct STDERR to the same destination as STDOUT. The file clio.clstdio.x was edited to fix the bug (a missing goto after the entry point for the pseudofile stderr was causing control to pass to the entry point for stdgraph). ------------ From: /usr/tody/ Tue 16:39:09 17-Apr-84 Package: imio Title: readwrite access to an image trashes the user area When opening an existing image for read write access, specifying the size of the user area as zero (because it does not need to be used by the calling program) IMIO would read only the standard header, but would then write out the full header at imunmap time. The user area would thus be overwritten with uninitialized data. Immap was modified to record the amount of header data actually read if opening an existing image, and imunmap was modified to write only that amount out when updating the header. ------------ From: /usr/tody/ Tue 16:43:55 17-Apr-84 Package: mii Title: bug fix in miipak The wrong array was being byte swapped in the call to bswaps. A simple error, fixed by replacing the name of the unpacked data array (spp) by the name of the packed array (mii). ------------ From: /usr/tody/ Thu 20:05:06 26-Apr-84 Package: vops Title: bug in ARAV, the rejection average procedure A bug in ARAV caused all pixels to be rejected in the first iteration (it was never tested and never worked). ------------ From: /usr/tody/ Fri 13:41:26 27-Apr-84 Package: system Title: "sleep" task added to system package The new task system.sleep provides a means for generating delays. The single integer argument specifies the number of seconds of delay desired. ------------ From: /usr/tody/ Tue 22:03:33 01-May-84 Package: imio Title: bug fixes Problems were reported accessing images of more than 2 dimensions: the data was not being written into the image properly. The problem turned out to be the routine which calculates the char offset into the image given the coords of a line. The routine would use the size of axis 3 in the calculation when it was supposed to be using 2. It would also use 2 when it was supposed to be using 1, but we have not seen it up until now because the test images have all been square. All programs which access images of 2 or more dimensions should be relinked. A second problem, reported some time ago, was also fixed. Imdelete would crash when trying to delete an image with a zero length header. Problem was lack of a return statement or error checking in immap after call to imerr; added return statement and it works fine now when there is a header read error. ------------ From: /usr/davis/ Mon 13:12:38 07-May-84 Package: dataio Title: pdsread mods For the sake of uniformity the pdsread task in dataio has been renamed rpds. Rpds now calculates the minimum and maximum data values in the pds image and enters them in the header. ------------ From: /usr/davis/ Mon 13:41:38 07-May-84 Package: dataio Title: rcamera mods Rcamera now calculates the maximum and minimum data values in the camera image and inserts these values in the IRAF image header. ------------ From: /usr/davis/ Mon 16:21:05 07-May-84 Package: dataio Title: rfits mods Rfits in the dataio package now calculates the data minimum and maximum values and puts these in the IRAF image header. ------------ From: /usr/valdes/iraf/ Tue 10:52:31 08-May-84 Package: system Title: interpolator routine The image interpolator library /usr/irafx/math/interp has a clget routine to select one of the interpolator types. The routine is called clginterp. interpolator_type = clginterp (param) char param[ARB] The CL request the parameter param, which is a string type parameter. The string is interpreted into one of the interpolator types in interpdef.h. The possible strings including abbreviations are: nearest linear 3rd order poly 5th order poly cubic spline This routine is implemented using clgwrd and will set an error condition if the input string does not match any of the allowed strings. ------------ From: /usr/davis/ Thu 17:58:52 10-May-84 Package: dataio Title: FITS writer Task wfits has been added to the dataio package. ------------ From: /usr/tody/ Fri 09:07:24 11-May-84 Package: system Title: count would not read from STDIN Count would no longer read from the standard input because of a recent change to ACCESS. Deleted all the file type checking from COUNT; now it counts each file in the list whether or not it is a text file, just like any other utility which operates on a list of files. ------------ From: /usr/tody/ Fri 10:06:11 11-May-84 Package: images Title: imdelete now works in verify mode The IMDELETE utility has been modified to fix a bug in verify mode. The code was using the old TTY routines and had to be modified to open and close the termcap database. While I was at it, I modified the SYSTEM.DELETE program to only access the termcap database in verify mode, making noninteractive deletes more efficient. ------------ From: /usr/davis/ Wed 09:58:49 16-May-84 Package: dataio Title: rcamera bugfix A bug in rcamera resulting in picture that were shifted left one column has been fixed. The problem was that rcamera was reading the first data value as the last header value causing all the pixels to be offset by one. ------------ From: /usr/tody/ Fri 11:41:18 18-May-84 Package: fio Title: file fault bug fix FIO divides a binary file up into a sequence of segments the size of the file buffer. When seeking to the very last char in a segment, FFAULT would fault in the next segment, and then write the first char of data into the char preceeding the first char of the buffer, overwriting the MEMIO header word. When the buffer was subsequently deallocated MEMIO would detect the overwrite and a fatal error would result. There was no problem when seeking (faulting) to any other char in a segment. ------------ From: /usr/tody/ Mon 10:21:25 21-May-84 Package: math Title: bevington library installed The Bevington library has been installed in lib$ for our scientific users. The limit on the number of datapoints the REGRES procedure can handle has been increased from 100 to 1000 points. These routines should not be used in serious software. ------------ From: /usr/tody/ Thu 13:15:47 24-May-84 Package: softools Title: bug fix in SPP compiler When called with the -F flag followed by one or more filenames and then some libraries, i.e., xcompile file.x, fort+, lib=library which is translated to xc -F file.x -llibrary the compiler would not set the -o file because by the time it saw "file.x" it thought it should only make the fortran. Fixed so that it always sets the -o file if it sees a filename. ------------ From: /usr/tody/ Sat 12:47:10 26-May-84 Package: system Title: handy type coercion functions for booleans We often need to convert the return value from procedures like clgetb into one of the integer values YES or NO, and vice versa. To make this easier I have added the following functions to the system: bool = itob (integer_value) int = btoi (boolean_value) ------------ From: /usr/tody/ Sat 12:58:24 26-May-84 Package: system Title: database package renamed The DATABASE package has been renamed DBMS (for database management system, as in all the trade journals). This yields a package prefix of "db>" for the package, and makes it possible to abbreviate DATAIO. ------------ From: /usr/tody/ Sat 16:55:06 26-May-84 Package: system Title: enhancements to HELP The HELP utility now handles paging of multiple help blocks or source files (option op=source) properly. The original program would pause at the end of every screen of text but not at the end of every help block or file. The program also discriminates now between the "more" prompt at the end of a screen of text and that at the end of a block or file. If one answers no to the "more" prompt at the end of a screen, the rest of the block or file in question is skipped, i.e., the program advances to the next block or file. If no is given in response to the prompt at the end of a block or file, HELP quits. Help blocks with multiple keywords are now handled a bit differently too. A single help block may be referenced by several keywords in the help directory; the same block will be printed regardless of which keyword the user enters. Previously HELP would always use the first keyword as the title name, producing an identical manual page regardless of the keyword requested. Now the program prints a manual page named for the keyword requested; and it just so happens that several keywords are linked to the same manual page. An entry is made in the printed manual for each keyword so that the user does not have to know what keyword to look under to find the documentation. ------------ From: /usr/tody/ Tue 20:28:04 29-May-84 Package: system Title: more enhancements to HELP The following enhancements have been made to the HELP utility: [1] Help now generates "manual page" style output, if the standard output is redirected and pagination is enabled. Just pipe the Help output to Page or Lprint and Help will generate manpage output by default. A command such as help 'package.*' | lprint will print all of the manual pages for the named package. [2] Help can now be run on a file or files not installed in the help database. This is useful for developmental packages, for specifications documents, and so on. The "file_template" switch, if enabled, causes the template to be interpreted as a filename template rather than as a module name template. For example, help doc$crib.hlp, f+ | lprint will print the current version of the crib sheet on the line printer. ------------ From: /usr/davis/ Mon 08:35:51 18-Jun-84 Package: dataio Title: reblock bugfix Reblock was not appending the tape file number to the input file correctly when the parameter file list contained only a single entry. ------------ From: /usr/tody/ Tue 11:38:41 19-Jun-84 Package: vops Title: call to error removed from the ACHT routines The call to "error" has been removed from the ACHT routines so that they can be called from Fortran programs (they are used in the new Fortran simple interface to IRAF images). If an ACHT routine is called to convert to an unknown datatype it does nothing. This is in keeping with the convention that the string operators, vector operators, formatted i/o operators, and math library routines be purely numerical procedures. ------------ From: /usr/tody/ Sat 16:33:03 23-Jun-84 Package: vops Title: interface modification The calling sequence of the ABAV (block average) vector operator has been changed and the code has been internally modified. The new calling sequence is abav[silrdx] (a, b, nblocks, npix_per_block) and partial blocks are not permitted. Computation of the block sum is carried out in enough precision internally to guarantee that there will be no problems with overflow for the integer types. ------------ From: /usr/tody/ Sat 21:46:08 23-Jun-84 Package: imio Title: bug fix in IMIO When opening a new or new copy image the image dimensions were not getting copied into the internal array IM_SVLEN, used in the IMIO code where the full (non-section) image dimensions are required. IMIOFF was modified to set this array when the dimensions of a new image are processed. ------------ From: /usr/tody/ Sun 18:41:04 24-Jun-84 Package: memio Title: error checking added to SALLOC The SALLOC procedure will now detect a bad datatype code (arg 3), producing a fatal error. ------------ From: /usr/tody/ Tue 10:41:20 26-Jun-84 Package: system Title: minor mod to RENAME RENAME has been modified so that it queries for the old filename(s) before the new name(s). The order of the queries now agrees with the order of the positional parameters. ------------ From: /usr/hammond/iraf/ Tue 16:53:24 26-Jun-84 Package: dataio Title: new task installed: rrcopy Task "rrcopy" has been installed in the dataio package. This task will read IPPS rasters stored on rcopy tapes and convert the rasters to IRAF images. A rcopy format disk file can also be converted. ------------ From: /usr/hammond/iraf/ Tue 16:56:37 26-Jun-84 Package: local Title: new task installed: fields Task "fields" has been installed in the local package. This task is a list processing tool used to extract whitespace separated fields from the specified files. ------------ ------------ From: /usr/davis/ Wed 08:31:28 27-Jun-84 Package: dataio Title: new parameter in dataio readers It is now possible to specify a tape file number offset in the dataio tasks rpds, rfits, rcamera and rcardimage. The readers will begin begin numbering the output files at "offset + tape file number" instead of "tape file number as before". The offset parameter is useful in cases where one wishes to maintain the same root output file name but read from a different tape. ------------ From: /usr/tody/ Thu 17:49:07 28-Jun-84 Package: fio Title: file fault now zero-fills on a write-only file The file fault procedure was modified to zero the file buffer when faulting "in" a block on a write only file. Previously the contents of the file buffer were uninitialized and if the buffer was not completely written into, garbage would be written out to the file. ------------ From: /usr/tody/ Mon 14:49:47 02-Jul-84 Package: help Title: bug fixes in HELP The following two bug fixes were made to the HELP utility: [1] File mode will now find unnamed help blocks, i.e., blocks beginning with a ".help" not followed by a keyword. [2] Help block name keywords may now contain underscore as well as alphanumeric characters. ------------ From: /usr/tody/ Mon 14:52:40 02-Jul-84 Package: fio, imio Title: optimization A number of small mods were made to FIO and IMIO to enchance efficiency. No interface changes were involved. ------------ From: /usr/tody/ Tue 14:20:23 03-Jul-84 Package: fio Title: whitespace in filenames, numbered files The filename mapping procedure now deletes trailing whitespace from a filename. It no longer maps numbered files, hence numbered files may now be referenced just like any other file. ------------ From: /usr/tody/ Wed 16:44:53 04-Jul-84 Package: images Title: improvements to IMHEADER utility The IMHEADER utility program now accepts filename templates, and hence can be used to print the headers of a list of images. There is both a long and a short form. The name of the pixel storage file is printed, along with a note telling whether or not the file exists. ------------ From: /usr/tody/ Thu 19:27:24 12-Jul-84 Package: spp Title: buffer space in compiler increased The size of the dynamic memory area in the RPP phase of the SPP compiler has been increased to avoid the "in dsget: out of memory" compiler errors we have been getting occasionally. ------------ From: /usr/tody/ Thu 19:38:04 12-Jul-84 Package: plot Title: GRAPH can now put image title on a plot If the parameter graph.title is set to "@imtitle" and the first operand in the list of operands to be graphed is an image, GRAPH will now use the title of the first image as the plot title. This has been changed to be the default action. ------------ From: /usr/tody/ Thu 22:12:43 12-Jul-84 Package: plot Title: variable line widths The plotting interface to the versatec has been changed to permit variable line widths. The appearance of plots on the versatec is much improved, particularly the contour plots. Try it! ------------ From: /usr/tody/ Fri 21:21:38 13-Jul-84 Package: plot Title: new parameter for CONTOUR A new hidden parameter has been added to CONTOUR permitting a zero point shift to be applied to the data before plotting. This makes it easy to adjust the percentage of dashed (negative) contour lines in the plot. ------------ From: /usr/tody/ Thu 16:51:39 19-Jul-84 Package: plot Title: minor mods and bug fixes CONTOUR and SURFACE will now print a warning message if used to operate on a very large subraster; possibly the user did not know about image sections or forgot to use one. The warning message can be disabled with the hidden parameter "verbose". GRAPH had a bug (now fixed) introduced by the recent mod to use the image title as the default plot title. If called to graph a list from the standard input, GRAPH would try to immap the input operand (STDIN) and would discover that it was not an image, which is fine, but then there would be no data left to plot since STDIN is sequential. ------------ From: /usr/tody/ Fri 10:41:45 20-Jul-84 Package: cl Title: new _curpack builtin A new builtin task "_curpack" has been added to the language package in the CL. The function of this task is to print the name of the current package. Used by HELP to order searches of the help database. ------------ From: /usr/tody/ Sun 22:41:02 22-Jul-84 Package: mkmanpage Title: now supports both CL format and SPP (procedure) formats The MKMANPAGE utility can now be used to generate manual page templates for both CL programs and library procedures. Do an lparam for additional details. ------------ From: /usr/tody/ Sun 22:43:44 22-Jul-84 Package: xtools Title: cleanup The XTOOLS package has been reorganized and the code cleaned up somewhat. The package is now online in HELP too. The CLGINTERP routine has been moved from math$interp into XTOOLS. ------------ From: /usr/tody/ Sun 22:46:07 22-Jul-84 Package: help Title: new relase of the HELP program The HELP program has undergone a major revision. Highlights include pre-compilation of the help database, improved error diagnostics, a much larger help database, and speedy response (but it still takes forever to spawn the process the first time). Details will follow soon. Try "help '*.', op=dir" or softools.hdbexamine for a peek at the database. ------------ From: /usr/tody/ Mon 19:28:52 23-Jul-84 Package: cl Title: dictionary size increased, process cache more efficient We had a dictionary overrun recently so I doubled the size of the CL dictionary (which is currently fixed in size at startup time) from 7500 ints to 15000 ints (60 Kb). This will permit more complex CL scripts to be run. The stack area is also fixed in size but judging from the report I heard it was the dictionary which was being overrun. While I was at it I made a mod to improve the efficiency of the process cache. The process cache is no longer flushed when a command is sent to the OS with a "!" escape. The improvement in response should be significant. The process cache is still flushed by CHDIR or when the environment list is modified. Both of these cases will be fixed in the future as well. The builtin command FLPRCACHE has been added to flush the process cache. An explicit command to flush the process cache has been added to XCOMPILE and MAKE to avoid a UNIX abort while trying to rebuild a process which is sitting in the cache. There may be other times when this is a problem, however; beware that the executable file is protected while the process is running, and that an occasional "flpr" may be necessary to explicitly flush the cache. The inconvience should be minor compared to the significant improvement in response we should see for commonly used commands like dir, clear, etc. In particular, the following will not work when a process is running: - The file cannot be written into, e.g., by COPY or even by the UNIX "cp". - The process can be run under the debugger, but you cannot set a breakpoint (cannot write into the executable file which is mapped into memory). Beware of these sorts of problems, it can be hard to figure out what is going on! Finally, a new intrinsic function STRIDX was added. Function is similar to the program interface version. The EDIT script was recoded to use this function, eliminating the use of a temporary file and speeding up EDIT a bit (we still have a ways to go, though). math/bevington/*.f Changed illegal unix style continuation in several Fortran files to Fortran continuation. math/llsq/sva.f Fixed hollerith fields of format statements at end of procedure. sys/clio/zfiocl.x Variable 'nchars' was multiply declared. sys/clio/clgfil.x This file (and the FNTGFN code) was completely rewritten as noted elsewhere in the revs notices. sys/fio/filopn.x The VFN (mapping file) was being opened with read write access even when the file was to be opened read only. This would cause an unnecessary delay when multiple users would try to read from the same file at the same time. sys/fio/vfnmap.x Variable file_exists was never used in vfnmap due to the commenting out of the .zmd file code. The declaration was commented out as well. sys/fio/fcopy.x The call to fmkcopy would fail when the input file was a standard stream. Changed to use fstdfile() to test for a standard file, calling fmkcopy only if not a standard file. sys/fio/vfntrans.x Escape character processing was mapping \X into X for all X, causing file.\X extensions to in effect not be escaped (a later routine would map the extension). Changed to allow escaping of the $ metacharacter but no other, i.e., \$ maps to $ without logical directory expansion, but \X maps to \X for all X != $. sys/fmtio/fmtstr.x sys/fmtio/strtbl.x Did away with the experimental ; in argument lists, using ',' instead. This feature turned out not be useful and has been deleted from the SPP language. sys/fmtio/strtbl.x Modified to use STRSRT to sort the field list. sys/fmtio/fmt_setcol.x sys/fmtio/fprfmt.x Deleted declaration for unused variable 'i'. sys/gio/glabax/glbverify.x Deleted commented out code used to test if log scaling made sense. The associated function gt_linearity() was declared but never used, causing a warning message. sys/gio/gtickr.x Similar problem with variable 'nticks'. sys/gio/nspp/mcscale.x min/max type mismatch fixed. sys/gio/cursor/grccommand.x The unreachable code should have disappeard with the completion of all the grccommand options, hence this routine was not changed. sys/gio/nsppkern/pwry.for This routine was deleted when the new character generator was intalled recently, hence the illegal \\ in fortran problem is gone. sys/mtio/mtupdlock.x Could cause no write perm problem on tmp$ as temporary file name "tmp$mtlok1234ll" was too long for VMS and would cause a mapping file reference. Changed the "mtlock" to "mt" to eliminate the mapping file reference. This is supposed to be a low level routine (it is called by a FIO device driver) hence full filename mapping is not desired here. sys/mtio/rewind.x This was changed in the VMS version of IRAF because the SIRM specified a channel number for the argument to ZZRWMT, rather than the unit number used in the code. No changes were made as it is the SIRM which is in error. REWIND takes as argument a drive name such as "mta", rather than the file descriptor of an open file. call rewind ("mta") This permits rewinding of a drive which is not open, and forbids rewinding of a drive which IS open. This consistent with the philosophy of MTIO in which MTOPEN can only be used to access a single file at a time. It is not safe to assume that a tape drive can be rewound while opened to a particular file on the tape; this does not work on some systems (e.g., AOS, at least as of 2 years ago). sys/libc/mathc.c lib/libc/libc.h,math.h Found a way to eliminate the extra procedure call without adding VMS dependence. The LIBC stuff is nearly machine independent and we should strive to keep it so. The (&(x),&(y)) construct used here (where x and y may be constants) is nonstandard C and should be avoided. sys/libc/cfchdr.c In the VMS version XOK was always returned since FCHDIR does not return a function value. This is not right because FCHDIR takes an error action if it cannot change the directory. I modified c_fchdir() to error check FCHDIR and return ERR if an error occurs. CHDIR in builtin.c checks for this function value. sys/libc/free.c Reference to 'ptr' changed to 'buf'. sys/tty/ttyctrl.x Remove declaration for unused variable 'i'. There were other modifications to the TTY files as a result of the recent effort to eliminate unix like filenames from the system directories. pkg/softools/boot/spp Many changes here. The missing argument in errchk.f was fixed a long time ago. There was also a missing argument in xppcode.c. The 'elusive rpp bug' (jiand problem) was fixed some time ago. Other changes for reordering declarations were reported in the revs notices recently. pkg/softools/boot/spp/rpp/rpprat/lndict.r pkg/softools/boot/spp/rpp/rpprat/outstr.r The declarations for 'cupper' were deleted in the VMS version to avoid a 'declareed but never used' warning message from the compiler. This was an error because the integer function cupper() is conditionally used in the body of the procedure if a switch definition is defined in the rpp include file. I replaced the declaration but put a conditional compilation around it to that it would be declared only if used. pkg/cl/exec.c There was a typo: operator '*' used where '&' was intended. This was fixed some time ago but not recorded. pkg/cl/builtin.c The error checking of the return value from c_fchdir was removed in the VMS version due to a related change to c_fchdir(). I left the error checking in for the reasons noted above. pkg/cl/pfiles.c The operator '&&' was used where '&' was intended, resulting in a nasty bug in pfcopyback(). The weak typing in C strikes again. sys/fmtio/parg.x Changed the default precision of a %f from 1 decimal places to full precision. The former default was causing numbers such as "0.01" to be printed as "0.0". dev/graphcap Added entries for vt100s,vt240,vt240o,vt100so from the VMS graphcap. dev/termcap Added entries for printronix,qms from the VMS graphcap. dev/ Added files edt.ed, emacs.ed. lib/cl.par Made cosmetic changes as in the VMS version. lib/cllogout.cl Commented out automatic magtape deallocation as in VMS. This has caused problems for users multiply logged in, and greatly slows down logout as well. lib/clpackage.cl Added SET defs for "language", "showall", and "editor". Added a "lexmodes = yes" to make command mode the default. lib/login.cl Did NOT add the statement set logfile = uparm$logfile" because the logfile name is controlled by a CL parameter rather than by an environment name. I do not understand why this set stmt was present in the default VMS login.cl file. lig/make.h Added a CFLAGS macro and a .c.o entry which uses it. lib/libc/ctype.h Added #ifdef vms condtional to define "vms_ctype_defs" as a globalvalue. Presumably this is necessary for the VMS C compiler to be able to link in the initialization module ctype.o from the library. lib/libc/iraf.h Added a def "UNIX42BSD" to specify the exact version of UNIX used on the host system. Not yet used anywhere. lib/libc/kernel.h Modifed plotter table entries for the NSPP devices to submit the rasterization tasks as a reduced priority (UNIX only). lib/libc/[kx]names.h Not modified - VMS needs a different version of these files. lib/libc/spp.h Changed define of INDEFL for VMS, added VFN definitions. lib/libc/stdio.h Modified getchar(), putchar() macros to use fgetc() and fputc(). Don't know why this is necessary on VMS, but it cannot hurt and these routines are used too infrequently for it to be worth worrying about efficiency. pkg/help/ pkg/help/lroff/ There were a number of changes recently to make structure definitions more portable, to eliminate ENTRY statements, and to avoid type mismatches in argument lists. pkg/help/lroff/lroff.hlp The documentation for ".in" incorrectly specifed the action as move to col "n+1". Changed to move to "current+n". pkg/language/ Moved in the entire VMS version of this directory to capture the new manual pages therein. Deleted the junk file "words" from the root directory. pkg/lists/ Added the SDAS task "columns" to the package. Some source changes were required, in particular the task was installed int the x_lists.e executable to save disk space. Not tested. pkg/lists/words.x The variable last_nscan was incorrectly declared to be a function. sys/gio/stdgraph/stg[gr]cur.x sys/gio/stdgraph/stginit.x sys/gio/cursor/grccmd.x A new cursor mode option ":.cursor N" was added to allow the user to select the logical device cursor to be used for cursor mode input. The cursor may also be selected under program control by using GSETI to set the desired cursor number and then calling GSCUR to set both the cursor number and position. The kernel will continue to use the indicated cursor so long as cursor reads request cursor number 0; this is always the case for cursor mode cursor reads. For cursor selection to work for STDGRAPH devices, the graphcap entry for RC should be implemented as a case statement, with logical cursor numbers of 1, 2, etc (see the entry for the device "vt640" for an example). sys/dev/graphcap.vt640 sys/gio/stdgraph sys/gio/cursor More changes were required to support setting of logical cursors in cursor mode. Setting a logical cursor with :.cursor now locks the logical cursor until unlocked by a subsequent ":.cur 0" or ":.init", i.e., the locked cursor N will always be referenced regardless of the cursor number specified in a GIO call. In the process of debugging this feature a bug was discovered in the stdgraph encoder. The encoder case statement ($1..$2..$$) was not being processed correctly. sys/pkg/plot A new task SHOWCAP was added to the PLOT package. SHOWCAP is used to interactively inspect and check out graphcap entries. Complex graphcap entries can be encoded and the output dumped in ASCII (rather than graphically) to verify that the correct codes will be are are being sent to the terminal. The task is implemented as an interactive interpreter and is self documenting. SHOWCAP accesses the graphcap database like any graphics task, hence any errors in the links to the database should be detectable by running SHOWCAP. It may be necessary to relink both SHOWCAP and the CL to make sure that both are using the same version of tty$cacheg.dat. Graphcap entries may be cached in object form in libsys.a by running the task MKTTYDATA in package softools, then rebuilding libsys.a. sys/osb/bytmov.s sys/vops/lz/amov.s The VAX memory copy instruction AMOVC3 cannot move more than 65K bytes in a single call. All routines which use this instruction were modified to check for very large blocks and to call AMOVC3 repeatedly to copy such blocks in 65K or smaller segments. sys/etc/symtab/* ***** NEW PACKAGE!! ***** A new system package SYMTAB was added to the system library libsys. SYMTAB is a very general and efficient symbol table package for use in IRAF system utilities, applications, and other places (possibly in HELP for faster help database searches, or a future version of the CL which implements macro expansion). sys/etc/environ.x The hash function used in the environment package was replaced by the function used in the new SYMTAB package, which is simpler and better. pkg/softools/* This entire package has diverged so far from the UNIX version that I have decided (for the present) to support two completely different implementations of the SOFTOOLS tasks, one for VMS and one for other systems. At some point this package will have to be cleaned up and the existing (gawdawful) code thrown away. sys/fio/fmkdir.x pkg/system/mkdir.x sys/libc/cfmkdir.c A new library procedure FMKDIR was added to the FIO package for creating new directories. A new task MKDIR was added to the system package to provide the same facility at the CL level. The FMKDIR procedure was added to xnames.h and a new procedure c_fmkdir() was added to the C runtime library. pkg/system/edit.* The edit script was changed to pick up the name of the editor from the environment, rather than from a parameter. This permits use of a single global variable to select the editor everywhere. pkg/system/lprint.x For reasons I don't understand (despite the comments in the code) the VMS version of this task insists on a separate LPOPEN call for each INPUT file to be printed. This does not make any sense - it should not matter to the LP driver where the input is coming from, or how many files input is coming from. Furthermore, printing each file as a separate VMS print job (presumably with a separate job header etc.) is a drastic change to the LPRINT utility which would render it unusable for making listings of packages (a directory full of files). The typical command to print a full listing of a package is "lprint *.x,*.h,*.com": these files must be submitted to the system as a single print job. I will check into this further when testing the merged system. pkg/system/stty.cl Modified to permit setting the baud rate, screen size, etc. on the same command line when the terminal type is set. STTY with no arguments shows the status; STTY with any combination of arguments sets the arguments. The VMS version of this mod was not quite right, by the way, because hidden parameters are not counted in $nargs. Hence a "stty baud=1200" has $nargs=0 (no positional args), and "stty baud=1200" was being treated the same as "stty". Note: STTY should not be used in the login.cl file because it is slow (it has to spawn the system process to access termcap). Use explicit SET statements in the login.cl file instead, leaving STTY for the interactive user. sys/Makelib sys/gio/Makelib Added entries for the system libraries libcur.a (cursor mode) and libstg.a (the stdgraph kernel). These libraries are required to link the CL hence are installed in the system library. I did not add the library libgkt.a (NSPP kernel) because it is really just a "pkg.a" type library, and is not used outside of the package. Updating of this package shoud done by some sort of make facility, since MKLIB cannot link and install the process. sys/fmtio/cto[il].x I agree that CTOI and CTOL should have the same semantics and that neither should call GCTOL, which is different. Rather than implement CTOI as a call to CTOL I have simply duplicated the same code in both CTOI and CTOL. One does not really save any memory by having CTOI call CTOL because mostly likely a process will only link in one or the other, and it will almost always be CTOI which gets used. If CTOI calls CTOL then the result is something which not only runs slower but which requires MORE core memory. CTOI, by the way, was originally implemented as a call to GCTOL (like CTOL). I wanted something simple and fast (formatted i/o is the slowest thing around), hence I don't like the extra procedure call in this context, even though I admit it is better software engineering (more structured). I noticed in the VMS CTOL that the "if (str[ip] == 'I')" was commented out, presumably because the code works the same whether it is there or not. The reason it is there is for efficiency - it is much faster to compare two integers and branch than to call an external procedure like STRNCMP. The test for 'I' eliminates the call to STRNCMP in virtually all calls to CTOL, since indefinite numbers are rare. By the way, I noticed that the new version of CTOI did use the new IS_INDEF macros, which indicates that somebody out there in VMS land reads the revs notices pretty carefully. Good show. sys/fmtio/* All occurrences of (NNN == INDEF_) were replaced by equivalent IS_INDEF constructs. Probably still a lot of the old constructs left in the other parts of the system. sys/libc/ato[il].c These LIBC procedures were NOT converted to calls to CTOI and CTOL. Formatted i/o is a major consumer of cpu cycles and the original C versions of ATOI and ATOL are optimal C code. The VMS version of ATOI has to pack the numeric string, call CTOI which calls CTOL which makes an unnecessary call to STRNCMP - 5 procedure calls where formerly there was only one. The amount of memory saved is not enough to be worth modifying an existing, functional library procedure, even if it did not make things slower. sys/libc/calloc.c The type coercion (char *)NULL added here and elsewhere is strange - UNIX lint accepts just a NULL which is a legal pointer value in C (see K&R pg. 192 and the example on pg. 97). The type coercion added is probably harmless but should not be necessary - does the VMS C compiler issue a warning message or something? If not and the message is only from VMS lint, then we should leave the NULL as is. I suspect that the VMS C does not realize that 0 is a special number in C and is complaining about an integer to pointer assignment. If that is the case, the problem is with the VMS C compiler and the code should not be changed (unless we are forced to do something because the compiler treats the "error" as fatal). sys/libc/cclose.c Added error checking and function value. sys/libc/cnvtime.c Move c_cnvdate out into a separate file. sys/libc/cenvget.c (and elsewhere) It is common in C sources at many sites to declare C functions as follows typespec function_name (arglist) rather than as typespec function_name (arglist) This is done because [1] there are software tools, e.g., a cross reference utility, which can only recognize a function declaration of the first form, and [2] when editing the file, a "beginning of line" pattern match will find the function declaration (rather than every call to the function plus the declaration for the procedure itself). The C syntax is so subtle that without this convention it is difficult to find things in C source files containing many functions. If C had a PROCEDURE keyword like SPP then I would use the second form, but since it does not and the ^func style is common in UNIX land, I would like to keep things as in the first example. sys/libc/cenvget.c (and elsewhere) I did not change the if (...) return (a); else return (b); constructs to the form if (...) return (a); return (b); The first form is more readable and does not bother UNIX lint. Modern compilers are smart enough to compile the same code in either case, hence there is no efficiency gain in the second form. If the VMS C compiler or lint complains it is in error and should be fixed (rather than IRAF). sys/libc/cerror.c Moved each procedure into a separate file. Added NCHARS as in VMS version. Went one step further and changed all the "return(strlen())" to "return (nchars)" since we now have the NCHARS. sys/libc/cread.c,cwrite.c Did not replace the calls to c_erract and c_error by direct calls to the SPP ERRACT and ERROR procedures. Since errors rarely occur doing so does nothing to make the code go faster; no memory is saved since c_erract is very small (5 longwords) and its size is offset by the extra code needed to call the Fortran procedures by reference. Also the code is prettier if the C binding is used. pkg/dataio/imtext/* This task was found to be extremely inefficient and the output code (ascii encoding) was rewritten, speeding the task up by more than a factor of 10. sys/fio/putline.x In the process of profiling the optimized version of WTEXTIMAGE it was discovered that text file i/o is now dominating the timings (the output text file is enormous). Examination of PUTLINE revealed that it could be optimized substantially, hence the procedure was largely rewritten. This is a major output procedure hence is worth careful coding. I rewrote PUTLINE, speeding it up by more than a factor of 2. sys/os/zfiotx.c Examination of the UNIX version of ZPUTTX revealed some considerable inefficiencies. Every character was being compared to newline even when it was known in advance that this was not necessary. Characters were being output one at a time with PUTC. I moved the test for newline outside the loop, making a loop where newline is checked for and another where it is not. In the latter case data is now output in a block without using PUTC. sys/libc/cxwhen.c Moved c_xgmes() out into a separate file. sys/libc/fclose.c,fflush.c,fopen.c,etc. Modified to make direct calls to the SPP procedures (as in the VMS version) to eliminate the slight overhead involved in the extra call to the procedures of the C binding. sys/libc/* Did not modify these routines to use FGETC,FPUTC rather than GETC,PUTC, since the inline macros are more efficient. Why was this changed for VMS? If it was to save a little memory I think that is a bad tradeoff, the amount of memory saved is insignificant and cpu cycles are usually in shorter supply. sys/libc/system.c Eliminated the call to c_oscmd to avoid an unnecessary and redundant string unpack/pack operation. sys/osb/*.c A number of procedures were changed in the VMS version of the system, converting register variable declarations to register argument declarations. I did not keep these mods as the resultant code is the same in either case and I prefer the former style. sys/etc/main.x A variable was placed in a fake common to defeat optimizations by the VMS fortran compiler that would cause failure in a future magic return from ZSVJMP during error recovery (a very subtle bug found by Jay T. at STScI, after a great effort). sys/etc/qsort.x The array argument x[nelem] where X and NELEM are both arguments would cause a VMS Fortran runtime error when QSORT was called with NELEM=0 (I am surprised that the compiler makes such a check at runtime). X is now dimensioned ARB hence this error should no longer occur, even if QSORT is called to sort an array of length 0. I did not add the test for NELEM==0 since the routine as written will work for this limiting case (the array X will never be referenced). sys/fio/rename.x A change was made to this routine to force the host system dependent mapped filename to come out a certain way. I did not keep the rev because FIO does not guarantee that a filename will be mapped in a certain way, and the external world should not depend upon names always being mapped in a certain way. A change which would slow down rename by a factor of two is not justified. sys/fio/vfnmap.x A number of changes were made here to fix bugs which were independently fixed on our UNIX system. In particular there was a bug where a null length mapping file would be created when a VFN was opened for updating but no update ever occurred. I fixed that independently some time ago (apparently with fewer changes) and will wait and test my version of the routine on VMS. sys/fio/vfntrans.x It was not correct to remove the handling of \ escapes completely in the routine which expands logical directory names. The problem was fixed by having \$ map to $ as before (allowing $ in filenames), but having all \C map to \C rather than C, preserving escaped filename extensions. sys/fio/isdir.x I did not add this routine to FIO, but may do so at some time in the future if it proves useful enough (the name would have to be changed). It was improper for the Makefile in IMAGES to refer to the routine directly in the SYSTEM package, since SYSTEM is a package and not a library (I was not aware of the reference to ISDIR in the IMAGES makefile). The solution for the present is to duplicate the code in both packages. There are an infinite number of routines that could be added to the system libraries; the libraries must be kept small to be comprehensible by programmers, hence I never add a procedure to a library until it becomes clear that it is best to do so. sys/fio/fseti.x sys/fio/fset.h sys/os/zfiomt.c [UNIX] A new fset option F_VALIDATE was added. This is used to validate the contents of the active FIO buffer after a read error, in order to try to recover at least part of the record. The caller must tell FIO how many chars to validate. The next read will return the contents of the buffer. The UNIX magtape driver routine ZZRDMT was also modified to backspace a record and retry the read if a read error occurs. The output buffer (usually the FIO buffer) is zeroed prior to the retry to avoid garbage data in the buffer if error recovery is attempted. My thinking is that the magtape drive will get partially through a dma transfer, detect a parity error, and then abort the rest of the dma transfer, in which case at least part of the buffer should be valid. # If a read error occurs validate whatever is in the buffer # and repeat the read, returning data from the validated # buffer w/o any more io. iferr (status = read (magtape, outbuf, maxchars)) { call fseti (magtape, F_VALIDATE, record_length_in_chars) status = read (magtape, outbuf, maxchars) } The caller must know the expected record length for this type of error recovery to be successful. --------------------------- CL2 integration Phase 1: quick onceover of code and first compile. --------------------------- Moved in new CL wholesale. Deleted all junk files. Inspected Makefile; looked up to date. Tried to compile the CL. Died on first file, "cl.x" (would not compile on UNIX for some reason - I did not investigate). Looked therein and horror of horrors, a hacked version of the IRAF Main. I do not want to support two versions of this very important procedure, so I deleted the hacked version so that The Real One will be pulled in from libsys. It looks like the hacked version of the iraf main was used to avoid the reference to the library routine "cmain" in LIBC (a funny library reference was required to link this module which perhaps caused problems on VMS). I guess a better solution is needed, but it must not involve a second copy of the iraf main. Doing so violates the LIBC interface, resulting in possible failure of the CL whenever changes are made to the iraf VOS. Second try. Lots of error messages, shown below. Actions taken to fix each bug are shown in parenthesis. *.c: CLDEBUG redefined everywhere (-DCLDEBUG compiler flag removed in UNIX Makefile) grammar.y: 120: RETURN redefined ./lexyy.c: 3: BEGIN redefined ./ytab.h: 22: RETURN redefined (etc) (RETURN is also defined as an opcode in opcodes.h. BEGIN is an internal (definition produced by YACC. Solution was to use Y_ consistently for all (tokens in the parser. Incidentally, RETURN is also defined in eparam.h. (Should have been called CR there, since CR is the standard ASCII name for (that key). linker unsatisfied external: _AMOVI (The CL used to have a C function called "amovi" in main.c. This was removed (in the VMS version and a direct call to the VOS procedure AMOVI substituted (instead, presumably in reponse to a library conflict on VMS caused by VMS (not distinguishing between C and Fortran externals in libraries. This is a (violation of the LIBC interface; I solved the problem by changing the name (of the C function to "cl_amovi".) ( (NOTE: the fact that we are doing this at all (playing with the ZSVJMP vector) (is a subtle violation of the LIBC interface, more serious than the use of (amovi, and should be changed (this one is my fault)). "gram.c", line 1116: warning: struct/union or struct/union pointer required (typo: misplaced parenthesis in statement 'curr = coderef(curr->c_args);' (should have been 'curr = coderef(curr)->c_args;') "eparam.c", line 311: warning: illegal combination of pointer and integer, op = It was not obvious what was causing this error, but I decided to postpone that for the moment to clean up the EPARAM code. This code stands out in the CL like a sore thumb, as it is written in a style completely different than the rest of the CL (and at odds with the iraf standard). Even worse, the style is not consistent as if several people with different coding styles had worked on the code. Now that the code is written and people other than the original author(s) have to maintain it, it must conform to the rest of the CL. (busy reworking eparam) Added call to c_ttycdes to ttyexit() to close tty descriptor. Old code was allocating a new tty descriptor on every call, never returning old ones. Replaced environment variables "standout" and "showall" by the single variable "epinit". A name such as "standout" has no obvious association with EPARAM and could easily conflict with other names. For example, use SET declaration "set epinit = nostandout showall" to disable standout and enable showall. Additional parameters can easily be added without cluttering the environment list. Usage is consistent with "cminit", used to init cursor mode. Simplified the code a bit in various spots; some pretty messy stuff in there (admittedly the rest of the CL is messy too). Obviously will have to check this when I get it running again against the original version to see if it operates mostly the same. Added "ehinit" to give user control over EHIST. Current parameters [no]standout, bol, eol. (now for edcap.c) Warning message is now printed if editor is named for which an edcap file cannot be found. Input EDCAP format for keystrokes generalized. Now supports the more readable ^X, \[befnrt] formats as well as octal escapes. GRIPE: The string editing facility is really tied into its use in EPARAM and EHIST, with all sorts of changes occurring internally to the values of global variables used by the calling program. The string editor should have been coded as a self-contained, general purpose subroutine that modifies a string using the EDCAP screen editor interface. (edcap.c and eparam.c now cleaned up and compile with no errors. I am familiar (with them now and ready to tune them when the CL runs (from looking at the (code I am sure that some tuning will be necessary). The rest of the new (code looks pretty good, with the exception of numerous ELSE clauses which (are not connected to the preceding ifs. Try again to make the CL...) CL runs! Sill getting compile time warning messages, however. The warning messages are as follows (numerous occurrences in several files): "globals.c", lines 45-97 warning: illegal combination of pointer and integer, op = The problem is statements like char array[] = "string"; which get converted by XSTR into char array[] = (&xstr[N]); This causes our compiler to issue a warning message and seems like rather a dangerous construct (I am not even sure it is even legal). The following is however ok: char *array = (&xstr[N]); I converted a few of the array[] to *array to avoid this class of problem. The EDCAP initialization remained a problem; the question is, is XSTR worth all this? It makes the compile go MUCH slower, but how much memory does it really save? no xstr text data bss dec hex 267264 56320 115556 439140 6b364 xstr text data bss dec hex 267264 55296 114832 437392 6ac90 ----- ------ ------ 1024 724 1748 For less than 2K I don't think its worth it. Bye bye xstr, at least for now. For comparison, here is CL1.x (current version, no eparm): text data bss dec hex cl1.x 237568 49152 109744 396464 60cb0 There was a time once when the CL fit in a 64K partition. Now the CL grows by almost 64K every time we add a new feature! -------------------------- Phase 2: Test and tune CL2 history.c Removed the "#ifdef ST" dealing with the logfile. Put a compile time switch called SHARELOG in config.h. Define this when history.c is compiled if logfile sharing amongst processes is desired. BUG - system package is not getting defined at cl startup time parser never sees the 'package' statement in system.cl Found it - in decl.c, the procedure procscript() tries to seek to the beginning of the current line with the statement fseek (fp, ftell(fp), 0L); This is illegal in UNIX and in IRAF too, since LIBC emulates UNIX with very few exceptions. What this does is seek to the current position in the file; the file pointer is not moved. I think the VMS kernel has a bug. The original SIRM specified that ZNOTTX returns the offset of the current line, rather than of the next line; I changed that long ago to be consistent with UNIX and other systems and sent you all some mail noting the change. How many occurrences of this bug are there? decl.c: fseek(fp, ftell(fp), 0L); decl.c: fseek (fp, posit, 0L); Only two in CL. Fixed; will fix VMS kernel later. BUG - in eparam. Eparam is not catching interrupts. When an interrupt occurs eparam is simply interrupted, leaving the terminal with echoing turned off in raw mode, reverse video, etc. Evidently the VMS text file driver is turning off interrupts in raw mode, i.e., returning <ctrl/c> to the program as a character rather than generating an interrupt. Eparam 'knows' this and hence does not post an interrupt handler. The easiest fix is to change the UNIX text file driver to behave the same way in raw mode. The problem is that editors generally work in raw mode but do need to be able to catch interrupts. There are some subtle problems to be solved to provide a nice interface that does all this. I will make the following changes to the kernels to fix the problem. ********************************************************************* Kernel change [1] The terminal is placed in raw mode when ZGETTX receives a request to read 1 character (rather than, e.g., SZ_LINE characters). In raw mode echoing is turned off, control codes are not mapped by the OS driver (e.g. CR to newline on UNIX systems), *interrupt is disabled and the interrupt code or codes are returned like any other control characters*, and each character is returned as it is typed, one at a time. [2] Raw mode is terminated when any of the following events occur: - When a ZGETTX call is issued to read more than a single character. - When a special escape sequence is "output" with ZPUTTX. The sequence is arbitrary and should be chosen such that it does not conflict with the sequences used by the terminal. The sequence is only recognized when passed as a record, i.e., ZPUTTX checks for the sequence only if called to write exactly the number of characters in the sequence. The sequence is recognized and filtered out whether or not raw mode is in effect. The actual escape sequence selected is "\e-rAw" (5 characters) where \e signifies ESC (escape). Output of a newline no longer disables raw mode. This eliminates the need to scan the output for occurrences of newline when raw mode is in effect. Raw mode is turned on and off by the character count, rather than by a control function, so that the terminal interface may be totally data driven. The result of a data driven mode switch is that a process may easily command raw mode in a remote process or on a remote node in a network, without complex special control codes. ********************************************************************* os/zfiotx.c (UNIX) Changed to disable interrupts during raw mode. sys/fio/fseti.x If the operand FD is STDIN fseti(fd,F_RAW,NO) will now send the sequence \e-rAw to STDOUT followed by a flush to immediately turn raw mode off. If FD != STDIN but has write permission than the sequence is sent to FD. lib/chars.h Added a new define called INTCHAR to define the interrupt character for SPP programs. sys/gio/stdgraph/stgrcur.x Will now terminate the cursor read when INTCHAR is received. The old interrupt handler was left in for the time being. Also added a putline call to write two null to the terminal after turning raw mode off with FSETI, to physically turn raw mode off in case the CL is interrupted before normal iraf i/o to the terminal occurs. pkg/cl/builtin.c Removed the ifdef vms around the beep, clear, etc. tasks. For the moment the ifdef remains for the magtape allocation code but I will get rid of that soon too. pkg/system/ Removed the beep, clear, sleep, and time tasks from this package since they have been moved into the CL language task. pkg/cl/edcap.c Changed show_edithelp to sort the command list and print it in two column format using STRTBL (like the package menus). This makes it easier to find commands in the list, and duplicate entries (e.g., several different ways of moving down) appear in the same place in the table. pkg/cl/eparam.c Displaying a large parameter set seemed slow, particularly when toggling back and forth between the help screen and the parameters. I tried to speed it up by optimizing the e_display_key routine to output fewer characters; helped only a little. The name of E_DISPLAY_KEY was changed to E_DRAWKEY to avoid a name conflict with E_DISPLAY on systems that only keep 7 or 8 characters of a C identifier. pkg/cl/eparam.c The action EPARAM takes when incorrect input is entered is as follows (integer value entered for a boolean parameter): [1] message is printed at bottom of screen, says respond with yes or no. [2] I type anything. [3] Screen is repainted, then immediately repainted again, leaving the cursor positioned to the first parameter even though I was several parameters into the set. I tried this on both the (unchanged) VMS system and on the UNIX version of CL2, with the same results. Certainly this must not be the way errors are supposed to be handled. I changed the error handling code to print the error message at the bottom of the screen and ring the bell (IRAF uses the bell to bring errors to the attention of the user), leaving the cursor at the former keyline so that a different value can be entered. The error message is limited to one line of text. (EPARAM seems to work pretty well now, although I have yet to test the rarely (used features - multiline prompts, arrays, etc. Now lets have a look at (EHIST). History Editor ------------------------------------ I added an environment variable EHINIT to initialize the history editor, with "standout", "nostandout", "bol", and "eol" forming the set of currently recognized switches. If EHINIT is not defined at all then history editing is turned off and the old CSH style history mechanism is in effect. The history editor is quite useful for its most common function - editing one line command blocks - but has some serious limitations, as noted below. Please note that despite all the criticism recorded here, I feel that the current history editor fulfills its primary function admirably and am quite happy with it. The current history editor has two major problems: [1] it cannot emulate VI, and [2], it is not useful for editing multiline command blocks. The editor as coded knows that is an "insert mode" editor, rather than a "command mode" editor or an editor which supports both modes. Hence I can prepare an EDCAP file for part of VI but the commands never get executed, since they are not escape sequences. There is no mechanism for switching between command and insert mode, which VI requires. In VI one types "i", "a", "A", "o", "O", "cw", "r", "s", etc. to insert, add, or change characters. Escape is used to terminate insert mode and return to command mode. The current string editor is always in insert mode (it doesn't know about modes). In VI there is no undelete char, word, line, etc. Instead there is the much more powerful "undo", which undoes the last modify command, no matter what kind of operation it was. Despite the fact that ehist cannot emulate VI, I was able to prepare an insert mode, control code type EDCAP file called "dev$vi.ed" which is easy for VI users to learn to use but which does not emulate VI. This is acceptable for editing single line history records. Editing of multiline command blocks does not work well enough to be useful. True VI emulation is a must for this type of application. The help feature does not work for history editing. Upon returning from help the current line is drawn in the middle of the help text; if editing a command block, only that one line is drawn and the rest of the block is editable but not displayed. Tabs are common in command blocks but are not handled correctly by the editor (they are treated as spaces - I added a klude fix to map them into spaces so that at least the cursor will be positioned correctly). It would be nice if the command block to be edited could be printed at the current position rather than at the bottom of the screen, although I realize there are technical problems in doing this. Relative rather than absolute cursor motions should solve the problem and are more efficient than absolute cursor addressing. My impression from using the history editor and studying the code is that it may have been a mistake to try to use the same editor for both EPARAM and EHISTORY. The parameter editor is really quite a different beast, more of a forms editor than a text editor. All that is really needed for EPARAM are cursor motions and field replacement (and the help facility). I would like eventually to have a history editor which emulates VI and and be used to edit multiline command blocks. We should strip the current EPARAM code down, producing a compact package which is used only to edit CL parameter sets. A general screen editing facility should be generated from scratch which has no ties whatsoever to EPARAM (it could be called by EPARAM to edit withing a single field - parameter value- if desired, but not to navigate about a parameter set). The screen editor subroutine, e.g., SCREDIT, would take as input a sequence of chars and produce a sequence of chars as output, using a self contained screen editing capability to perform the transformation. scredit (in, out, start, editor) int in, out # input and output file descriptors char start[] # defines initial cursor position char editor[] # editor language to be used Here, IN and OUT are FIO file descriptors which in the case of the CL would be associated with memory resident strings with STROPEN. START is a string symbolically defining the position within the file (line, column) to which the cursor is initially to be positioned. Without further study I cannot say whether the same code should be used to support both command mode and insert mode editor languages, or whether two separate codes should be used internally. The SCREDIT facility should be coded in SPP and placed in a system library since it would be self contained and usable in any program, e.g., in a portable IRAF text editor, or in a DBIO table editor. I suggest we leave EPARAM and EHIST much as they are now, for the moment. I will add the string editor subroutine to the program interface as part of the DBIO implementation. When this becomes available, the EPARAM and EHIST code will be restructured to use the new facilities, and we will also be able to start providing screen editing in the user interfaces to various IRAF programs. In particular, we will then have a simple but useful text editor as part of IRAF, usable on all IRAF systems (probably limited to small files, e.g., < 100K) and usable to edit files resident on remote nodes accessed by the kernel server (this is a problem we currently face with our network here). --------------------------------------------------- Now we proceed on to the language features. 1. Packages, Tasks, Parameters The CL handling of these checks out ok as far as I could determine. I tried package loading and unloading, verified that package list was as it should be, storage is being freed as it should be. The naming syntax package.task.param.p_field works as it used to. Parameters may be used in expressions and set in assigment statements as they used to. Have not yet retested parameter indirection, but we have packages that use it so will find out soon. 2. Expressions Simple expressions worked fine, however I did find some problems with the more subtle forms; these are discussed below in the context of bug fixes. 3. Control Flow A goto with an undeclared label causes an access violation. Trying to declare the label (with syntax "label:") caused a syntax error. FOR, BREAK, NEXT check out. Haven't checked SWITCH CASE yet. 4. Procedures I could not get procedures to work at all, at least when entered interactively from the terminal. Perhaps I will try again later, but this feature is not needed at this point and I have little time. I always get the message "illegal variable defn's". Typing in a procedure with null arglist and no declarations caused "illegal parser state" after the "begin". The declaration "int procedure junk()" was a syntax error. By the way, there are two reduce/reduce warnings in the grammar. In case I don't have time to look into these, I should point out that these usually indicate a serious problem in the grammar. Shift/reduce conflicts on the other hand are normal and are no problem provided the precedences have been defined properly. ------------------------------------- pkg/cl/gram.c The following statements would fail with a syntax error: = gcur # cl parameter - global graphics cursor I can see that the recent introduction of declarations forces us to reserve the type specifiers as keywords, else a declaration would look like a task statement. I could not bring myself to change the name of either the datatype "gcur" or the global cursor parameter of the same name, so I built the damn things into the grammar. I hate to do that but we would have a poorer user interface if we used different names in the two contexts. Having to reference "cl.gcur" is not acceptable. The keywords "gcur" and "imcur" are now recognized in expressions, permitting both "= gcur" and statements like "print (gcur // 'abc')". The form "gcur=" is however not permitted. sys/clio/clreqpar.x Changed the form of a runtime parameter request from "param=" to "=param". Perhaps we can do away with the former type of statement at some point, as it is no longer used anywhere. This was changed at this time so that the CL will receive "=gcur" requests rather than "gcur=" requests; the latter form will now cause a syntax error since gcur is a keyword marking the start of a variable declaration. pkg/cl/grammar.y The declarations syntax looks more complex than it ought to be. The parser now has twice as many states as it used to. This is not necessarily a problem but could indicate a "syntax error" prone grammar and/or a slow grammar. The last time I profiled the CL the parser turned out to be the biggest consumer of cpu cycles. Just turn on yydebug and see how many states it visits to parse a trivial statement and you will see why. NOTES on the grammar: [1] The form of an initializer should be "var = const" or "var = { key=value [, key=value] }", i.e., there should always be an '=' following the variable or array specifier if there are any initializations, whether or not these are compound. Always using '=' for initializations should simplify the grammar, in addition to making the CL consistent with C and SPP+. ------------------------------------------------------------ [Added Later] On second thought I am not sure about this one. I can see why the current syntax was chosen - it is because the initializer is a list of assignments rather than values as in C. The syntax is that of a structure with initializers for the individual elements, although it does not look like it when entered all on one line, e.g.: int param { default = 0, (alias ??) min = 0, max = 1, prompt = "prompt" } I missed that earlier - I guess the syntax chosen is the best after all. ------------------------------------------------------------ [2] I note that "char" is being used to declare string variables. This seems wrong; char is an integer datatype in SPP, as in C. There should be an explicit "string" typespec corresponding to type "s" in the parameter files. Note that SPP also has a string type. Syntax, e.g., string alpha = "text" (1) string beta[5] = { "a", "b", "c" } (2) char zeta[100] = "text" (3) In case (1) the length of the alpha string variable is the default, e.g., 64 chars. In case (2) BETA is an array of strings, each of the default length. In case (3) ZETA is a conventional character array as in SPP or C (****** but the actual amount of storage allocated should (****** be 101 chars, as in SPP). [3] It looked like the grammar included some support for array constants, e.g., int array[n]; array = (1,2,3,4) (1) or userproc (a, b, (1,2,3,4), c, de) (2) Case (1) assigns an array constant expression to an array variable. Case (2) passes such a constant in an argument list. If you folks are already doing that, congratulations, it is a great idea. By the way, can the N in "[n]" be a runtime variable or an argument, if "array" is a local variable? If we are going to support arrays and expressions involving arrays in the language (and we already do), then clearly we need to support array valued constants as well. -------------------------------------- Run time testing -------------------------------------- pkg/cl/builtin.c Put checking for error status return back in chdir. Deleted all the _static in the procedure declarations. Why bother? Static procedures have caused portability problems, and as long as the iraf procedure naming convention is followed (using a package prefix), there should be no problem with name collisions with other externals. Using static procedure names has one very real disadvantage, namely you cannot debug these procedures normally with the debugger since the symbol table info aint there no mo. pkg/cl/opcodes.c Deleted all the "_static" and the #ifdef vms. Add an "o_" prefix to all the opcode function names to avoid name collisions with other externals. pkg/cl/config.h Deleted the #ifdef vms definition for _static. lib/libc/libc.h No change, just some comments while checking up on LIBC due to a bug. There were a couple of things in the VMS libc which I did not keep: defines for u_upkstr,u_upksize (not used anywhere, defined locally in c_sppstr), and an #ifdef vms to define import_xnames. I suspect the latter was just because someone got lazy - in any event, iraf standard does not permit nested includes since I feel include file dependencies are potentially very dangerous and should be evident from looking at a file. Also, by avoiding nested includes, it is easy to prepare the list of dependency files for the Makelib files. sys/fio/fstati.x Was erroneously using "fp" rather than "ffp" when returning the value of F_MAXBUFSIZE. sys/mtio/* sys/os/zfiomt.c [UNIX] The datatype of the "drive" argument to the MTIO zz-routines ZZOPMT and ZZRWMT was changed from integer to character string. This is consistent with all of the other ZOPEN procedures, and provides more flexibility in how the drive is named. The magtape naming convention was also changed slightly to reflect this change, e.g., if the magtape is named as mta.1600[3] then file three of magtape unit A will be opened at 1600 bpi. The former syntax was "mta1600[3]". The significance of the new syntax is that everything between the "mt" and the "." (or EOS) comprises the DRIVE string sent to the kernel to access a drive. The contents of the DRIVE string are therefore no longer constrained to be "a", "b", "c", etc. and the string may be used to convey additional information. In particular (and this is the primary reason for the change), the drive string may now contain a node name prefix, permitting use of a drive on a remote node. For example, to reference logical unix A on node "draco" (drive "draco@a") either of the following will work: mtdraco@a.1600[3] mt.draco@a.1600[3] The syntax "draco@mta.1600[3]" would have been more pleasing and more consistent with node names in filenames, but the original MTIO specifications guaranteed that magtape names passed to MTOPEN must begin with the prefix "mt", hence the former syntax was chosen. Note that the usual magtape references "mta", "mtb", etc. are still recognized (the "." delimiter after the "mt" is optional). -------------------------------------------------------------------------- KERNEL CHANGES sys/os/zfiomt.c Change the "drive" argument of procedures ZZOPMT and ZZRWMT from an integer to a string. Change device table accordingly to permit table lookup via string match. Use of "a", "b", etc. for logical drives on a node is still recommended for consistency, although additional flexibility is now possible. -------------------------------------------------------------------------- sys/osb/mii.x Moved each procedure in this file into its own file to avoid loading the whole thing in each image. Modified the low level routines MIIPAK16, etc. to be self contained so that they can be called by the user rather than load the entire package. This involved moving the byte swapping code into these lower level procedures. sys/osb/bytmov.c This routine had a bug which was never discovered since we use an assembler version instead on the VAX. Rewritten to eliminate the bug. sys/gio/gopen.c dev/graphcap Added device "stdvdm" (alias "vdm") to the graphcap. VDM is the virtual device metafile, used to spool graphics output in a file. GOPEN was modified to accept "vdm" as an abbreviation for "stdvdm" and to use a default file ("uparm$vdm") if stdvdm is not defined in the environment. sys/gio/stdgraph/* Replaced all calls to TTYCTRL with calls to STG_CTRL, to permit use of graphcap encode/decode instructions in all graphcap entries. sys/etc/onentry.x Bkg filename was not being packed before call to ZOPNTX. Noticed this in passing; the standard onentry has never been tested as a means for running iraf tasks in the background. ------------------------------------------------------------------------------- Commencing installation of networking software (package KI) The KI is a sysgen option, not required for operation of iraf. ------------------------------------------------------------------------------- sys/ki/ Added package KI (the kernel interface). sys/os/zfioks.c Added a new device driver ZFIOKS, the kernel server device driver. This driver is the iraf interface to the networking facilities provided by the host system. sys/cl/prcache.c Changed the names of the CONNECT and DISCONNECT procedures to PR_CONNECT and PR_DISCONNECT. The unix networking system library includes a connect procedure, causing a library conflict. lib/knet.h lib/libc/knames.h sys/*/*.x Added <knet.h> to LIB and added includes of this file to all VOS procedures which do kernel i/o. Added a KNET section to the C include file <libc/knames.h>. These two include files determine whether or not a system is to be configured with networking capabilities. A system can be configured without networking simply by defined "NOKNET" before knames.h is loaded in a C file and by commenting out the defines in <knet.h>. sys/etc/main.x The inchan, outchan, device, etc. arguments passed to the IRAF Main by ZMAIN refer to the real kernel; if networking is used these must refer to the kernel interface instead. A subroutine call was added to map these values is networking is in use. This was done in such a way that the kernel itself did not have to be modified. One of the goals of the kernel interface design was that the KI shall require no modifications to the kernel beyond the addition of the ZFIOKS driver. sys/etc/envnext.x,environ.x,Makelib Two new procedures env_first and env_next were added to the environ package to permit traversal of the environment list by external programs. sys/etc/envscan.x Changed to use stropen to open the input string as a file, so that a string buffer containing many set statements may be processed in a single call. sys/fio/getline.x Getline was assuming that ZGETTX returns an EOS delimited string; this just happened to be the case with the unix kernel, but is not required by the si specs. Changed to EOS delimit the string using the count returned by zgettx. sys/libc/ckimapc.c Added ki_mapchan to LIBC to permit recovery of the host system channel code or pid and node name in the CL, e.g., for "prcache" output. Note that when the KI is in use KI channel numbers are returned to the VOS rather than OS channel numbers. pkg/cl/prcache.c Modified the prcache output display to print both the OS pid in both decimal and hex and the node name. The display formatting was improved to eliminate the blank area associated with the hex number. This code will work the same whether or not the system is configured with networking enabled. The information now printed for each process in the cache is the following: [v] node@pid(pidX) bits processname where V is the VOS pid (a small integer, typically one digit), NODE is the name of the node on which the process is running, PID is the host pid in decimal, PIDx is the host pid in hex, BITS are the process status bits (hibernating, running, etc.), and PROCESSNAME is the name of the process. A process may be flushed or locked by referencing either the number V or the name of a task in the process. dev/graphcap,termcap Added a new kernel string type parameter DD to the entry for each plotter or printer. This string is passed to the device driver at device open time and replaces the device tables formerly given in kernel.h (and should replace the table used in VMS as well). The scenario is as follows: the device open procedure GOPEN, LPOPEN, etc. (or a graphics kernel such as the NSPP kernel) calls TTY to fetch the graphcap or termcap entry for the device. The value of the parameter DD is obtained and passed as the "osfn" argument to the device open Z-routine, e.g, ZOPNPL or ZOPNLP. The syntax of the DD entry is determined by the device driver. For the existing drivers I have adopted the following syntax: :DD=node@device,string,string,...: where the comma delimiter may be any character, the value being set by the first alphanumeric character following the device name. The colon character : may not be used in the entry, of course, because it is a termcap metacharacter. The node name and delimiter are optional and are removed from the string by the kernel interface; the z-routine sees only the characters following the node delimter character @. The two logical nodes names "plot" and "print" are reserved for use to identify the node to be used to process graphics output or print text. The graphcap entries, for example, usually reference the "plot" node, e.g., ":DD=plot@versatec,...". The plot node is identified in the host name table dev$host (see KI documentation). If no node has the alias "plot", the default node will be used instead, e.g., the local node. sys/tty Given the above revision, it is no longer necessary to place any system level information in the environment entries for devices - nobody made use of that capability in any case. The name parsing code in TTY is however harmless and will be left in for the time being (in fact I extended the syntax for some reason before adopting the graphcap type solution). sys/gio/nsppkern Added support for the DD parameter. The value of this parameter is now passed to ZOPNPL at device open time; formerly only the device name string was passed. pkg/system/lprint.x Ditto; added support for DD parameter. pkg/images/display/tv/iisopen.x pkg/images/display/new/iism70/iisopen.x (throwaway or prototype interfaces). Added explicit node reference for the IIS so that the affected programs will work on all NOAO nodes. This is good enough for these interfaces since they are due to be replaced later this year. The only change required was to replace the (explicit UNIX) pathname "/dev/iis" with "lyra@/dev/iis". lib/libc/kernel.h [UNIX] The device tables for the plotter and printer devices were deleted since this information has been moved to graphcap and termcap. [Probably there should be printcap, rather than putting the printers] [in termcap, but that can wait... ] ----------------------------------------------------------------------------- KERNEL CHANGE sys/os/zfiopl.c [UNIX] sys/os/zfiolp.c [UNIX] Kernel drivers rewritten to obtain plotter information from the ZOPN argument rather than from the kernel.h table. ----------------------------------------------------------------------------- sys/tty/ttyopen.x In the process of revising graphcap and termcap I found and removed several unnecessary restrictions on the termcap format. Blank lines are now permitted, a capability entry may now extend over several lines and each line need no longer begin with a colon (leading whitespace on each line is omitted), and the colon character may be included in an entry via a conventional backslash escape \:. For some reason the original UNIX termcap format was more restrictive but I see no reason to preserve these restrictions. sys/gio/cursor If the system failed to open a plotter subkernel during a :.snap the graphics system would be left in a bad state. This would usually happen when the user entered an invalid device name for the snap output. Cursor mode will now recover properly from such an error. pkg/cl/builtin.c Reinstalled builtin task "gflush"; this was lost when the new CL was installed. pkg/cl/errs.c Reinstalled call to XER_RESET in errs.c to initialize the error stack when error recovery takes place (necessary since the iraf main is not being allowed to do error recovery). There should be a LIBC mechanism for doing this. pkg/cl/edcap.c Had to turn off raw mode to print the editor keystrokes because UNIX disables output processing of the newlines when in raw mode. IRAF kernel drivers which have to do such output processing themselves should also disable such processing when in raw mode, like UNIX. pkg/cl/main.c Added a call to gflush to the shutdown procedure to flush graphics output and terminate any graphics subkernels when the user logs out. pkg/cl/lexicon.c Added some context mode logic to recognize [] as command mode metacharacters only when they occur on the left hand side of an assignment statement. Hence an array reference such as v[4] = 123 is legal in command mode, as is contour m74[*:8,*:8] A reference to an array element on the right hand side of an assignment is legal since the rhs is interpreted in compute mode. An array reference in an argument list or expression must be parenthesized to be recognized as such, like *, /, etc. sys/gio/cursor/grcpage.x The help cursor mode feature used to display a "[hit return to continue]" message at the end of the help screen. This has been changed to a more informative message which includes prompts for the hidden keystrokes - in particular it is now easy to get cursor mode help. sys/etc/main.x sys/clio/clcache.x sys/fmtio/clscan.x Added parameter caching to CLIO. The iraf main now permits parameters to be set on the command line when a task in invoked. If a parameter is set on the command line when the task is invoked then CLIO will automatically satisfy all CLGET requests with the cached value of the parameter without generating a query. This is useful for tasks that have many (dozens of) parameters, since all of the parameters with values set at task invocation time (e.g., hidden params or params set on the command line) can be cached, greatly reducing the number of runtime queries to the CL and speeding task execution. This might save several seconds of clock time when executing a task with a couple of dozen parameters, since handshaking over the IPC is expensive. Parameter cacheing is implemented in CLIO using the new SYMTAB package, providing efficient hash table lookup. This was probably not necessary but I computed the object size overhead associated with using SYMTAB and it probably only adds 1K or so (SYMTAB is compact). The iraf main command line syntax now permits both i/o redirection and parameter set arguments of the form param=value and param[+-]. Whitespace delimits arguments. Newline may be escaped for multiline command blocks. For example: taskname 3 < $ 6 > mc\ param1=value\ param2+ This executes task "taskname" indicating that the stdin (pseudofile 3) has been redirected in the parent (dummy $ filename), stdgraph (6) is to be redirected by the iraf main to the file "mc", the value of the parameter param1 is to be set to "value" (may be quoted) and the value of param2 is to be set to "yes". In the case of a task with many static parameters the CL will pass a command consisting of many lines, but it takes only one IPC packet for the entire command block hence all the parameters are set with no additional expense. ----------------------------------------------------------------------- *** N O T E *** None of this changes the semantics of CLIO in the slightest and the changes are all internal. Note however that as soon as the revisions are made in the CL all executables will have to be relinked before they can be used with the new CL, or a syntax error will occur due to the multiline command blocks. ----------------------------------------------------------------------- pkg/cl/exec.c Modified task execution via pr_connect() to support parameter cacheing. The startup command is now a multiline command, hence all old executables must be relinked. I expect that runtime parameter cacheing will be a significant optimization, noticeable even for tasks with only a half dozen or so arguments. pkg/cl/cl.x pkg/cl/main.c Modified to eliminate the reference to the LIBC procedure "c_main", used to provide the ONENTRY procedure to gain control during process startup. I moved the ONENTRY into cl.x instead, elimating the funny backwards linking used formerly. The new configuration is straight- forward and should be portable. At some point I would like to replace this construct by an emulation of the K&R C main, i.e., with argc,argv. This would not be difficult and would make it much easier to port C programs to the IRAF environment. XC and MKLIB, for example, could be nice UNIX/C programs (actually iraf programs internally) callable from either the host system or the CL. This has been attempted in the VMS version of softools but we do not have a good interface yet. pkg/cl/main.c There has been a bug in the system for some time now that goes like this: [1] interrupt, nothing happens; [2] interrupt again, CL complains about unknown task `xmit', goes through error recovery; [3] IPC protocol is trashed and syntax error occurs at next attempt to run a task in the affected subprocess, [4] flpr and get write to IPC error from child. I suspect this is due to the way PRPSIO (pseudofile i/o) handles the input buffer; the buffer pointers are not updated until sometime after the "xmit" command was first read, and an interrupt in between states causes the XMIT directive to be read as data by the CL. The fix was however not to modify PRPSIO, but to call F_CANCEL on the t_in and t_out to the subprocess in ONINT when X_INT is sent to the subprocess. pkg/cl/prcache.c In pr_connect, the call to fprintf to format and output the startup command to the subprocess was replaced by calls to fputc and fputs. Fprintf has a builtin limit on the size of a %s string which could be a problem in this application (due to the long list of parameter assignments). Also, the flexibility of fprintf is not needed and fputs is more efficient. pkg/cl/eparam.c Recall that in raw mode, no output processing is performed hence newlines are not converted into crlf. Modified the repaint code to output \r\n instead of \n (switch to raw mode could also have been used). pkg/cl/history.c This code was assuming "ehistory" whenever a command beginning with "ehi" was entered, regardless of the characters following the "ehi". Hence a command such as "ehinit" would cause ehistory to be run instead. Changed to do a rigorous match. Any abbreviation is permitted, including "e", hence ehistory is a very privledged command indeed (I guess that is what was intended, else the normal minimum match would have been used instead). sys/ki/* lib/chars.h The network node delimiter was changed from @ to !, in response to a request from STScI. Either metacharacter will do just as well, but ! is a bit easier for a touch typist to type, is used for similar purposes in other networking software, but is not used in filenames on any of our major host systems. sys/fio/fntgfn.x Removed the code for the !...! escape notation. The decision has been made to not implement filename escapes using this notation. Conventional backslash escapes will be used instead; the code for this is already largely in place and all we really need to do is test it. pkg/system/edit.cl Tried to write a version of the editor script which uses the new array stuff to manipulate filename strings. I entered the declaration char fname[80] to define a character array for holding filename. Turns out the CL understood an array of 80 strings, each of some default length. This is supposed to be compatible with SPP? (see notes on string variables earlier). Dropped out into the CL to try some things interactively, since it was clear I do not understand the new features. Typed a "cl" to push a new context on the stack; made some declarations; when I typed "bye" I expected the variables to go away but they did not. Should have. Tried it again with integer decls and those went away at the bye. Cannot redeclare a variable within a block; I understand why but should be able to. First decl or so worked but then I got a segmentation violation in strcmp from paramfind. Did not look into it, perhaps it is a bug I introduced into the CL. Repeated the experiment somewhat differently and it worked fine. Logged out and back in to initialize things (ugh). Typed "char ss" to get a char array. Tried "ss = 'abcdef'; =ss" which worked as expected. Tried "=ss[3]" and it printed the entire string, rather than just one char. Also, if the [3] is a string rather than char index, it should have given an out of bounds error. Perhaps it was my own undoing (although I don't think I changed much that could cause such problems), but even without the bugs I think we have a serious problem here. We need to resolve the syntax and semantics of these new features before they are released for people to use. Here it is: char fname[80] int i1, i2 while (fscan (file_list, fname) != EOF) { # Determine starting and ending indices of filename. i1 = 1 for (i2=1; fname[i2] != 0; i2=i2+1) if (fname[i2] == '!') i1 = i2 + 1 fname = substr (fname, i1, i2) # Call host system editor. (wanted to use fname[i1] # in arglist but didn't dare). print ("!!", editor, " ", osflags, " ", fname) | cl } (and I still haven't figured out how to strip node prefixes from pathnames in the editor). ---------------------------------------------------------------------------- Snapshot of system sent to STScI ---------------------------------------------------------------------------- dev/graphcap Set the multifile count parameter MF to one for the dicomed. The default MF count (16) was causing overflow of the scratch area since the dicomed frames can be very large. pkg/cl/main.c pkg/cl/history.c A bus error would occur during CL logout when attempting to close the logfile. Removed the "ifdef ST" stuff in main.c (when !ST close_logfile was being closed without the logfile name argument, causing the crash). Also added error checking and a warning message to the open logfile code in history.c, in case the logfile cannot be opened. sys/os/zfioks.c [UNIX] The kernel server device driver was modified to post a handler for X_IPC and X_INT during a read or write on a KS channel. If one of these signals occurs while reading or writing a KS channel the signal is caught and converted into an i/o error (status ERR) on the channel. sys/os/zfioks.c [UNIX] The KS driver was further modified to use the user login to authenticate on a remote node (formerly a fixed public login was used). The login name and password are taken from a file ".irafhosts" in the user's home directory on the local node. If this file is not found a similar file in dev$ is read to obtain the name of a public login. The user may instruct the system to query for the password rather than risk placing their password in their .irafhosts file, even though the .irafhosts file would normally be read protected. The format of the .irafhosts file is as follows: alias1 alias2 ... aliasN : loginname password (etc.) If the host alias "*" is encountered the login name and password given in that record are used. If the password "?" is encountered the actual password is obtained by querying the user terminal (this will fail for a batch job). Comments lines and blank lines are permitted. Since this mechanism is implemented at the OS device driver level it is considered to be machine dependent, although it expected that a similar mechanism will be used on non-UNIX IRAF hosts. sys/fio/fwatio.x Modified to mark the file buffer empty if an error occurs on a file write. If this is not done then close() will try again to flush the buffer during error recovery, possibly causing error recursion. sys/ki/kbzard.x Fixed a typo, ks_awrite was being called where ks_await was intended. sys/fio/awaitb.x The fill code, used to pad reads out to an integral number of chars, was being executed even if zawait returned ERR. sys/ki/kbzcls.x sys/ki/ktzcls.x Modified to not return a close status of ERR if the error bit is set on the kernel server channel. Close is called during error recovery, and close would post a second error before the error message associated with the original read or write error was output. sys/ki/ksawait.x In the event of an i/o error on the kernel server channel, was setting the error bit on the channel when it should have been calling ki_error() instead. Because of this the error recovery designed into the KI was not being exercised. sys/ki/kghost.x The syntax of the host name table was changed to a better form which is also closer to the form of the irafhosts file. Formerly the format of a line was node@server_startup : alias1 alias2 ... aliasN This was changed to alias1 alias2 ... aliasN : node@server_startup This is more logical since the server startup string is logically a function of the node names. It is also simpler because the startup string is delimited by EOL rather than colon, hence it is no longer necessary to escape colons to include them in the startup string. (Rebuilt the image display task and tried it over the ethernet - worked as ) (advertised. Speed is only slightly slower than a direct connection. ) ******************************************* *** EDIT *** *** decent editor interface surfaces !! *** ******************************************* pkg/cl/builtin.c pkg/system/system.cl Since the old edit.cl script converts vfn's into pathnames, and pathnames now include nodenames, it is no longer possible with the current script to edit files not in the current directory. The edit script is glacially slow in any case, and offers a poor interface to the host editor. EDIT has been changed to a builtin task with no parameter file. The editor name is taken from the environment. The argument list is a list of strings. Each argument string is filtered through FMAPFN to convert VFN's into OSFN's, and then appended to the OSCMD string with a leading space. Most non-filename strings are unaffected by the filename mapping, hence system dependent arguments can be passed through to the host editor. Most importantly, this interface allows the editor to be called up about as fast as an OS escape or even a host command language call. To edit a list of files (note each file is a separate argument, unlike most iraf commands): cl> ed file1 file2 file3 To edit via a template: cl> ed *.x Note that this template is merely a string to the editor interface, i.e., it is expanded by the host editor rather than by IRAF. Filename mapping is passing the character * on through unchanged - it is not a VFN metachararacter. This is fast and (at least on UNIX) is a much better interface to the host editor, since the host editor will retain its context when called once to edit a set of files, but will lose its context when called N times to edit N files. pkg/cl/builtin.c A new task BACK has been added to the language package. BACK is used in conjunction with CHDIR (aka cd) to return to the previous default directory. BACK;BACK toggles between two directories. sys/libc/cfpath.c Added C_FPATHNAME to the LIBC interface (req'd by BACK). NOTE -- It would be useful to be able to define "cd" as "chdir", "pd" as "back", "pwd" as "cd;back", etc., etc., in one's login.cl file. Given that we now have string pushback in FIO (ungetline) and general hashed symbol table facilities (SYMTAB), little work remains to implement DEFINE in the CL. Note that SYMTAB maintains its database in a position independent form suitable for passing to bkg CLs. BUT -- The semantics of the CL define facility, in particular the scoping rules, have yet to be worked out. pkg/system/system.cl pkg/system/help/* pkg/help (defunct) Moved the HELP directories from pkg$ to pkg$system. Merged x_help.e into x_system.e. Hence, one executable now contains all of the system tasks required by the typical user. The intention is that the x_system.e process be locked in the process cache upon login. HELP, DIRECTORY, TYPE, PAGE, etc. will then be perpetually available without having to do a process spawn (although on a busy system considerable pagein may be required if the process has been idle for a while). TODO -- Rework HELP to use the new SYMTAB package to maintain the help database. This will greatly speed up the database search, the main efficiency bottleneck remaining in HELP. TODO -- All VMS IRAF process should have a large page fault cluster (and large working set size) to decrease pagein time, due to the large text segments involved. Shared libraries would be even better... lib/login.cl Added commands to the default login.cl file to spawn the system proc and lock it into the process cache (leaving 3 open slots). Also cached pfiles for directory, help, type, and page. In combination with runtime parameter cacheing and the fast edit this should produce a noticeably faster system. NOTE -- The MKIRAF host system task should use (does use on UNIX) this default login.cl file as input. When the new version of the system is released we should request that all users run MKIRAF to insure that all the new goodies are functional and that the user's old parameter files do not cause problems. When installing a new version of the system it is possible for a user's obsolete parameter file to test as newer than the package parameter file, since the user's pfile is updated continually. This can cause a task to abort myseriously with no obvious connection to the pfile. pkg/system/doc/*.hlp Changed all manual pages to reflect the recent change to command mode as the default input mode of the CL. The SYNTAX of the command in the USAGE section is given formally using parens, commas, and square brackets (for optional args). This is arguable; perhaps the parenthesis and commas should be dropped from the syntax. Only the positional arguments are shown in the command syntax. The EXAMPLES are given in command mode and typical abbreviations are permitted. (The System seemed noticeably more responsive when editing all these (help files.) NOTE -- The new manual pages in the language package are written in style inconsistent with that of the rest of IRAF (not SDAS). They should either be brought into conformance with the rest of the system, the standard manual page format should be amended, or some combination of the above. NOTE -- All manual pages for the IRAF applications packages will be updated within the next couple of weeks. BUG IN CL PARSER While not strictly speaking a bug, this is a serious problem. The CL recognizes "for", "next", etc. as keywords regardless of their position in the input text. Use of these strings as parameter or task name abbreviations or as data strings causes an EXTREMELY subtle syntax error, which most users will be unable to cope with. Perhaps some additional context sensitivity in the lexical input is required. pkg/cl/history.c Using the existence of the "ehinit" environment variable to trigger automatic editing of the recalled command is incorrect. If the user does not wish to verify and/or edit when using the normal history mechanism, they should still be able to define EHINIT to set the parameters for EHISTORY. The solution was to change the history code to autoedit recalled history records only if "ehinit" is defined and the switch "verify" is included in its value string, e.g., set ehinit = "nostandout eol verify" If autoedit is not desired the verify keyword may be omitted, or alternatively "noverify" may be substituted. pkg/cl/history.c A long standing bug in the history mechanism was fixed. The bug would cause a history record to be trashed every now and then; the record would print as garbage and would be unrecoverable. The cause of the bug was the fetch_history() function, which was unaware that it was reading from a circular buffer. sys/gio/cursor/prpsio.x Fixed a horrendous bug in PRPSIO that I am sure has been there for months (the bug is still present in the old system; I have been working at the system level for ages so was not aware of it). The bug would cause the CL to hang up during error recovery following an interrupt of a subprocess. It would be necessary to interrupt twice and then flush the process from the cache, because the IPC protocol would be trashed. BUG (unresolved) Twice now the CL has gotten into a funny state where it can no longer map any filename. This bug is NOT present in the old system. I have not been able to trace the bug down yet (it does not happen very often), but it is quite serious as the user must log out when it occurs. The symptoms are misleading; it appears to be a bug with FILEWAIT but in fact it is more serious. pkg/softools/make.cl Changed to set IRAFDIR rather than IRAFLIB, since the makefiles now build all pathnames using IRAFDIR. BUG (unresolved) Once again, when a subprocess died, the CL (which was reading from the process) was fed garbage input in the form of a single letter duplicated many times. The lexical analyzer, given this input, tries to build a very long identifier token and overflows its token buffer, corrupting memory. This is awkward to fix since the standard LEX lexical analyser does not do bounds checking. The best solution is probably to not use LEX, but use a hand crafted lexical analyzer for compute mode (we are already using one for command mode). YACC is useful, but LEX is generally not worthwhile for production software. NOTE: Infinite error recursion occurs when a parser error occurs causing error restart. The ERRLEV variable in main.c, used to detect error recursion, is cleared just before the parser is called under the assumption that error restart has completed. I have yet to find any surefire way to detect and prevent error recursion. sys/os/zfmkdr.c [UNIX] The MODEBITS are improper for a directory file. Replaced with a mode value suitable for a directory file on UNIX. ---------------------------------------------------------------------------- Snapshot of system sent to ST. Identical snapshot (minus binaries) saved in lyra!/tmp2/tody (for diffs). ---------------------------------------------------------------------------- sys/mtio/mtgetpos.x The pointer "drive" was not being dereferenced as "Memc[drive]" in two procedure calls. sys/ki/kopdpr.x The node name of the second filename argument "bkgfile" was not being stripped before the call to zopdpr to spawn a detached process (e.g., bkg cl) on the local node. sys/mtio/*.x,Makelib sys/os/zfiomt.c sys/ki/(kiextnode.x,kzopmt.x,kignode.x) The revised MTIO was tested and a number of bugs were found and fixed. Since I already had to do that I went ahead and changed the drive name syntax to node ! mt... The explicit test for a "mt" prefix to determine if a file is a MTIO file or not (as is currently done in many of the DATAIO programs) is now considered a violation of the MTIO interface. A new integer function MTFILE has been added to the interface to perform this test without need to build explicit knowledge of the filename syntax into an applications program. YES|NO = mtfile (filename) -------------- 26 Aug 85 sys/gio/cursor/prpsio.x When reading 0 chars from a stream, would return 0 as function value. Changed to return EOF when 0 chars are read, since that is what the caller expects. sys/mtio/mtalloc.x,mtdealloc.x These routines were not dereferencing the pointer "drive". -------------- 27 Aug 85 sys/ki/kzwrmt.x Was calling zawrbf when it should have been calling kzwrmt. sys/etc/(sysid.x,gethost.x,Makelib) Added GETHOST; high level interface to zghost. Added SYSID, used to format labels for various type of output, e.g., plots. Returns a string which identifies the version of IRAF, the user, the host machine, and the time and date. SYSID should be used routinely to label science plots and other output. sys/etc/envscan.x Envscan would truncate the value string at the first whitespace encountered, even if the value string was quoted. This bug has been lying there undiscovered since the code was first written. sys/fio/fmapfn.x sys/fio/vfnmap.x Modified FMAPFN to return a legal host system filename if the file resides on the local node, i.e., the node name is stripped from the osfn if the file is local. FPATHNAME should be called if an absolute pathname, including node prefix, is desired. The packing action of vfnmap was a nuisance here. Vfnmap was split into two procedures, vfnmap (unchanged) and vfnmapu (vfnmap w/o packing of the osfn). Fmapfn now calls vfnmapu to get an unpacked OS filename. (this packing and unpacking is of course a mess - everything should be unpacked until the kernel is called). [This bug was discovered when trying to edit with "ed lib$*.h" - the ] [resultant unix command was "vi lyra!/tmp3/unix/lib/*.h" which would ] [have worked were it not for the node prefix. ] --------------------- 28 Aug 85 sys/mtio/mtfilspec.x The generalized magtape name syntax turns out not to be quite upward compatible with the orignal syntax. An old name such as "mta1600" must be given as "mta.1600". There is no good reason to subject the users to such a forced change in the syntax, hence mtfilspec.x was changed to be compatible with the old syntax while still permitting general drive name strings. The new rule is: [1] if the first character following the "mt" is a letter, the drive name is a sequence of letters delimited by the first nonalpha; [2] if the first character following the mt is not a letter, it is taken to be the drive name delimiter character and characters are accumulated until the delimiter is reached. The result is the obvious syntax, e.g.: mta, mta1600, mta1600[1], (etc.) # local node or node!mta, node!mta1600 # remote node sys/os/doc/os.hd An error in the pathname for the logical directory doc was preventing access to the OS manual pages via help. USAGE NOTE - When you recompile the help database you must flush the system task (containing HELP) from the process cache before the changes take effect (the old database may still be in a buffer in the process memory). sys/fio/fwtacc.x The file wait code was not correctly testing (via FINFO) for a file's permanent access permissions before entering the wait loop to wait for access to the file. lib/finfo.h Added defines for the file permission bits. sys/clio/clcache.x If a task has no pfile the CL will send its command line arguments on in the command line passed to the IRAF main using the parameter names "$1", "$2", etc. Modified parameter cacheing to support this type of parameter to avoid unnecessary CL queries. ------------------- 29 Aug 85 sys/imio/imrdpx.x sys/imio/imwrpx.x The computation of the max physical pixel index, used to check for an out of bounds reference, was in error. -------------------- 30 Aug 85 sys/imio/db/idbpstr.x A ficticious image header parameter "IM_NAXIS" was being set when IM_NDIM was clearly intended. sys/tty/ttyputs.x AND was being called with an argument of type CHAR. ----------------- 3 Sep 85 ------------------------------------------------------------------------------- TODO * TODO * TODO * TODO * TODO * TODO * TODO * TODO * TODO * TODO * TODO ------------------------------------------------------------------------------- This is a continuation of a bug report given earlier. I plan to fix this sometime in Sep-Oct before releasing the new system (on UNIX) for general use. The bug is the syntax error caused by use of reserved keywords in expressions and argument lists. This happens with "gcur", "for", "next", "min", "max", etc., etc. The solution is to add context sensitivity to crackident(), the function used to turn idents into keywords. When the parser recognizes a statement it will push a new context and the lexical analyzer (crackident in this case) will change its actions accordingly. BUG (cl) =gcur aborts with an `EOF while evaluating expression' when EOF is returned in a cursor read. Should just return "EOF". I fixed this already in the old system and will transfer the fix to the new CL. ---------------- 08 Sep 85 sys/fio/awaitb.x sys/fio/await.x This is not really a bug, but a logical inconsistency. AWAIT, when called after an AREAD has completed, is supposed to zero fill the last char of data if the number of bytes read was not enough to fill a char. The filling operation was being performed in AWAITB instead of AWAIT, but it does not make any sense for the byte oriented routine to be zero filling chars. The zero fill code was moved from AWAITB to AWAIT. ----------------------- 17 Sep 85 lib/fio.h fseti.x prpsio.x Raw mode to the terminal from a subprocess via the CL was not working, it was never being turned on in PRPSIO. Added code to turn raw mode on and off to PRPSIO. Moved define for RAWOFF from fseti.x to fio.h so that it could be used both in fseti.x and prpsio.x. (tested) sys/gio/gki/gkigetwcs.x sys/gio/cursor/gtrctrl.x Discovered that APPEND mode (gopen) was never tested. In append mode the subprocess reads the world coord information back from the CL via the PSIOCTRL pseudofile. gki_getwcs had a bug; I also changed the operation to return the WCS data without a header on the process CLIN, which is easier than trying to return a SETWCS instruction via the pseudofile, as the GIO specs state. Only system code uses the GETWCS instruction so this mod should not affect any existing software. ----------------------- 20 Sep 85 sys/etc/main.x The main was calling clc_mark once and then calling clc_free after each task call to free storage, without doing another mark. This is illegal because clc_mark calls STMARK which creates a marker in the symbol table. STFREE frees this marker; calling STFREE again without another STMARK causes whatever was at the marker position to be used as if it were a marker, causing the symbol table pointers to be corrupted. The solution was to add a call to mark after the call to free in the main, e.g., call clc_free (clc_marker) call clc_mark (clc_marker) ------------------------ 27 Sep 85 pkg/softools/mktags.x A new task MKTAGS was added to the softools package. This task will find all procedures declared in a set of files, printing either [1] a sorted listing giving for each procedure the name, source file, line in the file, and the declaration, or [2] a "tags" database file, used by VI to edit procedures rather than files. ------------------------ 8 Oct 85 sys/clio/gexfls.x This file was moved from GIO to CLIO to avoid a library conflict. The object module is referenced by code in libsys but formerly it was archived in libex, which is searched before libsys at link time. sys/tty/ttyputs.x PUTC was being called to write the value returned by the integer function AND, causing an integer argument to be passed to a procedure expecting a short integer argument. Changed the PUTC to PUTCI. ------------------------ 9 Oct 85 sys/vops/absu.x sys/vops/Makefile sys/vops/vops.men sys/vops/vops.syn sys/vops/ak/Makelib Added a new vector operator ABSU for block-summing a vector (like a block average but not normalized). pkg/cl/config.h The size of the CL dictionary was increased from 15000 to 20000 due to dictionary overflow in a large script. pkg/cl/exec.c The parameter cacheing code would pass the ")param" literal value of an indirect parameter, rather than the value of the referenced parameter. Changed to not cache parameters which are indirect. pkg/cl/opcodes.c When testing the previous bug fix the CL crashed in debug mode due to an invalid OT_STRING test inside an if(cldebug) block in procedure o_inspect. Changed to ((p_type&OT_BASIC)==OT_STRING) which should fix the problem. ----------------------- 13 Oct 85 sys/ki/irafks.x The kernel server would die on a segmentation violation when it received a garbage datagram, due to lack of error checking on the p_sbuflen parameter used to unpack the text buffer. Added a min(SZ_SBUF to prevent this. sys/fio/fredir.x Added a subprocedure FREDIRO which is like FREDIR, but redirects a stream onto an already opened stream. (10/23) sys/os/zoscmd.c [UNIX] Minor changes. Will print error message if stream cannot be redirected. Uses fork instead of vfork when possible for a little more speed. Moved some code to before the fork in the hopes of making the apparent ZOSCMD related bug go away; I cannot find anything wrong with the code nor make it fail in tests. sys/os/zpanic.c [UNIX] Panic shutdown messages from bkg jobs are now echoed on the console to provide a permanent record. sys/os/zmain.c [UNIX] SIGTSTP is now ignored in bkg processes. Hopefully this will prevent bkg jobs from being suspended when the CL is suspended. sys/os/zmain.c [UNIX] Bkg processes are now spawned at a reduced priority. sys/os/zoscmd.c [UNIX] The local files of the parent were being inherited by the child (SH), causing OS commands to fail unpredictably due to no more file descriptors. Added FCNTL calls to close files >=3 if exec succeeds. (10/31) sys/osb/f77pak.f OP was not getting set if the loop terminated without seeing the EOS. (11/4) ---------- Thu Nov 1 21:23:16 MST 1984 1. CL Modifications A new version of the system has been released on all three 4.2 BSD UNIX vaxes at NOAO. The system has undergone major revisions, primarily to remove UNIX dependencies in preparation for the port to VMS. A number of enhancements have also been made in the process. Those modifications or extensions of most interest to the user are summarized below. 1.1 Process Cache: The process cache is the key to good interactive response. While the operation of the process cache is completely automated and need not be understood, the sophisticated user can obtain improved interactive response by tuning the cache. prcache show what is in the cache prcache task [,task] lock a task in the cache flprcache flush cache (except locked tasks) flprcache task [,task] unlock a task and remove from cache Where "task" is the name of a logical task, e.g. "directory". When task A is present (locked) in the cache all tasks present in the same physical process as task A are also cached (locked). If one locks too many processes in the cache deadlock will occur. Flushing a task from the cache initializes the process and may fix certain classes of bugs. The process cache is no longer flushed when the current directory is changed, when the environment changes (except when a redefinition is uncovered by a bye), or when a task aborts. 1.2 Background Jobs: Background jobs are submitted by appending & to the command line, as has always been the case. There is a limit on the number of bkg jobs that can be running at any one time (currently 4). cmd & run command "cmd" in the background jobs show status of background jobs service service a query from the last bkg job submitted service N service a query from job N kill N [,M,...] kill job number N A bkg job needing service (waiting for parameter input) will timeout and abort after a set period of time (currently 3 hours). 1.3 I/O Redirection: Is now compatible with the cshell, e.g., >& to redirect both the standard output and the standard error output, ditto |& and >>&. The standard error output of an OS escape issued from a script task or subprocess is now redirected properly, hence output from XC and MAKE may be spooled. 1.4 OS Escapes: !command Now sends "command" to the shell defined by the UNIX environment variable SHELL or to /bin/sh if SHELL is not found. The cshell user can therefore now issue ! cmd >& spool & from within the CL. !!command Always sends the command to the Bourne shell (/bin/sh). 1.5 Timing Tasks The cpu and clock time consumed by a compiled task may now be measured by prefixing the task name with a $, e.g.: set | $sort | $match tty | $count The times given are measured by the IRAF Main and hence do not include the overhead of process initiation if it occurs. This feature is currently only available for external compiled tasks. It is not currently available for script tasks, intrinsic functions, or builtin tasks. 1.6 Package Loading and Logout The original package loading protocol has returned without the drawbacks of the original. When you type the name of a package the package will become the "current package" (first package searched when looking for a task), the prompt will change to reflect the new current package, and a "bye" will be necessary to return to the package. It is no longer necessary to issue a whole series of "bye" commands to log out of the CL, just type "logout". packages name the packages currently loaded ?? print menus for all loaded packages ?sys print menu for (e.g.) package 'sys' A package is "loaded" by typing its name, making the tasks therein known to the CL. Any number of packages may be simultaneously loaded and referenced without continually changing the "current package". Programmers: use the new "clbye" in package script tasks in place of "cl" if the "cl" is the last command in the file, i.e., if there is no epilogue. This saves a file descriptor. The function of "clbye" is otherwise completely identical to that of "cl". 1.7 The Null File, Standard Streams IRAF now has a null file, known as "dev$null". This is useful for discarding the output of commands. The null file may be referenced anywhere an IRAF filename may be used. The special files "STDIN", "STDOUT", and "STDERR" are also quite useful. These reference the standard i/o streams of the task in which they are found but may be used wherever an ordinary filename would be used, including in compiled programs. 2. C Runtime Library 3. New Process Control Facilities 4. Environment Facilities 5. Error Recovery 6. File I/O Modifications 7. Formatted I/O Modifications 8. Bit and Byte Primitives 9. Vector Operators 10. Kernel Modifications 11. Configuration Control Procedures ### Revision 1.1 Wed Oct 31 23:29:14 MST 1984 ------------ From: /u2/valdes/iraf/ Fri 13:49:39 02-Nov-84 Package: system Title: WIDSOUT: dataio and multispec packages The task 'toiids' in the multispec package has been moved to the dataio package and renamed 'widsout'. It converts IRAF images to IDSOUT text format. It is complementary to 'ridsout'. ------------ From: /u2/valdes/iraf/ Fri 13:54:22 02-Nov-84 Package: plot Title: _imaxes The task _imaxes, used in scripts dealing with the sizes and number of dimensions in images, has been moved from the plot package to the images package. The plot package automatically loads the images package now. The reason for this change is that there are possible uses for _imaxes in scripts, particularly in the imred package, which also use many of the images tasks but rarely need any of the plot tasks. ------------ From: /u2/valdes/iraf/ Fri 13:58:48 02-Nov-84 Package: images Title: imcopy and imtransform The tasks imcopy and imtransform have been generalized to allow multiple input and output image templates. The parameter names are input and output respectively. Each image in the input list is matched, in order, with each image name in the output list. If the input image and output image are the same then the operation is performed to a temporary file in tmp$ and, when the operation is complete, the temporary file replaces the input image. Image sections are allowed in the input list and are ignored in the output list. For simple uses these tasks work just as they did before (except for the change in the parameter names). This system is also used in imlinefit. A help entry is now available for imcopy. ------------ From: /u2/valdes/iraf/ Fri 14:05:18 02-Nov-84 Package: generic Title: new tasks colbckgrnd and chimages The new tasks colbckgrnd and chimages have been added. Colbckgrnd is a companion to linebckgrnd. They apply imlinefit to determine and subtract a line by line or column by column background for the specified images. They supply a generalized form of a function available on the CYBER. The task chimages is a script application of imcopy and imtransform suitable for in place modifications of images. This task can trim, flip, rotate, transpose, and subsample a list of images. The default function can be set by the user to always perform the same function on the images (such as a standard trim and rotation). Help entries are available for these and other generic tasks. ------------ From: /u2/valdes/iraf/ Tue 13:13:00 06-Nov-84 Package: images Title: shiftlines A new task, shiftlines, is available in the images package. It shifts image lines by a specified amount. There are five choices for the image interpolation and two choices for boundary extension. See the help page for further information. ------------ From: /u2/davis/ Thu 16:30:26 08-Nov-84 Package: curfit Title: curfit mods A piecewise linear (SPLINE1) polynomial fitting function has been added to the CURFIT package. A new routine CVSET can take coefficients derived external to the CURFIT package and store them in the curve descriptor for use by the CVEVAL and CVVECTOR routines. ------------ From: /u2/davis/ Thu 16:35:58 08-Nov-84 Package: local Title: Imsurfit installed in local A new task, IMSURFIT, which fits surfaces to IRAF images has been added to the local package. See manual pages for more documentation. ------------ From: /u2/valdes/iraf/ Fri 10:12:52 09-Nov-84 Package: multispec Title: New parameter in findpeaks A new parameter, edge, was added to the task findpeaks. This parameter specifies a minimum distance to the edge of the image for peaks found. ------------ From: /u2/tody/ Sun 21:29:59 11-Nov-84 Package: cl Title: parameter indirection, enumerated values Two new features affecting parameter access have been added to the CL, parameter indirection and enumeration of the legal types of a string valued parameter. 1. Parameter Indirection Parameter indirection allows a parameter to take any combination of its value, minimum, or maximum fields from another parameter. Indirection is signalled by an '@' prefixed field in the parameter file, e.g., order,i,h,@param specifies that the value of parameter "order" is the value of the parameter "param", which is located by the usual search path (task, package, cl). More general forms include order,i,h,@task.param and order,i,h,@pack.task.param.field The taskname of the current package my be specified by the shorthand name "_", e.g.: order,i,h,@_.order states that the default value of the task parameter "order" defaults to the same value as the package parameter of the same name. Minimum and maximum fields may be specified in the same fashion, e.g.: alpha,i,h,5 beta,i,h,10,@alpha Indirection may be overridden permanently via an assignment statement or temporarily on the command line. 1.1 Caveats Indirection requires string storage for the p_value, p_minimum, and p_maximum fields of a parameter. String storage for these fields is allocated only if indirection is specified in the parameter file (except for the value field of a string parameter which is always allocated string storage of SZ_FNAME chars). In general a parameter may not be made indirect after loading the parameter file. Multiple indirection is permitted. Infinite recursion will occur if a field of a parameter indirectly references itself. A field of a parameter may indirectly reference another field of the same parameter. Implementing parameter indirection was surprisingly difficult and involved substantial modifications to the CL. The most common cases have been tested but further bugs caused by these modifications can be expected to turn up. 2. Enumerating the Values of a Parameter The legal values of a string type parameter may now be enumerated in the parameter file. For example, device,s,a,"mta","mta|mtb|tty" or order,s,a,"5","3|5|7" The CL will accept only values enumerated in the list found in the p_minimum field. Minimum match abbreviations are permitted, with the CL always returning the full value of the matched substring. While enumeration of values is currently supported only for string valued parameters, integer values may be stored as strings (with restrictions on the use of such a parameter in CL expressions) as shown in the example above. ------------ From: /u2/tody/ Sun 22:23:42 11-Nov-84 Package: cl Title: parameter file names in UPARM The underscore has been omitted from the parameter file names stored in the UPARM directory. Old parameter files can no longer be referenced and the contents of the UPARM should therefore be deleted. ------------ From: /u2/tody/ Sun 22:26:34 11-Nov-84 Package: cl Title: command mode versus compute mode in the CL A new lexical analyzer has been added to the CL to permit a more terse mode for entering simple commands. The CL now distinguishes between two distinct modes of command input, known as "command mode" and "compute mode". 1. Command Mode Command mode is useful when interactively entering commands because it is the most terse mode of input. There are few special characters, most strings do not have to be quoted, and command arguments are delimited by whitespace rather than commas. Command mode is permitted only when working interactively, i.e., it is turned off when running a script task or an external compiled task. The features of command mode are summarized below: o Unquoted strings may contain any nonwhite character except "'\(|><{};=#. The characters !^&+- may be embedded in strings but are recognized as the OS escape, history metacharacter, etc. when the occur in the obvious context. o Whitespace delimits arguments, not comma. Commas may be embedded within argument strings. For example, delete file1,file2,file.* v+ would delete all files matching the template given as the first argument with verification enabled. o The sequence \c is replaced by the character c. A lone \ at the end of a line causes continuation on the next line. Quotes may also be used to escape characters but a quoted string is a distinct token (e.g. argument or value). o The OS escape need no longer appear at the beginning of an input line of text, e.g., sequences such as "time;!w" are now permitted. Newline may be escaped when building up long OS escape commands. When not used to escape newline the escape character is left alone, i.e., \c equates to \c. An unescaped newline marks the end of the command. o A string constant appearing on the righthand side of an assignment statement need not be quoted. Only the simple assignment statement is recognized in command mode (i.e., += etc are not recognized). The command mode syntax is consistent with the that of many command languages, including the UNIX shell and cshell, and VMS DCL. 2. Compute Mode The original command form of the CL is now known as "compute mode". In compute mode arguments must be delimited by commas and general expressions are recognized. Compute mode is automatically entered in any of the following circumstances: o Within parentheses. Compute mode expressions may be embedded within command mode commands. o Within braces. Command mode is turned off in a compound statement, e.g. within the body of a while loop. o Within scripts or when executing an external compiled task. Command mode may not be used within a script; the old compute mode scripts need not be changed. o When the CL parameter "lexmodes" is set to NO. This is currently the default since command mode is not fully tested and the documentation all pertains to compute mode. 3. CL Syntax The grammar (syntax) of the CL is the same for both command and compute mode. Mode switching is implemented by two separate lexical analysis filters. The lexical filter to be used depends on the runtime context and is selected upon entry to a command block. The only change to the grammar of the CL made to implement command mode was the elimination of the need to quote a string constant appearing on the right side of an assignment statement. String constants appearing in assignment statements within braces ({}) must be quoted regardless of the lexical mode of the CL. Conversely, a parameter name appearing on the righthand side of an assignment statement outside of a block must be enclosed in parenthesis to be interpreted as a parameter name rather than as a string constant. This change affects any existing CL scripts which assign a parameter to a parameter outside of a brace delimited block. Within a block the syntax and lexical form of the CL is that of a conventional programming language. ### Revision 1.1 Mon Nov 12 04:42:52 MST 1984 ### Revision 1.2 Tue Nov 13 04:10:28 MST 1984 ------------ From: /u2/tody/ Wed 10:42:07 14-Nov-84 Package: system Title: help for the system package Thanks to Richard Wolff, manual pages have been installed for the tasks in the system package. ------------ From: /u2/tody/ Wed 20:46:43 14-Nov-84 Package: plot Title: graph mods The x-clipping feature lost during the switch to the new system has been restored. If xautoscale is turned off and ux1,ux2 are set to points within the plot, no data is plotted to the left or right of the window. The default value of the parameter "title" has been changed from "@imtitle" to simply "imtitle". The @ was causing parameter indirection with the result that graph could not find the parameter "imtitle". This points out a serious weakness of the parameter indirection scheme; the value of a parameter may no longer begin with an @ as a normal character. On the other hand it is nice to be able to respond to a query with @param. This will have to be resolved at some point, hence the current scheme for implementing parameter indirection must be regarded as experimental. We are committed to having the capability, but how it is implemented may change. ------------ From: /u2/tody/ Wed 20:54:07 14-Nov-84 Package: cl Title: bug fixes The bug involving INDEF's in parameter files has been fixed. The new MAGNIFY task and others like now work properly again. Beware that if you ran any of these tasks while the bug was active, the parameter file may have to be unlearned. ------------ From: /u2/tody/ Wed 20:56:30 14-Nov-84 Package: system Title: allocate, deallocate bug fixes The system tasks allocate and deallocate had two bugs related to the system modifications made last weekend. They should be working properly now. ------------ From: /u2/tody/ Wed 22:57:26 14-Nov-84 Package: cl Title: cl bug fixes The following CL bug fixes have been made. Examples of statements that formerly did not work are shown. Bugs in the new lexical analyzer (lexmodes=yes): [1] spawning background tasks (& at end of line), e.g., make >& spool & [2] what is a number, e.g., x = .5 [3] the ? and ?? tasks work again. ? anywhere but at the beginning of a statement is not special, i.e., ? can be embedded in strings. Assignment statement bugs: [1] both of the following would formerly fail: s1 = [*:16,-*] and s1 = identifier Other: [1] An invalid CD or CHDIR is no longer passed to a child process. This turned out to be a kernel bug rather than a CL bug, but should not affect processes other than the CL. ------------ From: /u2/tody/ Tue 22:02:07 20-Nov-84 Package: fio Title: freadp, fwritep added The two i/o procedures FREADP and FWRITEP have been added to FIO to support direct access by IMIO of pixel data in the FIO buffers. These new procedures are useful in operations where performance is critical, but usage is a bit tricky hence their use in applications code is not recommended. charp = freadp (fd, offset, nchars) charp = fwritep (fd, offset, nchars) Both routines cause the FIO buffer to be filled with the file buffer sized file block containing the portion of the file starting at "offset" and extending for "nchars" chars. The referenced segment must fall entirely within a file buffer. An error action is taken if offset+nchars falls at a point outside the file buffer, or if the segment referenced is beyond BOF or EOF. A char pointer to the first char of data is returned as the function value. The referenced data must be used before FIO reuses the file buffer for a different region of the file. Successive calls to data segments lying within a single large buffer, i.e., which do not result in a file fault, cost about 180 microseconds per call for FREADP and 240 microseconds per call for FWRITEP on the VAX 11/750. No seek is required since the file offset is included in the call. ------------ From: /u2/tody/ Tue 22:19:22 20-Nov-84 Package: plot Title: plot expansion in GRAPH The modifications to GRAPH made to permit proper autoscaling in Y when specifiying an X subrange of data to be plotted were lost in the disk crash over the weekend. This mod has been repeated and plots should be properly scaled when GRAPH is used to expand a portion of a plot. ------------ From: /u2/valdes/iraf/ Thu 17:37:39 06-Dec-84 Package: images Title: Enhancements to imarith Imarith has been greatly enhanced. The enhancements are in performance, flexibility, and use of lists for operands and resultant images. Performance has been enhanced by providing type specific operators to perform the operation in any desired datatype. If the calculation datatype matches the operand and result pixel datatypes then no datatype conversions are performed and the operation is particularly efficient. The flexibility enhancement allows the user to select the pixel datatype of the resultant image(s) and the calculation datatype(s). Specially recognized datatypes allow defaulting to the datatype of highest precedence from the operand images or to the datatype of either operand. The list enhancement allows either operand to be a list of images and constants (they can even be mixed). The result parameter can also be a list of images. The number of operand elements can be either one (in which case it is used repeatedly) or match the number of elements in the result list (in which case the elements are match one-to-one). It is also possible to have the result image match one of the operand images (apart from sections). In this case the operation is performed to a temporary image and, when completed successfully, the temporary image replaces the operand image being overwritten. There are options to print the operations being performed as diagnositics and to print the operations without performing the operation for debugging. ------------ From: /u2/valdes/iraf/ Thu 17:47:55 06-Dec-84 Package: images Title: Enhancements to imcopy Imcopy has been enhanced to use general image template lists. The number of input images can be either one (in which case the input image is repeatedly copied to each element of the output list) or or match the number of output images (in which case the image lists are matched one-to-one). If the input image and output images are the same (except for sections) then the copy is performed to a temporary image and, when completed successfully, the temporary image replaces the original input image. ------------ From: /u2/tody/ Fri 18:09:13 07-Dec-84 Package: imio Title: optimized version of IMIO installed Testing of the optimized version of IMIO has been completed and the new interface installed. The optimizations improve the efficiency of line by line i/o, the most common type if i/o to images. In short, the overhead required to access the pixels will be minimal provided no image section is given, datatype conversion is not necessary, there is only one input line buffer, boundary extension is not in use, bad pixels are not being set to INDEF, etc. If all these conditions are true IMIO can skip a lot of code and the pointer returned by a get/put line or section call will point directly into the FIO buffer, eliminating a memory to memory copy. If any of the fancy IMIO options are used additional expense will of course be incurred. There is little difference in the timings for the integer and real datatypes. All modifications to IMIO were internal to the interface, hence only relinking is necessary to take advantage of the increased efficiency provided by the new interface. The IMAGES and TV packages have been relinked. Loading a 512**2 type short image into the image display now takes only 2-3 cpu seconds in the simplest case (ztran=none). Clock time for this image load operation is still a minimum of 10-12 seconds due to the limitations imposed by the UNIX file system. Image arithmetic operations typically require several cpu seconds per operation. The inefficiencies of the UNIX file system will be overcome when a low level IRAF static file driver is implemented for 4.2BSD UNIX in coming months. ------------ From: /u2/tody/ Fri 19:04:42 07-Dec-84 Package: imio Title: image sections Minor changes have been made to the image section code in IMIO to make the section notation more deterministic and to remove unecessary restrictions on the use of sections. A section of the form [5] no longer refers to the highest physical dimension of the image; this was nondeterministic and required advance knowledge of the dimensionality of the image. In the new release of IMIO the number of dimensions referenced in an image section need not match the number of physical dimensions in the image. If too few dimensions are specified in the section then the higher dimensions will be set to 1. If too many dimensions are specified then the higher dimensions must be set 1 explicitly or it is an error. For example, if A is a 2-dim image, A[10:20] is equivalent to A[10:20,1], i.e., the section is 1-dim. If B is 1-dim B[*,1] is a legal section but B[*,2] is not. ------------ From: /u2/tody/ Fri 19:12:41 07-Dec-84 Package: libsys Title: error recursion A bug has been fixed in etc$xerfmt.x which would cause error recursion when error recovery occurred under certain circumstances. The error recovery code was calling ENVGETS, which takes an error action under certain cirumstances, causing error recursion. The call to ENVGETS was replaced by a call to ENVFIND which is functionally identical but does not query or abort if it cannot find an environment variable. ------------ From: /u2/tody/ Sat 14:41:37 08-Dec-84 Package: lprint Title: bug fix The bug which caused LPRINT to output garbage following the normal text has been fixed for the second time. Evidently this bug fix was lost in the disk crash in November. ------------ From: /u2/tody/ Sat 14:44:32 08-Dec-84 Package: libsys Title: process control bug fix Error checking was missing in PROPEN, used to open a connected subprocess. This would cause a segmentation violation in the CL when trying to open a nonexistent executable process file. ------------ From: /u2/hammond/iraf/ Mon 10:58:21 10-Dec-84 Package: dataio Title: wtextimage installed Task wtextimage has been installed in the dataio package for converting IRAF images to text files. The text file written consists of an optional FITS header followed by the pixel values. ------------ From: /u2/tody/ Mon 17:16:57 10-Dec-84 Package: cl Title: flprcache bug fix The flprcache (flush process cache) task had a bug that caused it to flush only the top process from the cache. The corrected function is to flush all processes not locked in the cache (locked processes can be flushed if a flag is set to break the locks). This bug may have been responsible for background jobs leaving dreg processes running after termination of the bkg CL. ------------ From: /u2/tody/ Wed 10:33:48 12-Dec-84 Package: images Title: new package IMDEBUG A new subpackage IMDEBUG has been added to IMAGES. This is a slightly cleanup version of the image i/o debugging tasks formerly stored with the IMIO source. Tasks MKTEST and CUBE are particularly useful for testing image operators. No param files or manual pages are presently available for these routines. ------------ From: /u2/tody/ Wed 17:45:55 12-Dec-84 Package: IMIO Title: bug fix The new (optimized) version of IMIO had a bug which was not discovered and fixed until today. The bug would cause image sections to be ignored (in effect). ------------ From: /u2/tody/ Wed 17:48:47 12-Dec-84 Package: images Title: IMCOPY can now mosaic images IMCOPY has been modified to permit use of an image section on the output image, which must be an existing image if a section is used. The input image or image section will be copied into the output section. Both sections must be the same size. This facility is particularly useful for mosaicing images, i.e., placing several small images into a larger image to form a single large image. ------------ From: /u2/tody/ Tue 15:35:41 18-Dec-84 Package: libbev Title: bug fix in regres.f The Bevington library procedure REGRES had an error in the Fortran source (a * that should have been a **) causing the coefficient errors to be computed incorrectly. The bug has been fixed and the task POLYFIT in local relinked to use the new version of the library. ------------ From: /u2/tody/ Thu 21:09:02 20-Dec-84 Package: fio Title: filename mapping bug fix The FIO filename mapping code had a bug which caused the mapping file (used for long filenames) to be corrupted when deleting a file with a name too long for the host OS. The mapping file would be unreadable with a checksum error following the deletion. The file vfnmap.x was modified in two places to fix this bug. This bug was discovered by the guys at STScI in the process of testing the new VMS version of IRAF. ------------ From: /u2/tody/ Thu 21:12:16 20-Dec-84 Package: system Title: obscure bug in the IRAF Main The IRAF Main, when passed a command using i/o redirection on the command line, would include the newline in the filename used for redirection. The code which extracts the filename was modified to recognize newline as a filename delimiter. ------------ From: /u2/tody/ Mon 13:42:05 24-Dec-84 Package: fio Title: bug fix in the filename mapping code The filename mapping package had a bug causing the mapping file to be left open when opened for writing but never modified. Procedure vfnclose in file fio$vfnmap.x was modified to close and unlock the mapping file in this case, as well as when it is modified and has to be rewritten. ------------ From: /u2/tody/ Mon 15:20:45 24-Dec-84 Package: spp Title: portability problems with INDEF Equality comparisions of the form if (pixval == INDEF) ... involve a comparison of two floating point numbers for equality and hence may not be portable to some machines or to some compilers. The following macro has been added to the SPP language definitions to make such constructs more portable: bool = IS_INDEFt (value) where the "t" stands for one of the standard type suffixes. One such macro is provided for each data type, e.g., IS_INDEFS, IS_INDEFR, and so on. The macro IS_INDEF with no type suffix is equivalent to IS_INDEFR. All programs which make equality comparisons involving indefinites should eventually be converted to use these new constructs. ### Revision 1.3 Mon Dec 24 22:13:31 MST 1984 ------------ From: /u2/valdes/iraf/ Thu 10:58:27 27-Dec-84 Package: local Title: imreplace Imreplace has been changed in several ways: 1. There is no output file. The replacement operation is done in place. 2. The input images may have sections to select the part of the image to be operated upon. This eliminates the need for the column parameters. 3. The lower and upper limits are hidden parameters. 4. The operation is more efficient. ------------ From: /u2/tody/ Wed 19:30:38 02-Jan-85 Package: fmtio Title: procedure CTOR added A new procedure CTOR has been added to FMTIO. This procedure merely calls CTOD and is provided for the sake of convenience, to avoid having to deal with double precision numbers in simple conversions. ------------ From: /u2/tody/ Thu 18:54:36 03-Jan-85 Package: images Title: magnify parameter file replaced The parameter file for the task MAGNIFY somehow got filled with zeros. I copied the version of the file from /u2/valdes/iraf/images into pkg$images. ------------ From: valdes$iraf/ Fri 17:42:46 04-Jan-85 Package: images Title: magnify Magnify has been fixed. The bug dealt with changing delta = (x2 - x1 ) / (npix - 1) to delta = (x2 - x1 + 1) / npix ------------ From: /u2/tody/ Fri 21:06:39 04-Jan-85 Package: system Title: new system functions Two new system functions, LOCPR and LOCVA, have been added to the ETC package. These integer functions are more convenient to use than the subroutines ZLOCPR and ZLOCVA, and they eliminate the need to call the kernel routines directly in the higher level packages. Use of these new functions should probably be restricted primarily to system code. ------------ From: valdes$iraf/ Thu 15:25:53 10-Jan-85 Package: images Title: imstatistics When using a large number of pixels the accumulation of the sum of the pixel values and the sum of the squares of the pixel values may suffer from roundoff error. This causes the standard deviation to be zero when the true standard deviation is less than a percent of the mean. This has been improved by changing to accumulation in double precision. ------------ From: /u2/tody/ Fri 12:38:59 11-Jan-85 Package: system Title: removal of filename restriction in IRAF Main The IRAF Main formerly restricted filenames in the context task > filename to only those filename beginning with a letter. This prevented the use of pathnames such as ../filename. The restriction has been removed and now the filename may contain any characters except whitespace, newline, or the redirection characters. ------------ From: /u2/tody/ Mon 10:43:09 14-Jan-85 Package: system Title: GIO installed The new GIO (graphics i/o) interface has been installed in a system library. Up to date documentation is available in gio$Gio.hlp. GIO is functional in the sense that the library is complete and programs written to use GIO can be compiled and run. The CL has not yet been modified to interactively interpret GIO metacode, however, so GIO can only be used at present to spool metacode into a file. Metacode produced in this fashion can be decoded with the GKIDECODE task in "gio$stdgraph/x_kernel.e" for debugging purposes. The task STDGRAPH in the same process is a mockup of the stdgraph kernel, and may be used (noninteractively) to make crude plots on the terminal from GIO metacode. ### Revision 1.4 Tue Jan 15 04:57:57 MST 1985 ------------ From: valdes$iraf/ Wed 17:45:04 16-Jan-85 Package: images Title: new output type for imlinefit A new output type for imlinefit has been added. The output type is called normfit. It outputs the fitted line normalized to have a unit mean. This is used to generate calibration corrections; for slit profiles in longslit spectroscopy in particular. Also the output pixel type is now always real. ------------ From: /u2/tody/ Wed 18:15:15 16-Jan-85 Package: cl Title: bug affecting list structured parameters A bug fix has been made to 'paramset' to permit setting the filename of a non-string datatype list structured parameter. This bug was the source of the problems experienced earlier this week with NORMALIZE and NORMFLAT (generic package). ------------ From: /u2/tody/ Mon 23:09:30 21-Jan-85 Package: system Title: error recovery bug fixes I spent several hours tracking down several of the "bugs" that have been annoying us when interrupting subprocesses. Specifically, a task will now stop spewing out output immediately when interrupted, and interrupt should no longer get stuck off. Software modifications: [1] etc$main.x Added an fseti call to cancel STDOUT during error recovery. [2] cl$pfiles.c Failure to write pfile was a fatal error for some reason; changed it to an internal error. [3] cl$main.c Calls added to the ONINT interrupt handler for the case interrupt of a subprocess: c_fseti to cancel stdout c_prredir to redirect stdin, stdout, stdgraph to 0. This causes etc$prfilbuf to ignore pseudofile directives buffered in IPC at the time the interrupt occurs. [4] fio$fseti.x The F_CANCEL code in fseti formerly did an LNOTE, LSEEK before and after clearing the buffer, to preserve the file offset. Removed these to force iop to point to beginning of buffer if i/o cancelled. [5] cl$prcache.c If a process was interrupted during startup IPC would be trashed, causing endless "IRAF Main: unknown task name" errors from the subprocess. Make the call to c_propen in cl$prcache.c a critical section to prevent interrupt of the child process until startup is completed (a critical section is a section of code which is uninterruptable). The effect of this is noticeable: a ctrl/c typed immediately after entering a command which causes a process to be spawned will be ignored. In the process of fixing this one I discovered a flaw in the exception handling code, which is patterned after the UNIX "signal" facilities. The idea was that one could call xwhen (signal, newhandler, oldhandler) .... xwhen (signal, oldhandler, junk) with some code in between, and the oldhander would be restored by the second call. This does not work if the code in between calls XWHEN for the same signal. [6] os$zxwhen.c A bkg process is spawned with SIGINT disabled. ZXWHEN would check to see if the signal was disabled when posting a new one. If the signal was disabled it would ignore the posting and leave the signal disabled. The bug occurred when ZXWHEN was called to disable the signal temporarily to protect a critical section; once turned off a handler could never be posted. Added a "first_time" flag to make the test only in the first call to SIGNAL for SIGINT. Zombie processes are still occuring. I found and fixed a bug some time ago which may have solved this problem, but many tasks have yet to be relinked and hence the fix has not been propagated. I will relink the whole system when the visitors currently using the system finish their reductions. ------------ From: valdes$iraf/ Tue 09:04:28 22-Jan-85 Package: images Title: New task blkaverage A new task specifically for block averaging two dimensional images has been added. ------------ From: valdes$iraf/ Tue 09:07:32 22-Jan-85 Package: images Title: Bug fix in sigl2 A bug in the sigl2 package for image block averaging and linear interpolation has been fixed. The partial blocks at the ends of the image were not being truncated so that the code was trying to read outside the image. This package is used in magnify and display. ------------ From: valdes$iraf/ Thu 17:33:02 24-Jan-85 Package: local Title: fixpix A new task called fixpix has been added to the local package. It replaces pixels in a set of bad pixel regions by linear interpolation from nearby pixels. The bad pixel regions are specified via a file (which may be STDIN) by lines of four coordinates consisting of the first and last columns and the first and last lines of the bad region. The direction of interpolation is determined by the shape of the region; i.e. it interpolates across the narrower dimension of the bad region. This task should be very useful in imred type applications for detectors with consistent bad full and partial columns and lines. ------------ From: /u2/tody/ Thu 20:25:18 24-Jan-85 Package: system Title: more bug fixes Many bug fixes have been made in the past two days to make the system more reliable, particularly where error recovery is concerned. A summary of the software revisions follows. [1] fio$xerputc.x Would overrun buffer if called repeatedly without a newline terminator. Modified to flush buffer when it fills, even if newline not seen. Buffer overrun was supposed to be impossible, but it turned out it could happen during error recursion. [2] etc$erract.x More logic was added to improve the heuristic for detecting error recursion. [3] cl$builtin.c The CLERROR procedure was modified to pop the last task off the control stack. Both BYE and ERROR signal task termination. It is necessary to pop the task to keep KILLTASK from sending X_INT to interrupt a task which has already terminated, if the task is an external executable. [4] os$zxwhen.c (UNIX kernel) The SIGTERM signal was getting blocked after the first exception, causing the process to become uninterruptable. For some reason UNIX identifies the signal as SIGINT in the argument list of the exception handler, even though it is the SIGTERM signal which was caught. The handler would unblock SIGINT rather than SIGTERM, leaving SIGTERM blocked. [5] cl$errs.c The ONINT cl interrupt handler is now reposted whenever CL_ERROR is called. This is necessary as CL_ERROR might be called while interrupts are disabled in a critical section. Additional logic was added to detect error recursion. The tests for an error occurring during process startup or shutdown were moved to the beginning of the procedure. [6] etc$main.x Modified to block interrupts during process startup, to prevent loss of sync of the IPC protocol. Added F_CANCEL calls for the standard streams during error recovery. Added a call to repost the standard exception handlers during error recovery (this will reenable interrupts if error recovery takes place while interrupts are blocked). [7] cl$exec.c The test to prevent "bye" from logging out of the CL was not working due to a precedence error in the test for a bkg CL. The CL will once again print the "use 'logout' ..." message if one accidentally types "bye" in the CL menu. [8] cl$main.c The ZDOJMP jump buffer of the IRAF Main is now restored when the CL main exits, in case an error occurs during process shutdown. A bug was fixed which would cause the LOGGINGOUT flag to be cleared when EXECUTE was called to process the logout.cl file. New routines INTR_DISABLE, INTR_ENABLE were added to make it convenient to protect critical sections of CL code. [9] cl$prcache.c The STDGRAPH stream of a task is now initialized to STDOUT by default when an external task is connected. [10] cl$pfiles.c Writing a parameter file is now a critical section to prevent learning of truncated parameter files. [11] cl$bkg.c The command to be executed in the background is now saved in the bkg file header to permit spying on what the task is doing with the octal dump program. Use PS to get the pathname of the bkgfile and then "od -c" to print the first few bytes of the file. [12] cl$bkg.c Various critical sections were identified and protected. The filenames of the bkg query files are now generated and deleted at bkg job spawn time, just in case there are any dreg files left over in uparm from a previous bkg job. If this were to happen JOBS might print a misleading status for the job. [13] fio$ delete.x, falloc.x, fmkcopy.x, frename.x, filopn.x All code segments that have the VFN database opened for writing are now critical sections, and all error checking in these sections is done explicitly. This is necessary to prevent an interrupt from corrupting the mapping file, to ensure that the lock on the mapping file is removed, and to ensure that the mapping file is closed by the process so that the same or another process does not block waiting for access to the file. [14] fio$rename.x RENAME will now copy the file and delete the original if the ZFRNAM (move) operation fails, allowing files to be renamed to different filesystems. [15] os$zoscmd.c (UNIX kernel) Will now trap interrupts and kill the OS command and then the parent process. ------------ From: /u2/tody/ Sat 19:27:06 26-Jan-85 Package: os Title: zfsubd smartened up The ZFSUBD kernel procedure, used to add a subdirectory to a pathname to produce the pathname of the subdirectory, has been smartened up to back up a pathname when the subdirectory is given as "..". Thus, if the current directory is "/iraf/sys/fio/" and the subdirectory is "..", the output pathname will be "/iraf/sys/" rather than "/iraf/sys/fio/../". The second form is a legal UNIX pathname equivalent to the first, but it is less efficient to evaluate and would lead to long pathnames and eventual overflow of a filename buffer. ------------ From: /u2/tody/ Sat 20:31:59 26-Jan-85 Package: fio Title: filename mapping bug discovered A filename mapping bug (file fio$vfntrans.x, procedure vfn_translate) has been discovered and fixed. The bug would cause sometimes cause a subdirectory name to appear twice in the mapped filename. I discovered it when the filename ".." was mapped to "../..", but the same bug could be expected for ordinary subdirectory names as well, provided they were recognizable as subdirectory names and appeared as the last field in a filename. ------------ From: /u2/tody/ Sat 20:41:08 26-Jan-85 Package: imio Title: modification date was not getting initialized A bug causing the image modification date to not be initialized at image create time has been discovered and fixed. The symptom was that the modify date would show up as zero (=1970). Tests for invalid minimum and maximum pixel values would consequently fail since an invalid limits time would also be set to zero, which would not test as older than the modify date. ------------ From: /u2/tody/ Sat 20:44:07 26-Jan-85 Package: images Title: new task MINMAX installed A new task named MINMAX has been added to the IMAGES package. The new task is used to compute or recompute the mininum and maximum pixel values in an image or set of images. The computed values may be printed, used to update the image header, or accessed from a CL script. ------------ From: /u2/tody/ Mon 21:26:43 28-Jan-85 Package: imio Title: error recovery bugfix in IMMAP The IMMAP procedure would not always close the header file when taking an error action, e.g., when the file opened turned out to not be an image header. ------------ From: /u2/tody/ Mon 21:30:26 28-Jan-85 Package: images Title: prettyprinting of title strings in IMHEADER The IMHEADER task will now strip trailing newlines and whitespace from title strings, printing the final string with a single newline. This results in a consistent looking title regardless of the DATAIO reader used to create the image (some readers blank fill to 80 columns, others add an unneeded newline). ------------ From: valdes$iraf/ Wed 14:52:04 30-Jan-85 Package: images Title: imsum replaces imaverage The task imaverage has been replaced by the task imsum. This new task provides several improvements. 1. The sum of the input images is provided. 2. The average of the input images is provided. 3. The median of the input images is provided. 4. The output pixel datatype is selectable and defaults to the appropriate highest precedence datatype of the input images. 5. The calculation type is selectable and defaults to the appropriate highest precedence datatype of the input images. 6. When all images are of the same datatype and the calculation is done in the same type then this operation is particularly efficient. ------------ From: /u2/tody/ Thu 19:13:33 31-Jan-85 Package: images Title: new task IMGETS added The new task IMGETS has been added to the IMAGES package. The new task is used to get parameters by name from the image header. The function of the task is not expected to be greatly affected by the upcoming DBIO mods. IMGETS returns the value of the parameter as a string in the parameter "imgets.value". To coerce this string into some other datatype, e.g., int or real, reference it in an expression with the either the "int" or "real" type coercion function (as in F77), e.g.: cl> imgets image HA cl> = real (imgets.value) 5.67544 cl> Note that the DATAIO readers tend to write the instrument parameters in the header with upper case names, and you must get the case right. Use IMHEADER to list the image parameters if there is a problem. The following standard image header parameters may also be accessed with IMGETS: name datatype pixtype string ndim int len[1], len[2], etc. int min real max real title string pixfile string ------------ From: /u2/tody/ Thu 19:21:34 31-Jan-85 Package: images Title: task IMHEADER modified The IMHEADER task has been modified to print the "user area" of an image header, if nonempty. The user area should be formatted as a string buffer containing newline delimited entries of the form "keyword = value" for this to work properly. ------------ From: /u2/tody/ Thu 19:58:17 31-Jan-85 Package: imio Title: changes to IMMAP The IMMAP procedure in IMIO has undergone the following modifications to support IMGETS, etc: [1] When opening an existing image or a new image, space is always allocated for a minimum size user area of 2880 struct units, regardless of the value given in the argument list (a larger space is allocated if the argument is larger than 2880). The purpose for allocating the extra space is to permit automatic reading of the user area regardless of the size given in the argument list, in case the image is subsequently referenced in a NEW_COPY mode call to IMMAP. [2] IMMAP will no longer abort if the size of the user area is different than that specified in the argument list. The choice of 2880 struct units is arbitrary but since this code is expected to be replaced shortly, I did the easy thing. The changes should not affect any existing code. ------------ From: /u2/davis/ Fri 09:03:49 01-Feb-85 Package: dataio Title: rcamera mods RCAMERA has been modified to read old 9-track camera format where tape files contained multiple images. A new parameter image_list has been added so the user can specify which images to extract from a file. The camera header parameters are now stored in the user area. ------------ From: /u2/tody/ Fri 18:22:40 01-Feb-85 Package: vops Title: modifications and additions to the VOPS package The following VOPS procedures have been modified: [1] ASOK - selection of Kth smallest element in a vector. Calling sequence changed to eliminate the internal vector copy, modifying the single vector argument in place. The procedure was tested and a bug found and fixed. [2] AMED - median of a vector Calling sequence changed to eliminate the internal vector copy, modifying the single vector argument in place. Calls ASOK. The cases npix=1,2,3 were optimized. [3] ALTR - general linear transformation of a vector Calling sequence modified to add a new constant making the transformation completely general, i.e., call altri (a, b, npix, k1, k2, k3) b[i] = (a[i] + k1) * k2 + k3 Examination of the optimized assembler source for the routines actually installed in the library showed that the installed routines named ALTRS and ALTRR were actually performing the function of AMAPS and AMAPR. The names of the assembler versions were changed to AMAP. A search of all IRAF source files showed that the modified vector operators were used only in the DISPLAY program, which was also modified. The following new vector operators were added: [1] ALTA - linear transformation via add and multiply A variant of ALTR: b[i] = (a[i] + k1) * k2 [2] ALTM - linear transformation via multiply and add A variant of ALTR: b[i] = (a[i] * k1) + k2 [3] AMED3 - vector median of three vectors Calculates the vector median of three input vectors, i.e., each point in the output vector is the median of the corresponding three points in the input vectors. [4] AMED4 - vector median of four vectors Calculates the vector median of four input vectors. [5] AMED5 - vector median of five vectors Calculates the vector median of five input vectors. ------------ From: valdes$iraf/ Fri 18:36:01 01-Feb-85 Package: images Title: Task linefit replace imlinefit The task imlinefit has been replaced by an improved version call linefit. The improvements are: 1. Ability to specify the output columns. 2. Pixels rejected from the fit may or may not be replaced when computing the output image lines. 3. The fitted points may be subsampled by binning them and taking the average of the column positions as the ordinates and the median as the abscissas for each bin. 4. The task efficiency has been somewhat enhanced. Greater efficiency enhancements are expected shortly. The tasks linebckgrnd, colbckgrnd, lineflat, and colflat in the generic package have been changed to use the new task. ------------ From: /u2/tody/ Sat 14:18:03 02-Feb-85 Package: cl Title: recent minor modifications The following modifications were recently made to the CL: [1] It was discovered that SCAN and FSCAN would recognize comma as a field delimiter in addition to whitespace. While this might occasionally be useful, it prevents the scanning of lists wherein the fields contain embedded commas, e.g., image sections. The scanner has been modified to recognize only whitespace as the field delimiter. Todo: recognize quoted string fields. Add formatted scan capability to permit more complex scans. [2] The iofinish procedure (exec.c), called when a task finishes, would take an error action if an attempt was made to close one of the standard streams. This interferred with the use of the special "files" "STDIN", "STDOUT", and "STDERR". The checking has been removed. The checking was put in back when the CL was interfaced to UNIX, which will physically close a standard stream if you tell it to. The IRAF i/o system works differently and the checking is no longer necessary. ------------ From: /u2/tody/ Sun 15:58:00 03-Feb-85 Package: os Title: interrupting a command sent to the OS The ZOSCMD kernel procedure has been modified in an attempt to fix the bug that occurs sometimes when "edit" is interrupted. Interrupting a command like "edit" at just the right point would sometimes leave both the foreground CL and the forked child CL reading from the terminal at the same time. The theory is that this occurs when the FORK system call is interrupted, a definite possibility since a fork can take a while for a process as large as the CL. Interrupts are now disabled before the fork and reset in the child process before the execl. The parent CL is now uninterruptable upon entry to the fork and until the child exits. I have given up on the attempt to kill a tree of child subprocesses when an interrupt occurs as there does not seem to be any way to do this in UNIX (e.g., a kill of a "make" command entered from the CL). It is easy to kill the process at the root of the tree, but the children of that process continue to run. Using process groups it is possible to kill all processes in the tree but unfortunately the CL and all processes except the CSH are killed too. Perhaps examination of the CSH source will uncover a way to do this properly. ------------ From: /u2/tody/ Sun 16:11:45 03-Feb-85 Package: softools Title: mklib buffers changed The MKLIB buffers have been increased by a factor of two to accommodate larger libraries. The recent additions to VOPS overflowed the old buffers (which were set to hold 1000 object modules). ------------ From: /u2/tody/ Sun 19:10:34 03-Feb-85 Package: tty Title: logical device names The TTY package has been modified to permit logical device names containing optional subunit, frame, etc. information in addition to the root TERMCAP or GRAPHCAP name for the device. This feature is most useful when any of several units of a device may be accessed, all of which have the same termcap or graphcap entry. Thus, for example, we might have devices "versatec.1" and "versatec.2". The format and contents of any subfields are device or installation dependent (the extra fields are passed on to and interpreted by a kernel). All we require is that the fields be set off from the root device name by a special character. To make such distinctions possible the root device name may contain only characters chosen from the set [a-zA-Z0-9_+_]. ------------ From: /u2/tody/ Mon 22:25:48 18-Feb-85 Package: libc Title: bug fix to FILBUF in the C runtime library I discovered that a section of code originally put into the FILBUF procedure in the C runtime library for debugging purposes was never removed. Two lines of code were deleted and the library updated. rev ------------ From: /u2/tody/ Thu 02:33:09 21-Feb-85 Package: os Title: new devices added Two new logical printer devices were added to kernel.h, "vup" and "vdown", used to select either the upstairs or downstairs versatec printers. The names are the same as used in the UNIX commands vpr, lpr, etc. A new plotter device named "dicomed" was added to the system to facilitate debugging the NSPP kernel, although the host system metacode translator is not yet installed. ------------ From: /u2/davis/ Fri 15:22:12 01-Mar-85 Package: curfit Title: curfit mods The CURFIT package has been revised. The routines should now run faster. A new routine cvacpts has been added (see help facilities). ------------ From: /u2/tody/ Tue 20:14:02 05-Mar-85 Package: system Title: new task GRIPE installed; BUGMAIL deleted. Following a suggestion from STScI (no doubt under the influence of AIPS) I have replaced the old, little used BUGMAIL task by a new task named GRIPE. GRIPE writes into the file "doc$gripesfile", which anyone can read, and in addition sends each gripe directly to a designated IRAF person, currently me, to insure that the GRIPE is read promptly. Please feel free to post gripes using the new task to ensure that they get recorded properly. ------------ From: valdes$iraf/ Thu 11:47:13 14-Mar-85 Package: images Title: Changes to imcopy Imcopy has been changed as follows: 1. The output name may be a directory in which case the input images are copied to the directory. 2. A verbose option has been added to print each copy operation as it occurs. 3. The choice of copying one file to several output files never really worked and is now not allowed. Thus, the input is a list of images and the output must be a list of matching images (the number of input and output images must be the same) or a single directory. This task now is similar to the file copy except that the file copy does not allow matching lists of input and output files. The help page has been updated. ------------ From: valdes$iraf/ Tue 18:11:20 19-Mar-85 Package: xtools$imt.x Title: image template enhancement Tasks which accept image templates (most of those in images and longslit) may now concatenate strings and templates to form new image names. This is particularly useful for forming new output image names based on the image names in the input image template. The concatenation operator is the double slash (//). Some examples are: image -> image image* -> image1, image2, etc. image*[1,*] -> image1[1,*], image2[1,*], etc. image*//.1 -> image1.1, image2.1, etc. new//image*//.bck -> newimage1.bck, newimage2.bck, etc. image01\[0-5] -> image010, image011, ..., image015 To test the image template expansion use the task "sections". This change is effected in the imt.x procedures in the xtools package. ------------ From: /u2/tody/ Tue 21:35:46 19-Mar-85 Package: osb Title: string conversion between fortran and spp Two subroutines have been added to the OSB package (the bit and byte primitives) for converting to and from SPP and Fortran strings. The new routines are named F77PAK and F77UPK. Usage is very similar to the existing STRPAK and STRUPK, used to convert to and from SPP and C strings. Unfortunately there is no standard specifying the representation of strings in different languages, and all three of the IRAF languages do in fact represent strings differently. f77pak (sppstr, f77str, maxch) f77upk (f77str, sppstr, maxch) Note that these routines are only callable from a Fortran subprogram since there is no (machine independent) way in SPP or C to declare a Fortran string type argument (the only exception is the use of *" for Fortran character constant arguments in SPP). ------------ From: /u2/tody/ Sat 14:47:18 23-Mar-85 Package: vops Title: histogram vector operator added to the VOPS package A new vector operator ahgm[csilrd] has been added to the VOPS package. The function of this new operator is to accumulate the histogram of a series of input vectors in the integer histogram output array. ahgm_ (data, npix, hgm, nbins, z1, z2) The range of the histogram is from greyscale values Z1 to Z2. Pixels with values outside the range Z1 to Z2 are excluded from the histogram. ------------ From: /u2/tody/ Sat 17:24:55 23-Mar-85 Package: images Title: graphics added to the imhistogram task The imhistogram task in the images package has been rewritten to use GIO. The default action is now to compute and plot the histogram of an image on the standard graphics output. The histogram may optionally be output as a list. ------------ From: /u2/tody/ Sun 18:28:56 31-Mar-85 Package: cl, etc. Title: recent mods A bug in the ABS function was fixed. In recent weeks many changes have occurred in that portion of the graphics system resident in the CL, i.e., the cursor read code, workstation transformation, pseudofile i/o, subkernel control, and the stdgraph kernel. Virtually all of the revisions have been isolated to the graphics subsystem. These revisions were not documented since the graphics subsystem was actively being installed and tested. Now that the graphics subsystem is stable I will begin to record revisions to the subsystem. Nearly all of the GIO modifications/additions were in the following new directories: sys$gio GIO library sys$gio/glabax axis labelling and drawing sys$gio/gki graphics kernel interface sys$gio/cursor cursor mode, pseudofile i/o, subkernels sys$gio/stdgraph the stdgraph kernel sys$gio/nspp the nspp kernel pkg$plot task implot added soon to be added: sys$gio/gkslib gks emulation sys$gio/ncarutil ncar/gks utilities sys$gio/display generic image display kernel The CL must now be linked with the libraries libcur.a (cursor) and libstg.a (the stdgraph kernel). Changes to the CL itself were minor; the QUERY procedure in modes.c was modified to read the cursor for gcur and imcur type parameters. Minor modifications were required wherever the word STDGRAPH appeared in the old CL. CL_ERROR was modified to call C_XONERR during error recovery to give the i/o system a chance to clean things up (e.g., turn off graphics mode). In addition to the graphics directories, minor changes were required in the FIO, CLIO, and ETC directories. A new (internal) file type SPOOL_FILE was added to FIO to support the wierd i/o required by the graphics system. A new fset parameter was added to find out how man chars remain to be read in the file buffer. CLOPEN and ZARDPS in CLIO were completely rewritten to permit bidirectional i/o on the graphics streams and to add support for all three graphics streams. A new stream PSIOCTRL was added; this is used by GIO to send control instructions to connect subkernels to graphics streams, save and restore WCS, and so on. CLOPEN will now direct graphics output to the null file if a task is run standalone, thus preventing garbage from being written to the terminal in debug mode. ------------ From: /u2/davis/ Mon 16:26:44 01-Apr-85 Package: images Title: imsurfit The task imsurfit have been moved from the local to the images directory. New parameters to permit pixel rejection and region growing have been addded. The user has the option of fitting selected columns, rows, subsections or a border around the image. ------------ From: /u2/tody/ Wed 15:56:57 03-Apr-85 Package: cl Title: parameter indirection The metacharacter used to indicate parameter indirection has been changed from @ to ) to eliminate confusion with list file indirection in a filename template. For example, try setting the CL parameter "s1" to ")version", i.e., and the command "=s1" will cause the value of "version" to be printed. ------------ From: /u2/tody/ Wed 15:59:39 03-Apr-85 Package: vops Title: additions to VOPS Two new vector operators have been added to the VOPS package. The ACNV operator convolves a data vector with a kernel of the same datatype. The ACNVR operator is similar but the kernel is always of type real. acnv[silrd] (a, b, npix, kernel, kpix) acnvr[silrd] (a, b, npix, kernel, kpix) where a[] pixel array of length npix+kpix-1 b[] pixel array of length npix (output, accumulated) kernel[] convolution kernel, type pixel or real For kpix=5, b[1] would be set to b[1] plus the dot product of the 5 point kernel and the 5 points a[1:5]. The one dimensional convolution operators may easily be used to implement higher dimensional convolutions. ### Revision 1.5 Wed Apr 3 23:59:56 MST 1985 ------------ From: /u2/tody/ Sun 17:30:28 07-Apr-85 Package: system Title: miscellaneous bug fixes A number of bug fixes were made to the system following written bug reports from STScI (for the VMS port) and CTIO (for the AOS port). All bugs reported by STScI were fixed with the exception of the asynchronous file i/o bug, which will be easier to track down later when we have a system with asynch. i/o. In particular the SPP bug which caused '&&' to be replaced by 'jiand' has been fixed, hence it will no longer be necessary to use the typed functions andi(), etc. in the VMS version of the system. The same bug would have cropped up in AOS. The bugs reported by STScI are summarized in their report, which is reproduced below. I have added comments enclosed in brackets, i.e., [[ ... ]]. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ From stsci Thu Jan 17 08:30:15 1985 Received: by lyra.noao.UUCP (4.12/4.7) id AA10600; Thu, 17 Jan 85 08:30:03 mst Date: Thu, 17 Jan 85 08:30:03 mst From: stsci (Space Telescope ) Message-Id: <8501171530.AA10600@lyra.noao.UUCP> To: tody Status: R ------------------------ NOTES file for IRAF (VMS) ------------------------ Contents: 1. Bugs 2. Portability Considerations 3. Source file name changes ------------------------------------------------------------------------------- 1. Bugs /pkg/cl/bkg.c In rbkgfile(), following fix: ... fgets (set, fp) ... TO ... fgets (set, SZ_LINE, fp) ... /sys/libc/qsort.c -- qcmp() --> (*qcmp)() /sys/libc/cfilbuf.c -- filbuf() --> (*filbuf)() /sys/vops/advz.x -- '$t' missing for generic function /sys/vops/advz*.x -- procedure advz[idx...] (extra files ?) /sys/vops/ak/aif*.x -- aiftrrr.x /sys/vops/ak/aff*.x -- afftrxx.x, afftxrr.x, afftxxx.x /sys/vops/ak/acjgx.x -- acjgxx.x /sys/fio/fdebug.x -- "int and()" statement missing /sys/fio/fputtx.x -- "int and()" statement missing /sys/fio/fgetfd.x -- "int or()" statement missing /pkg/cl/pfiles.c -- pfcopyback() 'pt' was going NULL before 'pf' in a for loop check; was changed to check for 'pt' being NULL instead of 'pf' /pkg/cl/builtin.c, binop.c, bkg.c, history.c, lexicon.c, debug.c, errs.c, exec.c, pfiles.c problem getting to next task by doing 'tp++'; due to VMS alignment of structures. Fixed with a macro definition in /pkg/cl/task.h for "next_task()", namely: #define next_task(tp) \ ((struct task *)((unsigned)tp + (TASKSIZ * BPI))) #define prevtask next_task(currentask) [[ not fixed; will be fixed when CL2 is merged (dct) ]] /pkg/help/lroff/output.x -- references constant MAX_INT "include <mach.h>" statement missing /pkg/lists/tokens.x -- last_nscan() --> last_nscan SPP/RPP compiler bug -- define and jiand (e.g. in <iraf.h>) if (x == FIVE && y == SIX) ... (SPP) if (x == FIVE & y == SIX) ... (RPP) if (x .eq. 5 .jiand. y .eq. 6) ... (FORTRAN) only occurs when a macro is replaced before the '&&', in this case FIVE. [[ see spp/rpp/rpprat/pbstr.r for fix ]] /sys/fio/vfnmap.x vfnunmap() -- when an OS extension was (un)mapped to a null IRAF extension, the character count returned was not right: (e.g.) doc.dir --> doc but nchars = 7 This caused problems later on in the directory task of the system package. The fix: ... { vfn[extn_offset-1] = EOS op = extn_offset - 1 <-- was missing } ... /sys/fio/vfntrans.x filename mapping for directories sometimes doesn't work; sys$ --> drb4:[irafx.sys]sys instead of drb4:[irafx.sys] made a quick fix in zopdir() and zfchdr() to deal with this for now; can't seem to locate the exact problem in vfntrans.x [[ this was fixed sometime ago ]] /sys/fio/vfnmap.x When deleting a file and the parameter subfiles=yes, and no mapping file is currently in directory, a null mapping file is (sometimes) created. Then doing a system.directory fails when trying to read this null mapping file. E.g. cl> delete aaabbb.ccc subfiles=yes (default .par file) delete() is called with 'aaabbb.ccc' and then 'aaabbb.ccc.zsf', which is degenerate; therefore, if a mapping file doesn't exist, it is created. Haven't fixed this yet; not sure about the best way to go. Some simple suggestions: -- Could somehow figure out whether the file is being added or deleted (i.e. was vfnmap() called from vfnadd() or vfndel()?). -- Or, whenever a mapping file is created, it has the appropriate header info in it rather than being null? -- Or, some special case of reading null mapping files? [[ the mapping file is now deleted if it contains no filenames ]] /sys/etc/environ.x env_hash() return (sum**2 / CONSTANT) causes integer overflow on VMS; changed to: (for now) return ( int( float(sum) / CONSTANT * float(sum) ) ) [[ changed to sum / constant * sum ]] arithmetic overflows: /sys/etc/urand.x and others... These seem to occur in a few places, sometimes purposely. An easy VMS fix is to use "FORTRAN/NOCHECK file", but it's a pain to have to remember which files need this special treatment. [[ no action; overflow in URAND is not supposed to happen ]] /sys/fio/ Seems to be a bug in the file i/o related to writing asynchronously to binary files. In VMS, this is REALLY asynchronous, and it caused problems when trying to write the help database; making VMS do it synchronously (like UNIX) fixed the problem, for now anyway. It may be that a single buffer is being played with between the write and the wait; at least that's what it looks like in some of the binary files that were written -- there were missing blocks and misaligned data, which were there when displayed on the terminal. I don't think it's a problem with RMS, since double buffering was used to test the binary file driver, with success. [[ no action ]] /pkg/system/directory.x - call strcat ("$", Memc[template]) --> call strcat ("$", Memc[template], SZ_FNAME) - should sort and output the file name list ONLY if (nfiles > 0) otherwise, an "adjustable array error" occurs in qsort() (i.e. an array with dimension of 0 is passed to qsort, causing VMS to complain) ------------------------------------------------------------------------------- 2. Portability Considerations (life would be easier if most source files and include files did not need to be mapped ... especially C source files) -- no underscores -- only alphanumeric -- case insensitive (i.e. lower case) -- (9 chars).(3 chars) (Notes: Some source file names and references to include files had to be changed to create the filename mapping in the first place, see list below. This is not so crucial now, since XC and XPP have been written/modified to handle filename mapping. But, it still does cause problems when making bug fixes to source files with degenerate names and trying to keep track of them with RCS. Of course, 99% of the mapping problems will disappear with VMS V4.0.) [[ I am considering eliminating all nonportable filenames in the system ]] [[ directories, though not in the applications directories. This would ]] [[ make it easier to move the system software, and once the system is up ]] [[ the system itself would map the filenames in the applications programs. ]] also, directories should be case insensitive: /sys/vops/AK --> ak /sys/vops/LZ --> lz [[ subdirectory names in VOPS, IMIO changed to lower case ]] sys/libc/cfilbuf.c -- filbuf --> vfilbuf (VMS) since 'FILBUF' gets translated to 'filbuf' on VMS, not 'filbuf_', which causes a conflict: FILBUF and (*filbuf)() SPP coding conventions that cause problems for VMS FORTRAN compiler: [[ the following are all bugs, not SPP features; all have been fixed ]] a) multiple declarations, e.g. int ip ... int ip, this, that files: /sys/fmtio/ctoi.x /sys/fmtio/sscan.x /sys/fmtio/chdeposit.x /pkg/system/sort.x b) 'real' function, e.g. real (dval, 0) --> real (dval, 0.0) /sys/fmtio/gctod.x c) array declarations in procedures, e.g. procedure proc1 (array, count) type array (count) int count # should be BEFORE array declaration # (tries to use 'count' as a REAL) /sys/fmtio/patmatch.x /pkg/utilities/t_translit.x makeset(), filset() [[ all such arrays that I could find are now dimensioned ARB ]] and(), or(), not() --> andi(), ori(), noti() ??? lots of references in /sys/fio/ to these procedures (added them for VMS, using andi(),...) xor() referenced in /sys/vops/...axorki.x (added xor() for VMS, using xori()) FIO file mode "APPEND" -- create the file if non-existent?? UNIX (and now VMS) do this automatically, but system reference document isn't clear about this. Without this feature, /pkg/system/bugmail.cl and revisions.cl don't work. [[ kernel should create nonexistent file opened in APPEND mode ]] I/O drivers ZMAIN passes the EPA of the read driver to IRAF_MAIN() ... This is OK in UNIX (and VMS, for now), but may have cases where I/O is to different devices and the drivers are NOT compatible. (I don't know enough of the internals of other operating systems to say for sure that this is a problem, but it seems like it could be.) Eg. STDIN from a file and STDOUT to terminal If the file and terminal drivers are incompatible, then things won't work. We may eventually have some problems in this area with VMS, especially if we start to support some network I/O access. Eventually, we may store the actual read/write function in the channel structure and check it whenever a read/write function is called; that is, override the higher level IRAF I/O at the kernel level. [[ this is unlikely to be a problem and can be fixed at kernel level if nec ]] zardlp() function referenced in /sys/etc/lpopen.x ?? (added dummy function for VMS) [[ no action ]] modf() function used in /pkg/cl/unop.c was missing from /sys/libc/ (added modf.mar to /sys/os/ for VMS) [[ call to modf deleted; modf.mar should be removed from kernel ]] ------------------------------------------------------------------------------- 3. Source file name changes /pkg/softools/boot/spp/xpp/xpp_main.c --> xppmain.c [[ removed underscore in C file name ]] Necessary for file name mapping: /sys/memio/ty_size.dat --> tysize.dat /sys/memio/coerce.x, salloc.x, sizeof.x include "ty_size.dat" --> include "tysize.dat" [[ used "include <lib$szdtype.dat>" ]] ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ In addition, Nelson Zarate at CTIO reported the following minor bugs which I have fixed. The AOS port is only just beginning. Once again a lot of code compiled with only a few errors. [1] libc$cfredir.c redundant "int fd" declaration. [2] libc$cfilbuf.c, libc$qsort.c pointer to function returning pointer to int problem [3] softools$boot/spp/rpp/rppfor/indent.f77 local variable LOGICO redefining common variable of the same name. [4] softools$boot/spp/rpp/ratlibf/catsub.f array argument declaration out of order, as reported by ST in other parts of the system. [5] softools$boot/spp/rpp/ratlibc/ create.c,open.c reference to UNIX stdio hidden structure field _file. Replaced by use of "fileno(fp)". ### Revision 1.6 Mon Apr 8 22:27:58 MST 1985 ------------ From: /u2/tody/ Wed 10:15:40 10-Apr-85 Package: softools Title: bug in the SPP compiler A bug was fixed in the first pass of the SPP compiler as a result of the effort to get IRAF running on the Jupiter supermicro system at JPL. The bug would cause a spurious "too many strings" error to occur during the processing of a large number of include files. ------------ From: /u2/tody/ Mon 21:44:33 15-Apr-85 Package: sys Title: use of ENTRY statement considered harmful In trying to port IRAF to the Jupiter / 4.2BSD system I found that the F77 compiler on that system could not compile ENTRY statements. This is not the first time I have had trouble with compilers and entry statements, hence I was forced to conclude that use of such statements compromises the portability of our software, despite the fact that entry statements are part of the Fortran 77 standard. I have removed all the entry statements from the files in the sys$ directories (the system libraries), and am in the process of removing them from HELP and XTOOLS (the latter so that I can compile the images package). Hereafter the use of ENTRY statements in SPP or Fortran code in IRAF is discouraged. ------------ From: /u2/tody/ Thu 00:21:31 25-Apr-85 Package: gio Title: new version of NSPP kernel installed The latest version of the NSPP/GIO kernel has been installed. This version differs from the previous one primarily in the way it handles put cell array. ------------ From: /u2/tody/ Thu 00:22:52 25-Apr-85 Package: clio,gio Title: nasty bug found and fixed The nasty bug that was keeping cursor mode SNAP from working has finally been found and fixed. The recent removal of ENTRY statements caused this bug (a memory overwrite bug) to resurface as a cascade of "Warning: cannot delete file" messages when writing to STDPLOT or PSIOCTRL (bug reported by both Dyer and Suzanne). In this form it was straightforward to find and fix. The bug dates back to the addition of the pseudofiles STDIMAGE, STDPLOT, and PSIOCTRL to the pseudofile i/o code in CLIO. The parameter defining the number of pseudofiles was not increased to allow for these new files, causing a write off the end of an array. All graphics applications programs should be relinked to capture the bug fix. All such programs are currently overwriting either part of the FIO or environment list data structures, depending on how they are being used. ------------ From: /u2/tody/ Thu 11:11:36 25-Apr-85 Package: softools Title: XCOMPILE enhancement The XC task will now produce, by default, an output executable file with extension ".e", unless the output file name is explicitly set on the command line with the "-o" switch. ------------ From: /u2/tody/ Thu 22:22:50 25-Apr-85 Package: graphcap Title: dicomed device entry The graphcap entry for the dicomed device has been modifed to use the NSPP kernel, allowing it to drive the physical dicomed device. Formerly the output was directed to the gkidecode kernel since the low level code was not then working. ------------ From: /u2/tody/ Thu 22:24:51 25-Apr-85 Package: gio Title: cursor mode enhancements 1. Snap The SNAP function in cursor mode is now fully functional. The cursor mode snap makes a hardcopy of the graphics on your screen on a batch plotter, e.g., the imagen or versatec. This gives a much higher resolution plot then is achievable with a video hardcopy device, and of course snap works from any graphics terminal. The snap will work at any magnification and will capture key A style axes, T or D text and lines, etc. Syntax is :.snap [device] where [device], if absent, defaults to stdplot or to the device currently connected to the STDPLOT graphics stream, if any (type :.sh to check the status of the streams). For example, :.sn imagen makes a snap on the imagen, and :.sn is the shortest form of the command. It takes only a fraction of a second to make a snap, but it may take several minutes for it to appear on the plotter. Versatec snaps are much more expensive to generate than imagen ones due to the need to digitize the plots (but the versatec makes prettier plots). Avoid generating a lot of snaps to the versatec all at once, or you will swamp the VAX with "vplot" rasterization tasks. SNAP does a frame advance after each snap, hence there is no buffering of the last snap in the kernel, as is the case for ordinary output to STDPLOT. With the implementation of SNAP cursor mode is now fully functional. I do plan, however, to add a DISPOSE function to flush the last plot out of STDPLOT. There are also still some bugs in the cursor mode code waiting to be fixed, e.g., =gcur before any graphics does not currently work, nor does it work after a :.init. The backup and undo commands, used to discard graphics instructions, do not always erase the graphics at present (a full screen redraw will however show that the line or text string is gone). 2. Waitpage The [hit return to continue] wait is now implemented as a text mode getline rather than as a cursor read. The cursor read would cause a switch to graphics mode, blanking the text memory on some terminals (giving the user no chance to read the text). ------------ From: /u2/tody/ Thu 22:42:47 25-Apr-85 Package: gio Title: nspp kernel The NSPP kernel will now generate lines of different widths. More precisely, linewidths less than 1.5 are drawn in "low" intensity, and everything else is drawn in "high" intensity. By default, the axes and tick labels of a plot are drawn in high intensity and the data in low intensity. Variable lien widths are not currently supported for the imagen and dicomed. ------------ From: /u2/tody/ Fri 17:38:19 26-Apr-85 Package: gio Title: more bug fixes The "write to IPC with no reader" bug, which would formerly appear at logout after a write to a plotter device, has been fixed. The bug was in the IRAF Main, i.e., file etc$main.x. GIO calls ONEXIT to post a procedure which is supposed to shutdown the graphics system at process exit. The procedure posted by ONEXIT was not getting called, however, until the main called the FIO cleanup procedure, which would close down the graphics kernel rudely, causing the error messages. The main has been modifed to call the ONEXIT procedures before cleaning up FIO. Typing "=gcur" at login, before generating any graphics, now works. The graphics kernel was being opened but the open workstation command was not being sent to the kernel, hence an attempt to read the cursor would cause EOF to be returned. A new builtin task called GFLUSH has been added to the CL package language. This may be called to flush any graphics buffered in STDPLOT to the plotter. ------------ From: /u2/tody/ Sat 15:12:41 27-Apr-85 Package: gio Title: new command .gflush A new : escape ":.gflush" has been added to the cursor mode escapes. This is the cursor mode equivalent of the "gflush" command in the CL, permitting disposal of plotter output from within cursor mode w/o having to exit to the CL. ------------ From: /u2/tody/ Mon 09:40:45 29-Apr-85 Package: fmtio/str Title: string sorting utility A new procedure has been added to the STR subpackage of FMTIO for sorting arrays of strings. strsrt (strp, sbuf, nstr) int strp[ARB] # array of indices of strings pointer sbuf # pointer to string buffer int nstr # number of strings Sorting is carried out by permutation of the indices of the strings in STRP. String storage is in SBUF. ------------ From: /u2/davis/ Mon 11:05:43 29-Apr-85 Package: dataio Title: rfits,wfits mods WFITS will now store the IRAF image name in the FITS parameter IRAFNAME. RFITS will optionally restore image files to disk with the old IRAF name if the parameter oldirafname is set. A new parameter scale has been added to RFITS. If the scale switch is set to no RFITS will dump the FITS integers directly to tape regardless of the value of bscale and bzero. Otherwise RFITS will check the bscale and bzero factors before deciding whether or not to scale the data. ------------ From: /u2/tody/ Mon 22:48:03 29-Apr-85 Package: fmtio Title: bug fix in patmatch PATMATCH was found to be grossly inefficient for a match at the beginning of a line, and was optimized accordingly. ------------ From: /u2/davis/ Tue 14:31:06 30-Apr-85 Package: images Title: new 2-D shift operator A new task IMSHIFT has been added to the images package. IMSHIFT will shift an image by fractional pixel shifts in both x and y using a set of 2-D interpolation routines. See the help file for more info. ------------ From: /u2/tody/ Mon 22:36:38 06-May-85 Package: cl Title: bug fix affecting i/o redirection A bug was discovered that would cause commands of the form task < file to fail. In such a case, the CL would open "file" as a binary file rather than as a text file, causing every other character to be discarded on our machine. The bug was due to a typo in exec.c with the operator "*" typed when "&" was meant (see the code which sets the STDINB flag). ------------ From: valdes$iraf/ Wed 16:13:37 08-May-85 Package: imred Title: bias package There has been a great deal of confusion about removing bias from CCD images. This has led to use of several different IRAF operators for this purpose. In particular a contant bias subtraction and a line-by-line bias subtraction have been used. Neither of these is correct. The bias does vary (at least for some detectors) so a constant is not appropriate. However, line-by-line subtraction introduces noise because the bias variations are expected to be low frequency. After discussion with George Jacoby and Roger Davies the following approach seems to be the correct one. For this discussion assume that the bias region occupies a set of columns. Average the bias columns to produce an average bias column. Fit a function of the line number to the average bias column. Subtract this average function from each column of the image. Note that this method encompasses the constant bias and high frequency line-by-line subtraction with a suitable choice of the fitting function (i.e. a polynomial of order 1 or a very high order function respectively). I have implemented this algorithm using the general curve fit package with interactive extensions. In the imred package there is a package called bias. In the bias package are two tasks colbias and linebias. Help information is available for these tasks. Because trimming of the bias region and unwanted portions of the image is very commonly desired and can result in considerable savings in execution time these tasks provide for specifying a "trim" region to be bias subtracted and output. The tasks can be run non-interactively, interactively, or partially interactive and partially non-interactive. I would like to remove the following tasks from the imred.generic package: dcbias, btt, chimages, biassub If there are no objections I will do this in a week or so. ------------ From: /u2/tody/ Wed 23:50:51 08-May-85 Package: plot Title: additions to IMPLOT task The IMPLOT task has been modified to generate a two line plot title, the first line of which tells whether the plot is a line or column plot, gives the line or column number (or numbers in the case of an average), and gives the image name. The image title appears on the second line. The :l and :c commands now permit a second numeric argument. If two numeric arguments are given, the indicated range of lines or columns will be averaged. ------------ From: valdes$iraf/ Thu 14:27:51 09-May-85 Package: curfit Title: New stat procedures Two procedure have been added to the curfit package. These are CVSTATI and CVSTATR. They are used to get parameters from the curfit package. The parameter macro values are defined in curfit.h. The parameters currently defined are: define CVTYPE 1 # curve type define CVORDER 2 # order define CVNSAVE 3 # Length of save buffer define CVXMIN 4 # minimum ordinate define CVXMAX 5 # maximum ordinate Primary usages are to get the length of the buffer needed for a CVSAVE and to determine the properties of a curve which has been restored by a CVRESTORE. ------------ From: /u2/tody/ Thu 18:07:47 09-May-85 Package: images Title: minor mod to HEDIT The interactive verification feature of HEDIT has been slightly modified since some users turned out to be confused by the original way things were done. One can now respond to the field update query with either "yes" or "no", in addition to explicitly entering a new value string. Either "yes", "y", or carriage return serves to accept the prompted value. A editing sequence involving many images can now be terminated prematurely by typing "q" in response to the "update header" query given after all fields in a particular image have been edited. The task will exit immediately without editing the remaining images. ------------ From: /u2/tody/ Thu 22:12:31 09-May-85 Package: imio Title: new release of IMIO installed A new release of IMIO was installed in the system on 29 April. The major changes in the new version of IMIO are summarized below. [1] IMIO now has the ability to perform (optionally) automatic boundary extension to satisfy out of bounds pixel references. [2] A preliminary database interface has been added to IMIO. Image headers are now variable length and use only the amount of disk space required to store the header (excluding byte packing). [3] Two new database utility tasks HEDIT and HSELECT have been added to the IMAGES package. Both use the new library subroutine EVEXPR, now installed in the FMTIO package. The new release of IMIO is upward compatible with previous versions and should require no changes to or even recompilation of existing code. The basic image header structure is unchanged hence existing images and image headers are still accessible. Copying of old images still on disk with IMCOPY may however be desirable to reduce disk consumption (the old headers were wasteful of storage). This release of IMIO introduces some database tools and concepts which should help us understand the role the DBIO interface and DBMS package will play in image processing applications in the near future. The current database interface has serious functional limitations and is inefficient for operations which operate upon entire databases (e.g., the select operation), but does provide a basic and much needed image database capability. 1. Modifications to the Current Interface 1.1 Boundary extension Automatic boundary extension is useful in applications such as filtering via convolution, since the convolution kernel will extend beyond the boundary of the image when near the boundary, and in applications which operate upon subrasters, for the same reason. When reading from an image with boundary extension in effect, IMIO will generate artificial values for the out of bounds pixels using one of the techniques listed below. When writing to an image with boundary extension in effect, the out of bounds pixels are discarded. By default, an out of bounds pixel reference will result in an error message from IMIO. Consider an image line of length 5 pixels. The statement buf = imgs1r (im, -1, 7) references out of bounds by 2 pixels on either end of the image line, referencing a total of 5+2+2=9 pixels. If boundary extension is enabled and the get section completes successfully then Memr[buf] will reference the pixel at X = -1, and Memr[buf+2] will reference the first inbounds pixel. When an image is first opened zero pixels of boundary extension are selected, and any out of bounds references will result in an error. To enable boundary extension imseti must be called on the open image to specify the number of pixels of boundary extension permitted before an out of bounds error occurs. include <imset.h> call imseti (im, IM_NBNDRYPIX, 2) If boundary extension is enabled the type of boundary extension desired should also be set. The possibilities are defined in <imset.h> and are summarized below. BT_CONSTANT return constant if out of bounds BT_NEAREST return nearest boundary pixel BT_REFLECT reflect back into image BT_WRAP wrap around to other side BT_PROJECT project about boundary Types of Boundary Extension The type of boundary extension is set with the imset parameter IM_TYBNDRY. If the BT_CONSTANT option is selected the constant value should be set with an imseti or imsetr call to set the parameter IM_BNDRYPIXVAL. Boundary extension works for images of any dimension up to 7 (the current IMIO limit). A single IM_NBNDRYPIX value is used for all dimensions. This value is used only for bounds checking, hence the value should be set to the maximum out of bounds reference expected for any dimension. Larger values do not "cost more" than small ones. An actual out of bounds reference is however more expensive than an inbounds reference. 1.2 Image Database Interface The image database interface is the IMIO interface to the database containing the image headers. In this implementation the image header is a variable length binary structure. The first, fixed format, part of the image header contains the standard fields in binary and is fixed in size. This is followed by the so called "user area", a string buffer containing a sequence of variable length, newline delimited FITS format keyword=value header cards. When an image is opened a large user area is allocated to permit the addition of new parameters without filling up the buffer. When the header is subsequently updated on disk only as much disk space is used as is needed to store the actual header. The new header format is upwards compatible with the old image header format, hence old images and programs do not have to be modified to use the latest release of IMIO. In the future image headers will be maintained under DBIO, but the routines in the image header database interface described in this section are not exected to change. The actual disk format of images will of course change when we switch over to the DBIO headers. While the physical storage format of header will change completely under DBIO, the logical schema will change very little, i.e., our mental picture of an image header will be much as it is now. The main difference will be the consolidation of many images into a few files, and real support in the image header for bad pixels, history, and coordinate transformations. In addition a number of restrictions on the "user fields" will be lifted, the remaining distinctions between the standard and user fields will disappear, and database operations will be much more efficient than they are now. 1.2.1 Library Procedures The IMIO library procedures comprising the current image database interface are summarized in the table below. value = imget[bcsilrd_] (im, field) imgstr (im, field, outstr, maxch) imput[bcsilrd_] (im, field, value) impstr (im, field, value) imadd[bcsilrd_] (im, field, def_value) imastr (im, field, def_value) imaddf (im, field, datatype) imdelf (im, field) y/n = imaccf (im, field) list = imofnl[su] (im, template) nchars/EOF = imgnfn (list, fieldname, maxch) imcfnl (list) where pointer im, list char[] field, outstr, datatype, template, fieldname Image Database Interface Procedures New parameters will typically be added to the image header with either one of the typed imadd procedures or with the lower level imaddf procedure. The former procedures permit the parameter to be created and the value initialized all in one call, while the latter only creates the parameter. In addition, the typed imadd procedures may be used to update the values of existing parameters, i.e., it is not considered an error if the parameter already exists. The principal limitation of the typed procedures is that they may only be used to add or set parameters of a standard datatype. The imaddf procedure will permit creation of parameters with more descriptive datatypes (abstract datatypes or domains) when the interface is recut upon DBIO. There is no support in the current interface for domains. The value of any parameter may be fetched with one of the imget functions. Be careful not to confuse imgets with imgstr (or imputs with impstr ) when fetching or storing the string value of a field . Full automatic type conversion is provided. Any field may be read or written as a string, and the usual type conversions are permitted for the numeric datatypes. The imaccf function may be used (like the FIO access procedure) to determine whether a field exists. Fields are deleted with imdelf ; it is an error to attempt to delete a nonexistent field. The field name list procedures imofnl[su] , imgnfn , and imcfnl procedures are similar to the familiar file template facilities, except that the @file notation is not supported. The template is expanded upon an image header rather than a directory. Unsorted lists are the most useful for image header fields. If sorting is enabled each comma delimited pattern in the template is sorted separately, rather than globally sorting the entire template after expansion. Minimum match is permitted when expanding the template, another difference from file templates. Only actual, full length field names are placed in the output list. 1.2.2 Standard Fields The database interface may be used to access any field of the image header, including the following standard fields. Note that the nomenclature has been changed slightly to make it more consistent with FITS. Additional standard fields will be defined in the future. These names and their usage may change in the next release of IMIO. keyword type description i_ctime l time of image creation i_history s history string buffer i_limtime l time when limits (minmax) were last updated i_maxpixval r maximum pixel value i_minpixval r minimum pixel value i_mtime l time of last modify i_naxis i number of axes (dimensionality) i_naxis[1-7] l length of an axis ("i_naxis1", etc.) i_pixfile s pixel storage file i_pixtype i pixel datatype (SPP integer code) i_title s title string The names of the standard fields share an "i_" prefix to reduce the possibility of collisions with user field names, to identify the standard fields in sorted listings, to allow use of pattern matching to discriminate between the standard fields and user fields, and so on. For the convenience of the user, the "i_" prefix may be omitted provided the resultant name does not match the name of a user parameter. It is however recommended that the full name be used in all applications software. 1.2.3 Restrictions The use of FITS format as the internal format for storing fields in this version of the interface places restrictions on the size of field names and of the string value of string valued parameters. Field names are currently limited to eight characters or less and case is ignored (since FITS requires upper case). The eight character limit does not apply to the standard fields. String values are limited to at most 68 characters. If put string is passed a longer string it will be silently truncated. Trailing whitespace and newlines are chopped when a string value is read. 2. Database Utility Tasks Two image database utility tasks have been implemented, hedit and hselect. Hedit is the so called header editor, used to modify, add, or delete selected fields of selected images. The hselect task is used to select images that satisfy a selection criteria given as a boolean expression, printing a subset of the fields of these images on the standard output in list form. Both of these tasks gain most of their power from use of the evexpr utility procedure, now available in FMTIO. The evexpr procedure takes as input an algebraic expression (character string), parses and evaluates the expression, and returns as output the value of the expression. include <evexpr.h> pointer evexpr() o = evexpr (expr, getop, ufcn) where o Is a pointer to an operand structure expr Is a character string getop Is either NULL or the locpr address of a user supplied procedure called during expression evaluation to get the value of an external operand. ufcn Is either NULL or the locpr address of a user supplied procedure called during expression evaluation to satisfy a call to an external function. The operand structure is defined in <evexpr.h> . The best documentation currently available for the operators and functions provided by evexpr will be found in the manual page(s) for hedit . Additional documentation will be found with the sources. The expression evaluation procedure is probably the single largest procedure in the system (in terms of kilobytes added to an executable) and should not be used unless it is needed, but it can greatly increase the power of a task in the right application. ------------ From: /u2/tody/ Fri 20:32:55 10-May-85 Package: gio Title: spooling of plotter output Plotter output is now spooled many plots to a metafile rather than being processed to the plotter a plot at a time. You will no longer get a plot out immediately after a cursor mode snap. Spooling many plots into a single metafile is much more efficient, reduces the load on the system, and produces more pleasing output since the plots will come out on successive pages of paper as a single print job (with no wasted paper). All plotter output is spooled, e.g., cursor mode snaps, plots written to devices such as "stdplot", "vup", "imagen", etc., and plots written to the dicomed (e.g., by crtpict). Versatec plots are also now more closely centered on a page (formerly part of the plot would be in the perforations between pages of paper). The spooled plotter output will be disposed of to the plotter device when any of the following events occur: [1] When you type the command "gflush" in the CL, or enter colon escape ":.gflush" while in cursor mode. A ":.init" will also do it. [2] When you log off. [3] When the graphics device assigned to the stream STDPLOT changes. This happens when a new device name is assigned to the "device" parameter of a graphics task which writes to STDPLOT, or when you make a "snap" on a different device. [4] When the "maximum frame count" for a device is exceeded. This can be specified individually for the different plotter devices, with the default value currently being 16 frames per metafile. This is enough frames to make things efficient, but not so many that we run out of disk space or lose everything in a system crash. If the system should crash before plotter output is disposed of to the system, the metafile should still be intact in the /tmp directory (e.g., "/tmp/gvp...") and may be disposed of manually with a command such as the following (this would plot on the upstairs versatec): plotX -Tver -W=3 /tmp/gvp* | plot -Tver -M75 -Pvup These comments apply only to the plotter devices. The plotter (NSPP) devices currently supported are the following: calcomp dicomed imagen versatec vdown vup ------------ From: /u2/tody/ Sat 00:23:44 11-May-85 Package: gio Title: bug fix, log scaling, etc. A serious bug in pseudofile i/o to STDPLOT (gio$cursor/prpsio.x) was preventing plots written to STDPLOT by a task from being sent properly to the NSPP kernel. This was fixed and in addition, the following enhancements were made: [1] Log ticks are now labelled as powers of 10 rather than in linear units, i.e., as a 10 followed by the exponent one half character higher. [2] The EPSILON tolerance test in FP_EQUAL[DR] and FP_NORM[DR] was relaxed to 10*EPSILON to make these routines useful in real life situations where there is typically a small accumulation of error to be allowed for. ------------ From: /u2/tody/ Sun 20:19:38 12-May-85 Package: plot Title: graphics kernels added to PLOT package The three existing graphics kernels have been added to the PLOT package as the following tasks: gkidecode stdgraph stdplot These tasks are used to process metacode, either from a file or from the standard input (e.g., from a pipe). At present, however, there is no way to redirect a graphics stream on the command line. In the process of installing these tasks the name of the old NSPP task was changed to STDPLOT. These tasks will already be useful for metacode translation. In the future the CL i/o redirection syntax will be extended to permit redirection of graphics output, e.g., plottask arg arg | stdplot dev=vup cell+ ------------ From: /u2/tody/ Thu 10:08:43 16-May-85 Package: gio Title: degenerate windows If a window (world coordinate system) is degenerate in either axis GIO will now increase the upper limit of the window in steps of EPSILON until there is enough separation between the edges of the window to define the coordinate system. Formerly GIO would take an error action ("no range in [XY]") when given a degenerate coordinate system. ------------ From: valdes$iraf/ Fri 13:37:34 17-May-85 Package: lib Title: Subdirectory for include files of xtools To keep the lib directory managable we plan to make subdirectories of lib available as standard search places for include files and libraries. As a start include files for the xtools procedures have been placed in the subdirectory xtools. Currently these include files must be reference with an OS path name. This will change later to be OS independent. The usage (in UNIX) is: include <xtools/icfit.h> ------------ From: valdes$iraf/ Fri 13:42:33 17-May-85 Package: xtools Title: New xtools Three packages of xtools have been added to xtools. These packages are in subdirectories of xtools. The new xtools are as follows. ranges (rg_ prefix procedures): A ranges package using : as the range delimiter (like image sections) and without steps. It contains tools for parsing ranges, merging ranges, ordering ranges, packing and unpacking arrays, and doing median and averaging of sets of points within ranges. It is used in the icfit tools and in some of the longslit tasks. Indirectly through icfit it is used in a number of other tasks. gtools (gt_ prefix procedures): Graphics tools. These tools provide for printing help files in response to : cursor commands, paging (clearing the terminal) text, and tools to use a structure containing window and axis transformation parameters. These tools are used in icfit and indirectly through icfit in a number of tasks. icfit (ic_ and icg_ prefix procedures): Interactive curve fitting. These tools provide a somewhat higher level interface to the curfit math package. It has been designed to be efficient and to use a set of standard parameters (sample points, median and subsample averaging, functions, orders, deviant point rejection, and rejection growing. A subset of the package provides an interactive GIO interface. These tools are used in linefit, lineclean, identify, response, illumination, background, linebias, colbias, lineflat, colflat, linebckgrnd, and colbckgrnd in the images, longslit, bias, and generic packages. A good example of usage is t_linefit.x in the images package. The source code is moderately well commented but there is little other documentation as yet. Include files for the various set and get calls are in lib$xtools. If you have any interest in using them see Frank Valdes. ------------ From: valdes$iraf/ Fri 13:44:31 17-May-85 Package: images Title: Linefit and lineclean The linefit task has been changed again. It now has interactive graphics options and is greatly simplified. The previous version tried to do too many things and was difficult to understand. The new task fits image lines and creates output images with the fit, the difference or residuals, or the ratio of the image values to the fit values. Cleaning of deviant pixels has been moved to the task lineclean (though elimination of bad points from the fits is still available). There is no minimum division threshold and no output column selection. It uses the icfit tools for interactively setting the fitting parameters. A new task lineclean has been added to use curve fitting to detect and replace deviant pixels. It is similar to linefit and uses the same icfit tools and interactive fitting parameter approach. These tasks are also used in the generic scripts lineflat, colflat, Linebckgrnd, and colbckgrnd. ------------ From: valdes$iraf/ Fri 13:45:48 17-May-85 Package: generic Title: Changes The generic package has been revised. The tasks dcbias, chimages, btt, and biassub have been removed. Note that bias strip corrections are available in the imred.bias package. The tasks lineflat, colflat, linebckgrnd, and colbckgrnd have been changed to use the new linefit task in the images package. ------------ From: valdes$iraf/ Fri 16:20:06 17-May-85 Package: icfit Title: Help for interactive graphics in icfit package Help for the cursor mode keys and colon commands in the interactive graphics of the icfit package is available as ichelp. This package is used in numerous tasks and this help page is relevent to all these tasks: linefit, lineclean, colbias, linebias, colflat, lineflat, linebckgrnd, colbckgrnd, identify, response, illumination, background. ------------ From: /u2/tody/ Wed 17:48:07 29-May-85 Package: clpackage Title: revision of the root package (clpackage) The DTOI package has been moved from the root to IMRED, where it more properly belongs. The SDAS package has been added to the root and will eventually contain the applications software from our friends at the Space Telescope. ------------ From: /u2/tody/ Thu 00:57:22 30-May-85 Package: plot Title: IMPLOT enhancements A new keystroke 'e' has been added to IMPLOT to permit plot expansion with a completely redrawn box (makes a prettier plot than the 'E' and 'A' keystrokes in cursor mode). Two new colon escapes ":x x1 x2" and ":y y1 y2" have been added to permit fixing of the plot scale in X or Y, disabling autoscaling. A bug was fixed which was preventing log scaling from working. The ? help text was moved out into the file "plot$implot.keys". The manual page was updated. ------------ From: /u2/tody/ Thu 17:45:52 30-May-85 Package: clio,fio Title: filename templates The filename template expansion code has been rewritten with the results outlined below. The external specifications of the packages involved have not changed. [1] Sorting, if enabled, is now done only when expanding a pattern rather than globally over the entire template. Thus a template such as *.h,*.x will be expanded with all the .h files first in alphabetical order, followed by all the .x files in alphabetical order. File lists are no longer sorted, e.g., forms "a,b,c", and "@filelist". [2] Template expansion is now substantially more efficient. One result of this is that directory listings are now somewhat faster. [3] The new routine CLPREW has been added to the CLGFIL package. The new entry point may be used to rewind the file list. [4] An OS escape syntax has been added to permit "quoting" if OS filenames containing template metacharacters, e.g., $ and []. Anything enclosed by OS escape characters, e.g., '!', is treated as a filename string. Hence, templates such as !vmsdir$file!,!drb0:[dir]file! or !vmsfile are now permitted. The only effect of the ! are to turn metacharacters off and back on; beyond that the ! are just stripped out. [5] A new "buffered" FNT package has been added, to permit convenient filename template expansion for templates not necessarily bound to CL string parameters. Procedures: fntopnb - Expand template, open a buffered filename list fntclsb - Close buffered list fntgfnb - Get next filename from buffered list fntlenb - Get number of filenames in a buffered list fntrewb - Rewind the list Calling sequences: list = fntopnb (template, sortflag) fntclsb (list) nchars|EOF = fntgfnb (list, outstr, maxch) nfiles = fntlenb (list) fntrewb (list) These entry points are provided only for extra flexibility. The old CLPOPN[ISU] should continue to be used for the common case when the template is a CL parameter, or when the list is associated with the standard input, which might be redirected. In other changes, the CLGFIL sources in CLIO have become somewhat smaller and simpler. A couple include files were removed from CLIO. The FNT package was moved from CLIO to FIO. ------------ From: /u2/tody/ Thu 18:19:17 30-May-85 Package: gio Title: bug fix to polymarker in NSPP kernel The polymarker function in the NSPP kernel turned out to be the same as polyline; was changed to draw points rather than vectors. ------------ From: /u2/tody/ Fri 20:12:34 31-May-85 Package: gio Title: gio bug fix The GINIT procedure was using an invalid struct pointer to initialize the current text attribute structure. ------------ From: /u2/tody/ Fri 20:14:09 31-May-85 Package: hedit,imio Title: bug fix involving addition of parameters to image headers The image database interface routines in IMIO were not properly adding new parameters to image headers. There were a couple of bugs, i.e., the newline of the new record was being overwritten with a blank, and the datatypes of string and numeric fields were not being set properly. A bug was also fixed in HEDIT that would cause a cryptic "parameter not found" message when attempting to "add" (edit) an existing image header parameter. HEDIT will now invoke the edit function when an attempt is made to add a parameter to an image that already contains such a parameter. ------------ From: /u2/tody/ Sat 20:07:58 01-Jun-85 Package: gio Title: new font installed in NSPP kernel The original character font of the NSPP kernel has been replaced by a new font from NCAR/GKS. The new font is generally better than the old one, particularly the lower case letters and certain digits, although there appear to be certain characters in the new font which could stand some tweaking (i.e., the characters ";:#"). Roman, bold, and italic versions of the new font are available via gseti-G_TXFONT and \f[RIB] escape sequences in text strings. ------------ From: /u2/tody/ Sun 18:37:36 02-Jun-85 Package: gio Title: postprocessing of metacode files The graphics kernels have been modified to permit processing of metacode files which do not contain the open workstation instruction. Such metafiles are commonly produced by the ":.write" command in cursor mode. ------------ From: /u2/tody/ Sun 18:39:53 02-Jun-85 Package: gio Title: new NSPP device "tek" added A new graphics device using the NSPP kernel has been added. The new device, named "tek", is useful for interactively inspecting fully vectorized plots such as would they would appear on the imagen, versatec, dicomed, or other NSPP devices. In particular, character generation is done by the NSPP kernel rather than by the graphics terminal. ------------ From: /u2/tody/ Mon 15:49:33 03-Jun-85 Package: gio Title: text generation in the STDGRAPH kernel The text generator developed for the NSPP kernel has been installed also in the STDGRAPH kernel. The "normal" text quality for the stdgraph kernel is whatever is provided by the hardware character generator on the graphics terminal. Hardware character generation will be used if the text quality (G_TXQUALITY) is normal or low; the software character generator is selected if the quality is set to medium or high. Software character generation permits arbitrarily sized characters with arbitrary up vectors, making it possible to prepare plots just as they will appear on when drawn on a plotter device which does not have hardware character generation. The new character generator also fixes errors in string justification present in the old kernel. Strings will now be justified correctly regardless of the text quality. The text quality may be selected by setting the G_TXQUALITY parameter before calling GTEXT. When working interactively, however, it will be easier to override the runtime quality attribute by setting the text quality in cursor mode. A new cursor mode command ".txquality" has been added for this purpose. For example, after drawing a plot with hardware character generation and while in cursor mode, type ":.txq h" followed by a 0 to redraw the plot using software character generation. The "T" keystroke may be used to interactively draw characters and inspect the font, etc. ------------ From: /u2/tody/ Mon 15:59:27 03-Jun-85 Package: cl Title: error handling bug fix A bug has been fixed in the CL affecting error handling. When error recovery takes place while within an IFERR in SPP, e.g., after a <ctrl/c> when the CL has called a system service which uses error handling, it is necessary to reinitialize the error handling system. In normal programs this is done by the error restart code in the IRAF Main, but since the CL has its own error recovery code it must perform the reinitialization itself during error recovery. ------------ From: /u2/tody/ Mon 16:35:51 03-Jun-85 Package: gio Title: character size adjusted Ideally the size 1.0 character should be the same size in NDC units on all graphics devices, otherwise text which is scaled relative to the size 1 character will come out a different size on the different devices. The NSPP devices turned out to have character sizes some 20% bigger than the normal character size for the VT640, our reference graphics device. The hardware character size of the VT640 could of course not be changed, so I adjusted the character size in the graphcap entry of each NSPP device to make it agree with the VT640/4012. As as result of this change default size characters on plots will now appear both taller and narrower than before (note that the aspect ratio of a plotter device is different than that of the terminal). An entirely reasonable plot still results. If precise control over the character size is desired, the character size may in NDC units may be set explicitly, rather than scaling relative to the device size-1 character. There is currently no way to change the aspect ratio of a character, since only one size parameter is used for both axes (this is consistent with GKS). ------------ From: /u2/tody/ Mon 18:51:33 03-Jun-85 Package: gio Title: error checking of ":.snap" The cursor mode command ":.snap" now checks for an error when trying to open the snap device, printing a warning message if there is a problem rather than going through error recovery. ------------ From: /u2/tody/ Tue 22:51:26 04-Jun-85 Package: gio Title: bug fix affecting log scaling A case turned up in which glabax would fail to log scale correctly. The step size between minor ticks was so small that the next major tick would never be reached, resulting in an infinite loop. This bug has been fixed and the log scaling code has been extensively tested. ------------ From: /u2/tody/ Wed 15:51:50 05-Jun-85 Package: cl Title: bug fix in CL The internal CL function "breakout" breaks a long form parameter name into fields, returning a pointer to each field (a long form parameter name is "task.param" or "package.task.param"). In order to return a pointer to an EOS delimited string the routine overwrites the periods in the input string with EOS, returning a pointer to each substring (needless to say, not a good practice). A bug was found and fixed involving the use of the breakout procedure in the run time opcode procedures. When such a procedure is passed an operand which is compiled into the dictionary, the first call to breakout was modifying the input operand but functioning successfully. In the second call the same input operand was passed (but modified in the first call), hence the bug. I added a new procedure called "breakcopy" which calls breakout on a local copy of the input operand. All calls to the breakout function in the CL were examined and parcelled into three cases: [1] breakout called on a local copy, no changes; [2] reentrant call to breakcopy possible, routine changed to make local copy of operand and call breakout directly; and [3] simple reference to a nonlocal operand, call to breakout replaced by equivalent call to breakcopy. Most changes were of the latter type and occurred in the file opcodes.c. ------------ From: /u2/tody/ Wed 16:06:37 05-Jun-85 Package: fio Title: bug fix in binary file i/o The procedure FWATIO was modified to mark the FIO buffer inactive regardless of the status returned by an i/o transfer to the buffer. Formerly the routine would not mark the buffer inactive following an i/o transfer which returned an error. This would cause the error to "stick" when trying to read the file (and probably when trying to write it too). ------------ From: /u2/tody/ Wed 19:03:06 05-Jun-85 Package: library Title: mathematical constants A new file <math.h> has been installed in the system library. This file contains definitions of various constants and macros useful in numerical applications. ------------ From: /u2/tody/ Thu 20:27:16 06-Jun-85 Package: fmtio Title: bug fix in CTOWRD (hence also CTOTOK, GARGTOK) The CTOWRD procedure was recognizing newline as a word delimiter even when the "word" in question was a quoted string or character constant. Since newlines are permitted in strings and character constants this was an error. CTOWRD is called by both CTOTOK and GARGTOK, hence the bug affected both of these routines as well. ------------ From: /u2/tody/ Fri 11:44:20 07-Jun-85 Package: spp Title: pointer conversions in structures Several macros have been added to <iraf.h> to facilitate pointer conversions within structure definitions. These macros convert a struct pointer into a pointer to some other datatype, e.g., to reference a field of a structure. P2C(p) *(struct) -> *char P2S(p) *(struct) -> *short P2L(p) *(struct) -> *long P2D(p) *(struct) -> *double P2X(p) *(struct) -> *complex By definition pointer to struct is equivalent to a pointer to int hence there is no conversion for integer fields. Also by definition int and real occupy the same amount of storage, hence pointer conversions are not necessary to reference fields of type real. Since an SPP pointer is implemented as an integer it may be stored in an integer field within a structure (we are taking liberties within struct declarations that should not be taken in regular code). Sufficient storage should be allocated for each CHAR or SHORT field of a structure to permit variations in the actual size of a char or short on the host machine. Allocate one full struct unit for each storage element of type char, short, int, long, or real. Allocate two struct units per double or complex element. When storing character strings in a structure, dimension the string odd and allow one extra char for the EOS. Double and complex fields must have double alignment with the pointer to the start of the structure, i.e., a double or complex field must be placed at an offset of 0, 2, 4, etc. struct units from the start of the structure. Example: define LEN_ASTRUCT (10+40+40) define SZ_NAME 39 define A_FD Memi[$1] define A_STATUS Memi[$1+1] define A_SCALE Memd[P2D($1+2)] define A_OFFSET Memd[P2D($1+4)] # (extra space) define A_NAME1 Memc[P2C($1+10)] define A_NAME2 Memc[P2C($1+50)] Note that the offset into the structure is always given in struct units. It is good practice to allow extra space for expansion. Likewise it is advisable to place char storage at the end of the structure to allow for expansion without having to recompute the offsets of the scalar fields. If these rules are followed the resulting structure definitions should be portable to most or all IRAF host machines without change. ------------ From: /u2/tody/ Sun 22:31:53 09-Jun-85 Package: fmtio Title: new procedures added to the strings package Two new procedures have been added to the strings subpackage of FMTIO. The new procedures, more general versions of the old routines STRIDX and STRLDX, are called STRIDXS and STRLDXS and differ from the original routines in that the first argument is a string defining a set of characters to be searched for, rather than a single character. index = stridx (ch, str) # first occurrence of CH in STR index = stridxs (set, str) # first occurrence of SET[i] in STR index = strldx (ch, str) # last occurrence of CH in STR index = stridxs (set, str) # last occurrence of SET[i] in STR I have been thinking of adding these procedures for some time since there have been numerous occasions on which they would have been useful. The addition of the new routines at this particular time was prompted by the redefinition of a character constant as an integer value. The following statement is now illegal since the first argument is an integer rather than char scalar: ip = stridx ('a', str) # BAD!! Rather than define a char variable to convert this into a legal statement, the following form may be used instead: ip = stridxs ("a", str) ------------ From: /u2/tody/ Sun 22:42:40 09-Jun-85 Package: spp Title: revision to the SPP language standard The original standard for the SPP language, which has remained unchanged since its first definition several years ago, defined a character constant (e.g., 'a') as a numeric constant of type CHAR. It has recently become apparent that the SPP language can be simplified by redefining the character constant as an integer constant. With this new definition the following are all legal integer constants, equating to 32 decimal: 32 40B 20X ' ' Likewise, the following are all equivalent to the decimal integer 49 (SPP specifies that characters are represented internally in ASCII): 49 61B 31X '0' + 1 TO_DIGIT(1) This change requires that all existing code which uses character constants in argument lists be changed, since it is illegal to pass an integer argument to a procedure expecting an argument of type char. A complete list of all such (newly) illegal statements has been prepared by automatic means and the affected code is in the process of being revised. ------------ From: /u2/tody/ Mon 19:23:58 10-Jun-85 Package: fio Title: new routines added to FIO Two new procedures PUTCI and GETCI have been added to the FIO interface. These are integer versions of the old procedures PUTC and GETC. ch/EOF = getci (fd, ch) # int fd, ch, getci() putci (fd, ch) ch/EOF = getc (fd, ch) # int fd; char ch, getc() putc (fd, ch) The new routines are functionally identical to the old ones but eliminate the char scalar argument and function value (which can cause portability problems). PUTCI is particularly useful for the output of character constants, e.g.: call putci (fd, '\n') This is certainly a kludge, but both types of procedures are useful depending on the application, and the revision is upwards compatible. In the ideal world a better solution would be for SPP to accept either char or integer scalar arguments but always pass an integer in the argument list, like C. This is beyond the capabilities of the current preprocessor because it does not parse expressions and keep track of their datatypes. ------------ From: /u2/tody/ Tue 09:53:06 11-Jun-85 Package: fmtio Title: argument datatype changed in DTOC, XTOC The datatype of the "format" argument to DTOC and XTOC has been changed from CHAR to INT, since the actual argument in typical calls is usually an explicit character constant, e.g., 'f', 'e', or 'g'. ------------ From: /u2/tody/ Tue 13:19:28 11-Jun-85 Package: system Title: minor revisions to match argument types Approximately 75 lines of code were modified to match argument datatypes (char to char, int to int) as a result of the recent change in the SPP language standard. The count of number of lines of code modified does not include numerous related modifications made in declarations, etc. Some of the affected routines were quite old and were cleaned up and made more efficient in the process. Most of the changes were procedure name changes, e.g.: putc -> putci stridx -> stridxs The following files were affected. PKG: utilities decod_tablst.x t_translit.x help/lroff breakline.x center.x justify.x lroff.x section.x do_ls.x output.x skiplines.x help prblkhdr.x hbgetblk.x hbgetblk.x hinput.x manout.x manout.x manout.x modtemp.x images imheader.x hedit.x imdebug/dump.x tv/display zopnim.x softools/yacc/debug ytab.x(dc.y) system page.x directory.x match.x beep.x plot graph.x SYS: clio clcmd.x zfiocl.x fmtio ctocc.x dtoc.x xtoc.x strtbl.x fmtstr.x parg.x gio nspp/encode.x stdgraph/stgencode.x glabax/glbtitle.x cursor/(rcursor.x,grcstatus.x,grccommand.x) debug realio.x nop.x clio.x strings.x fio.x fio2.x words.x plot.x mtio mtgetpos.x tty ttyputline.x ttyinit.x debug.x etc erract.x sys_ptime.x x_debug.x envlist.x propen.x main.x xerstmt.x fio x_debug.x, fchdir.x vfntrans.x imio immap.x db/imgetb.x db/imaddf.x db/imdelf.x fmtio chrupr.x chrlwr.x dtoc.x fpradv.x patmatch.x pargx.x mtio mtposition.x mtdevlist.x tty ttyputline.x For the most part the modifications were trivial and should not cause any problems. Routines where were more heavily modifed include the pattern matching code in FMTIO and the character output code in LROFF (HELP). ------------ From: valdes$iraf/ Thu 10:29:36 13-Jun-85 Package: xtools Title: Simple text database tools Some simple text database tools have been added to the xtools library. Some help text under the name dttext is available through the help task. The source code is in dttext.x in the xtools directory. These tools are useful for keeping multiple database text entries in one file. These tools are in use in the longslit package. ------------ From: /u2/tody/ Fri 13:26:57 14-Jun-85 Package: spp Title: SPP compiler revisions The first pass of the SPP compiler, i.e., the C program XPP, has been modified as follows: [1] A new file "decl.c" was added containing a package of routines for parsing argument lists and declaration statements. As arguments and declarations are parsed each argument or local variable is added to an internal symbol table. When a procedure has been fully processed a code generation routine is called to output the RPP declarations required by the second pass. Declarations are output in the following order: (1) scalar arguments, (2) array arguments, (3) local variables, (4) miscellaneous declarations (commons, errchk, etc.), (5) switch variables and string array names, (6) data statements. This ordering is independent of the order in which these items occur in the source. The output ordering is as required by Fortran. This revision was made to permit correct generation of Fortran code for the following type of case, which is quite common in IRAF: procedure alpha (fd, x, xlen) pointer im real x[xlen] int xlen begin Previously the Fortran generated for this procedure would have declared the OUTSTR array and then the MAXCH scalar. This is illegal Fortran because the length argument XLEN would be used to dimension X before being declared as an integer argument; some compilers would accept such a construct and compile it correctly, while others would conclude that XLEN was a real by applying the autotyping convention of Fortran. This is no longer a problem because SPP will reorder the argument list to suit Fortran. [2] In addition to reordering argument lists, the new code will also detect multiple declarations of the same identifier, printing a warning message for each such case found. [3] XPP (the first pass of the preprocessor) now prints the name of each procedure as it is processed, in addition to the name of each file processed. This gives the user more feedback on the progress of the compile, particularly when a large file is being compiled on a slow machine. Typically the filename will appear, then there will be a delay while all the header includes are read in, then the individual procedures in the file will be processed, then there will be a delay while RPP is called to generate the Fortran output file. Related changes were required in the lexical analyzer (file xpp.l) and in the file xppcode.c. In particular the way in which context sensitivity is handled was completely changed, removing all of the old context variables and replacing them with a single multistate variable named "context". Additional buffering was added (the symbol table and another buffer for miscellaneous declarations), hence now an entire procedure is buffered internally before any output is generated. Comments are no longer passed to RPP. By stripping comments and optimizing character output the performance of XPP was improved by 25%. ------------ From: tody$ Fri 21:48:40 14-Jun-85 Package: vops Title: precision of VOPS projection functions increased The VOPS function procedures ASSQ, ASUM, and ADOT have been changed to accumulate the sum of squares, sum, or dot product as a variable of type real, double, or complex depending upon the datatype of the procedure. The function value was also changed to return a floating point value of the correct precision. Formerly both the sum and the function value were of type PIXEL, making the procedures of little value for integer vectors. ------------ From: tody$ Sun 21:13:22 16-Jun-85 Package: system Title: filename changes All files in the lib$ and sys$ directories (excluding lib$xtools) with names longer than 9 characters or which contained underscore characters were renamed. This eliminates the need to mess with filename mapping (except for filename extensions) when bootstraping the system on a foreign host. There is no need to do the same thing for the applications packages since the IRAF VOS will handle the filename mapping once the core system is up. ------------ From: tody$ Sun 21:17:32 16-Jun-85 Package: math Title: nonportable constructs fixed in math packages The bevington fortran sources had nonstandard UNIX continuation (& in col 1). This was changed to Fortran standard continuation (char in col 6). The file llsq/sva.f was found to have errors in the hollerith format statements at the end of the file. The version of the file from VMS IRAF was substituted for the UNIX version, after verifying that the only differences were in the format statements. Neither library has been recompiled as yet. ------------ From: /u2/tody/ Thu 19:34:06 20-Jun-85 Package: cl Title: logout change, bug fix The CL will no longer automatically deallocate magtape drives when logging out. This greatly speeds up logout and permits the same user to logout of a second CL while the first still has a tape drive allocated (formerly the drive would be deallocated). Since deallocation is no longer automatic, people will inevitably forget to deallocate the tape drive. The next user to attempt to allocate the drive will get a "drive already allocated" error message, even if the user has deallocated the drive at the host system level. When this occurs delete the file "dev$mta.lok" or "dev$mtb.lok", making certain by first typing the file that it is not in use. If this file does not exist IRAF assumes that the drive can be allocated to someone else. The file MUST exist while a drive is being accessed. A recent bug fix has been made which will probably cure the bug which has been causing parameter values to be overwritten with garbage when a task exits. ------------ From: tody$ Thu 19:42:41 20-Jun-85 Package: gio Title: graphics kernels Recently all filenames in the iraf VOS were changed which were longer than 9 characters or which contained underscores. Some runtime files were affected, in particular the graphics kernels, "lib$x_stdplot.e" and "lib$x_stdgraph.e". These are now named "lib$xstdplot.e" and "lib$xstdgraph.e". The system tasks spawned by the graphics kernels when "gflush" is typed are now submitted with a reduced priority of 4. ------------ From: tody$ Thu 19:42:41 20-Jun-85 Package: system Title: merge A merge of the VMS and UNIX versions of IRAF is in progress. A number of changes have been made which have not yet been recorded by revisions notices. These will be summarized when the merge is complete. ------------ From: valdes$iraf/ Mon 14:11:37 01-Jul-85 Package: generic Title: Changes to lineflat and colflat The task lineflat for creating flat fields by ratioing the image lines by function fits to the image lines has been changed. It used to be a simple script calling the task linefit in the images package. Now it it is a complied task which is similar to linefit. This allowed the addition of a new parameter specific to the flat field problem. The new parameter is call minflat. If the fitted value is less than the value of this parameter the flat field value is set to unity. The task colflat used to be a script calling linefit. Now it calls lineflat and, hence, has the minflat parameter. ------------ From: valdes$iraf/ Tue 16:49:50 02-Jul-85 Package: gtools Title: Expanded gtools structure The gtools structure has been expanded to include a title, xlabel, and ylabel. The call gtlabax in place of glabax will call glabax with the appropriate strings. If the gtools pointer is null then glabax will be called with no strings. The initialization is null strings. Calls to gt_sets and gt_gets are used to set and get the strings with the parameters GTTITLE, GTXLABEL, and GTYLABEL from gtools.h. gt_colon has also been modified to respond to :title, :xlabel, and :ylabel. These changes do not affect any code except those which use the colon parameters title, xlabel, and ylabel. ------------ From: valdes$iraf/ Mon 15:28:57 08-Jul-85 Package: images Title: New magnify The magnify task has been rewritten using the new two dimensional interpolation and boundary extension packages. Thus, magnify will now do true two dimensional interpolation. The block averaging aspect of the old magnify has been eliminated (use blkaverage first if desired). ------------ From: valdes$iraf/ Tue 14:18:36 09-Jul-85 Package: local Title: New task imfunction A new task called imfunction has been added to the local package. It applies a function to the input pixel values to produce an output image. Currently only the log (base 10) function is available but other functions can easily be added. This task differs from imarith in that imarith is a binary operator and imfunction is a unary operator. ------------ From: /u2/davis/ Fri 15:25:04 12-Jul-85 Package: dataio Title: rcamera mod RCAMERA now writes the total time since last readout, and the total shutter open time in the IRAF image header as well as the integration time. ------------ From: /u2/davis/ Wed 11:51:13 17-Jul-85 Package: images Title: shiftlines bug Shiftlines was not calculating the fractional pixel shifts correctly in the case of boundary extension. Instead it was shifting by the nearest integer, i.e 1 if shift=1.5. ------------ From: /u2/davis/ Mon 09:32:44 22-Jul-85 Package: dataio Title: rfits mod The parameters blank has been added to the RFITS task. Blank values in the FITS image will be replaced in the IRAF image by the user defined quantity blank. Appropriate type conversion will be performed on blank. For example if blank = -1.5 and the output image type is real, blank pixels will have the value -1.5; if the output image type is integer, blank pixel values will have the value -1. ------------ From: valdes$iraf/ Mon 14:15:47 29-Jul-85 Package: images Title: Changes 1. The task revisions has been added to page revisions to the images package. The intent is that each package will have a revisions task. Note that this means there may be multiple tasks named revisions loaded at one time. Typing revisions alone will give the revisions for the current package. To get the system revisions type system.revisions. 2. A new task called fit1d replaces linefit. It is essentially the same as linefit except for an extra parameter "axis" which selects the axis along which the functions are to be fit. Axis 1 is lines and axis 2 is columns. The advantages of this change are: a. Column fitting can now be done without transposing the image. This allows linefit to be used with image sections along both axes. b. For 1D images there is no prompt for the line number. ------------ From: valdes$iraf/ Mon 14:17:40 29-Jul-85 Package: generic Title: Changes July 26, 1985: 1. Help page available for flat1d. 2. Background has been modified to use new fit1d task. It now does column backgrounds without transposes and allows image sections. ------ July 25, 1985: 1. A new task called flat1d replaces lineflat and colflat. It is essentially the same as lineflat except for an extra parameter "axis" which selects the axis along which the 1D functions are to be fit. Axis 1 is lines and axis 2 is columns. The advantages of this change are: a. Column fitting is done without transposing the image. b. The colflat script using image transpositions would not work the same as lineflat when used with sections. Now it is possible to mosaic several flat fields as need with multiple slits or apertures. c. Because no transpose is needed and it is not a script flat1d should work faster than colflat. d. The prompts for interactive fitting are now correct when doing column fits. ------ July 23, 1985: 1. The task revisions has been added to page revisions to the generic package. The intent is that each package will have a revisions task. Note that this means there may be multiple tasks named revisions loaded at one time. Typing revisions alone will give the revisions for the current package. To get the system revisions type system.revisions. 2. The tasks linebckgrnd and colbckgrnd have been combined into one task with the extra hidden parameter "axis". With axis=1 the task is the same as linebckgrnd and with axis=2 the task is the same as colbckgrnd.