02 SEP 2000: version 1.10
- Support devfs naming convention for Linux 2.2+ (Byron Stanoszek)
- Public release.
20 JUL 2000: version 1.9.5
- Added Matrox G100, G200, G400, G400 MAX support. (Tom Zerucha)
17 JUL 2000: version 1.9.4
- Added NVidia TNT and TNT2 support. (Tom Lees)
21 JUN 2000: version 1.9.3
- Removed the 1023 limitation on vertical scanline size, since some cards CAN
support it without squashing the screen. (Byron Stanoszek)
10 MAY 2000: version 1.9.2
- S3 code now uses a variable reference clock for S3 Trio 3D/2X boards.
06 MAR 2000: version 1.9.1
- Minor code enhancements and improvements.
--Older Versions--
15 MAR 1999 - version 1.9
- (1.8.1) remove "strip" option for "install" target (causes errors on older
systems)
- fix sync polarity bug in Ti3026 Sync-On-Green code (Giacomo COMES)
- update vgaset's licensing to GPL (Igor Grobman/Debian, Greg Lehey)
- fix the bogus H/V VGA register upper limits. Especially vertically the
limit was 4096, but should have been 1023 instead, which caused a lot of
modes that appeared to stretch far below the bottom of the screen). Also
fix all modelines that violated this constraint in the TextConfig file.
(with help from Udo van den Heuvel)
- (1.8.2) some changes to STM_reset script (Suggested by Igor Grobman, Debian)
- DOS compile fixes and enhancements by Molnar Laszlo. New version of cwsdmpi
included.
- don't bail out when ResetProg or FontProg doesn't exist (suggested by Igor
Grobman, Debian)
- some FAQ additions
- fix S3 Sync-on-Green code for Ti3026. (Timo Giesselmann)
- don't allow ANYONE to run STM from a non-virtual terminal, add "-x" option
to override this (suggested by Yann Dirson, Debian)
- set default fontpath to empty string, because "setfont" knows it as well
(suggested by Yann Dirson, Debian)
- add small text explaining how to install STM using the SYSV init method in
contrib/sysvinit (Michael De La Rue)
- add 40x25 mode (Otto J. Makela, Riku Saikkonen)
- (1.8.3) add Neomagic support (Pavel Machek)
- add "-k" option to disable software resize (hardware is resized) (Pavel
Machek)
- (1.8.4) add MACH64 Rage II support (should now support all MACH64 boards
with internal programmable clock)
- fix ET6000 MClk setting (use safer method recommended by Tseng)
- LOTS of work to program MACH64 DSP properly for textmode clocks. Thanks to
Marc La France (of XFree86 fame) for inputs on this.
- (1.8.5) integrate RIVA128 support from Martijn van Oosterhout.
- (1.8.6) fix bug in Neomagic code (option topleft didn't work)
- fix segfault in ET6000 "special" code
- force correct clockchip for some chips if not defined
- fix RIVA128 clock code (Martijn van Oosterhout)
- LOTS of extra free fonts (Frank Klemm)
- (1.8.7) Yikes... the RIVA code used a FIXED (#define) PCI base address.
use PCI code now.
02 DEC 1997 - version 1.8
- add Matrox Millennium II support (jurriaan <thunder7@xs4all.nl>)
23 NOV 1997 - version 1.7
- (1.6.1) use /dev/tty0 instead of /dev/console (finally!)
- fix two minor bugs (missing break in vga_prg.c, new include for libc6)
(Molnar Laszlo)
- added some more cards to PCI detected devices (avoids "PCI unknown device"
message). Some cleanups in the PCI debug messages (was too verbose).
- Andree Borrmann's DOS/DJGPP patches (now Mystique should work in DOS)
- documentation updates (SIGWINCH, GPM, SDRAM, how to raise textmode max
clock)
- (1.6.2) protect users from running STM from inside X or remotely (only
root can do that now). (Benson L Chow)
- add "hshift" parameter to modelines (Massimiliano Ghilardi)
- (1.6.3) add support for Mystique rev 3 (Martin Scherbaum, S.u.S.e.)
- change setpalette-to-kernel color ordering from straight VGA to ANSI. This
should make setpalette store the correct colors now. (Massimiliano
Ghilardi)
- some Alpha compilation fixes (Antti Barck)
- (1.6.4) add Cirrus Laguna clockchip support (CLGD546x)
- add ISO_FONT9 option (Tuukka Toivonen)
- partial fix for IBM RGB RAMDAC problem with standard VGA clocks
- change "PROBLEMS" file into a FAQ. Maybe people will read it then...
- some DJGPP/DOS compilation fixes (missing #includes).
28 JUN 1997 - version 1.6
- (1.5.1) fixed bug in 9-pixel char mode (Stuart Lamble)
- fixed SIGSEGV in Cirrus clock code (Stanislav Voronyi)
- fixed bug in ET4000 W32i special_svga.c (linear address map comparator)
- (1.5.2) one more Tseng register is reset in special_svga.c (CRTC 0x33)
- fix bug in setVGAreg for 'io' registers (missing "break")
- add MATROX Mystique support (Trey Boudreau)
- (1.5.3) add a few new fonts (Massimiliano Ghilardi/Michele Maltoni)
- add DOS screen resizing patches and font loading from Andree Borrmann
- add video modelines text in contrib (PAL/NTSC)
- better DOS manual file formatting by jkekoni@aurora.tky.hut.fi
- MClk max limit is now 100 MHz instead of 70 MHz.
- the LSM file in version 1.5 was 1 year off :-)
15 APR 1997 - version 1.5
- (1.4.1) added "-k" option to setpalette (as requested)
- explicitly set a few more registers on ET4000/ET6000 to get better
textmode restoration after X-server crash. Works much better now.
- SVGATextMode.c: added "-o" option, for "set80" functionality.
(setting all std. VGA regs by force).
- setVGAreg.c: added "-c" and "-s" options for defining a set-mask or a
clear-mask instead of an absolute value.
- DOS programs now use DJGPP cross-compiler from Linux.
(doesn't work yet...)
- setVGAreg: "idx" and "io" 'registers' added for generic IO access.
- updated some docs (there was some obsolete stuff)
- fix bug in resize_specified_vts() (endless loop resizing first specified
VT)
- only vga_prg.h includes asm/io.h now
- (1.4.2) fixed dos-gcc stuff in makefile: now "make dos" makes DOS
executables out of the box (finally -- no more rebooting or dosemu)
- some stuff added to allow grabmode (not the others!) to compile under
the Cygnus GCC for Windows. Clock probing does not work yet.
- user_time() code added from Harald Koenig.
- (1.4.3) added Trident TGUI support contributed by Massimiliano Ghilardi.
- Integration of XFree 3.2 clock code (XFree/common_hw).
THIS MAY CAUSE SOME SERIOUS NEW BUGS!
- (1.4.4) Mach64 support -- not finished yet. Not at all...
- 9-pixel wide characters now correctly display "box" characters (Stuart
Lamble)
- added PCI bus code from XFREE/common_hw/xf86_PCI.* This will be needed for
some of those PCI-only cards (Matrox Millennium et al).
- (1.4.6) Matrox Millennium added (not working yet)
- (1.4.7) ET4000 textmode restore enhancements
- Matrox Millennium code fixed -- ChipSet MATROX / ClockChip ti3026 (finished
and initial testing by Trey Boudreau <trey@treysoft.com>)
- Matrox Mystique -- ChipSet MATROX / ClockChip MGA1064 -- added to chipset.h
No other support for Mystique yet (Trey Boudreau)
- (1.4.8) MACH64CT added by Matthias Grimrath (m.grimrath@tu-bs.de).
- unlock_svga.c created, containing svga unlocking code from vga_prg.c
- special_svga.c created, containing svga special code from vga_prg.c
- Millennium MClk support added.
- (1.4.9) Millennium MClk support fixed and enabled.
- Accept range of frequencies when only one single frequency is given in
config file. This reflects the reality of so-called "fixed-frequency"
monitors.
- some mach64ct fixes by Matthias Grimrath.
- replaced misleading "CLGD542x" chipset with more general "CIRRUS" chipset.
22 SEP 1996 - version 1.4
- (1.3.1) added GIO_CMAP/PIO_CMAP to setpalette, so on kernels >1.3.3
switching VT's doesn't restore the old palette, but remembers the new one.
- kversion.c split out from ttyresize.c (needed it in setpalette)
- validate_mode slightly simplified (and faster).
- "8-bit" mode added to get/setpalette
- (1.3.2) updated E-mail address references
- S3Trio MClk programming support
- fixed problem with S3Trio support for standard VGA clocks (thanks Dirk!)
- (1.3.3) ET6000 support.
- improved ET6000 max. clock speed (register tweaking)
- (1.3.4) SiS clock selection fix
- some get/setpalette cleanups (nothing essential).
- some documentation updates
- updated ET4000 clock selection code to new XFree86 3.1.2C method.
This obsoletes the "AltClockSel" method.
- fix problem with "ClockProg" reporting an error:
"Clock selection: 0 MHz clock selected"
- PRELIMINARY S3 Virge support. UNTESTED.
- puny bug in setVGAreg fixed.
- performance tweaking ET6000 fixed (didn't work).
- added Makefile rules for creating source/binary/DOS distributions
21 APR 1996 - version 1.3
[This will be mainly a few bugfixes and cleanups on version 1.2]
- (1.2.1) fixed bug in validate.c, ATI MACH32 checking of number of clocks
(typo of " 2" instead of "32")
- filename checking in cfgfile.y improved (setting the fontprog to "setfont
-u something" is now possible). This was needed to allow setfont to set a
unicode map each time it is called (otherwise loading a raw font file erased
the original unicode mapping)
- added sync_on_green option for TI RAMDAC's (will somebody test this?)
- (1.1.2) some Redhat 3.0.3 (contrib) patches included (Makefile, TextConfig)
- added "echo" keyword to TextConfig syntax.
- some "Bool" type conflicts removed.
- VGA_CAN_DO_64KB as an option to allow 32k char modes (default was 16k)
- some contrib directories updated (setfont,...)
- kernels 1.3.78 and up have MAX_NR_CONSOLES defined _outside_ the __KERNEL__
defines, so __KERNEL__ must no longer be defined in ttyresize.c. It causes
conflicts with ___strtok. (thanks to Noel Cragg)
- 16 color textmode option added (default is 8-color with blinking)
- (1.2.3) missing "break" in validation caused error about S3 clocks when
using std. VGA chipset.
- some more standard VGA modes added, some doubles removed.
- slight re-ordering of TextConfig file to give it a little more structure.
Some contributer's names were omitted. Sorry about that guys, but it
was becoming a mess. You're in the CREDITS file.
- (1.2.4) some man-file updates, cleanups,...
- all clockchips (except Cirrus) will now prefer using the standard VGA
(hardwired) clocks of 25 and 28 MHz instead of always programming a new
one.
- CLOCKCHIP_X option now works for ET4000 + ICD2061a, too.
- included MSWindows grabmode executable.
7 MAR 1996 - Version 1.2
- (1.1.0) added some more 56 kHz modes (20-pixel fonts) to TextConfig file
- added contrib/setcolor: a kernel palette changer for 1.3.x kernels,
contributed by Bob McCracken.
- (1.1.1) mode grabber now stores messages in mode struct instead of printing
them. This allows other programs to use the mode grabber without being
bothered by unwanted output (e.g. scanmodes.exe for DOS).
- finally: the grabmode/clockprobe glitches are explained (but not solved yet).
- grabmode enhanced again to accomodate Interlaced mode detection.
Also restructured modedata.c code.
- Option "Number_Nine" removed. Was never used anyway...
- ! (1.1.2) BIG change in config file parser: now uses lex/yacc
! Most files have been changed for this. Many other files were obsoleted.
! !!! NOTE syntax change for font definitions !!!
- removed automatic "\n" printing in PDEBUG, PERROR,...
most "bugs" removed below were not in 1.1, they resulted from the massive
changes caused by the lex/yacc thing.
- (1.1.3) further integration of lex/yacc stuff.
- ALL clock frequencies are now in kHz (integers). This could cause lots of
nicely hidden bugs!
- (1.1.4) first shot with all-new-and-improved code seems to work. Speed
is about three times as fast! more major fun-stopper bugs removed.
- (1.1.5) further bugfixes
- one more "bugfix" for a kernel mis-feature: VT_RESIZEX doesn't know about
doublescan modes, and thus we have to lie to it about the number of video lines...
(this _was_ in 1.1)
- option "NoUnderline" has gone. Underlining is default "off" (i.e. set to
line 31) until defined otherwise using "UnderLine xxx".
- included screen-capture program contrib/consoletools/dscrn.c and Borland
C's gettext/puttext clones contrib/consoletools/gettext_settext_demo.c
written by Bob McCracken.
- (1.1.6) grabmode's probe now does less loops. Precision is still good enough,
speed is better.
- 9-pixel wide fonts allow upto 9/8 DACSPEED clock (see doc/PROBLEMS for reason)
- (1.1.7) further porting of misc utils to new data structures.
- So I discovered "lint". So what? All boolean values are now type "bool" instead
of "int".
- (1.1.8) get/setVGAreg reworked a bit.
- more cleanups. The more I clean up, the more mess I turn up... Go figure...
- remains of configfile.c split into file_ops.c and string_ops.c
- ClockProg updated to use new parser
- Due to all the changes, a bug in grabmode dissapeared as well (sync polarity
was reported wrongly)
- (1.1.9) clock validation fixed (was broken from 1.1.4)
- (1.1.10) fixed bug in clockchip clock validation
- flex "%option" lines removed (they needed v2.5 of flex, which not all people
have)
- clockchip limit checking moved to validate.c (at least it doesn't bail out
in the middle of VGA programming anymore, which was pretty stupid)
- more config file sanity checks added, and moved to validate.c
- added dump_cfgdata.c
- ET4000 clock selection modified a little (clears CLKDIV4 in addition to
CLKDIV2 just to make sure).
- ET4000 + ICS5341: safer clock #2 selection to avoid wrong clock setting
(prior versions needed "hibit_low" option to avoid setting 1/2 clock)
- preliminary support for ET4000 + ICD2061 (Diamond Stealth 32)
- (1.1.11) bug fixed in new clockchip code ("break" statements added)
- fixed several bugs in ET4000 ICD2061 clock code.
- ClockProg/STM now correctly passes clock index to clock program.
- ClockProg "index" parameter no longer required (but still allowed).
- (1.1.12) added grabmode parameters to specify known X/Y size of mode
(so grabmode has to do less guessing and more knowing)
- getbyte() renamed to getint(). (was int anyway)
- buggy calls to GMOFLG_ISSET macro expanded (only single options allowed)
- awesome bug removed in ChipSetRec structure (allowed options were all wrong)
- (1.1.13) bug in IBM RGB clockchip code removed (caused by transition to kHz
notation)
- Preliminary support for SiS, ARK (incl ICS5342 clock chip), RealTek, NCR,
Genoa GVGA, MX and ET3000 (how could this one have slipped my attention for
so long?) chipsets.
- (1.1.14) added "clockchip_x" option for icd2061a/ics9161.
- added warning to Makefile about syntax change.
- more grabmode tweaking... this will never end. Again some improvements for
grabbing "extended" modes.
- (1.1.15) removed bug in S3 HStext debug code (garbled output).
- added better error message when old FontSlect syntax is used.
- timeout bug removed in wait_vsync.c (it reported a Vsync timeout, and then
got stuck anyway).
30 NOV 1995 - Version 1.1
- (1.0.1) removed null-pointer writing bug that caused ELF and DOS code to
segfault in DefaultMode scanning code. Thanks to "checker"!
- `make newinstall' now copies old textconfig file to .orig file.
- (1.0.2) updated S3 clockchip code to XFree86 3.1.2, so now IBM RGB
RAMDAC's are supported. But they don't work....
This needs more work: - check for TI302X init code
- beta test IBM RGB, STG1703 and TI3026
- Cirrus clockchip stuff needs testing again
- set_clockchip_clock now uses kHz (as in XFREE3.1.2). This needs testing.
DANGEROUS!
- removed cut/paste bug in ICS5341 clockchip MClk programming.
- updated some docs (especially the INSTALL file was _old_).
- added DRAM speed options for S3 cards (actually sets M-parameter, whatever
that is). This allows S3 cards to use even higher text mode clocks.
- (1.0.3) Added S3 high speed text font loading. This can be enabled with
option "S3HSText" in config file. Now S3 can do over 80 MHZ in text mode!
- (1.0.4) experimental ali (Advance Logic) chipset support. Untested.
- (1.0.5) sent output from font loader through pipe to print it later on S3
cards with HSFont enabled: no more corrupted screens.
- better checking if HSText is allowed.
- added horizontal wrap detection to grabmode.
- (1.0.6) changed set_V_timings to avoid null sync width in vertical total
double mode.
- added bordercolor option. Nobody needs it, but I wanted to do something
simple for a change...
- due to popular (?) demand: programmable underline position.
- bug in resize_active_vts() tried to resize /dev/tty0../dev/tty15 instead
of /dev/tty1../dev/tty16 (some ELF distributions don't have /dev/tty0)
- added OTI67/77/78 chipset support. Easy. Untested.
- (1.0.7) fixed bug in IBM RGB RAMDAC code (had to multiply freq's by 1000)
- added AL2101 chipset (very close to generic ALI chipset).
- generalized and expanded number-of-clocks checking code
- more popular demand (some italian guy...): shorthand "stm" for SVGATextMode.
- status line of STM now adds "D" when doublescan mode is selected
- resize_active_vts() bug fix introduced another bug. Replaced that one with
another one...
- unlocked IBM RGB clock selection (locked by XFREE code), so soft reboot can
restore screen. Doesn't seem to work...
- (1.0.8) cleaned up VT_RESIZE code a little. STM now tries to free some more
memory (*1, *2 and *3) if VT_RESIZE fails. More chance of success.
- added check for 16k chars per screen in parse_modeline.c.
- added mention of MClk programming in manuals.
- doc/REPORT file removed. Seems not very useful anymore. Anyone disagree?
- always reset ClockDiv2 bit when option not defined. This caused trouble
on systems where this option was disabled after switching to such a mode.
13 OCT 1995 - Version 1.0 Friday the 13th release!
I hope you aren't superstitious!
- added MClk programming to ClockProg. Should increase text mode max clock
(and graphics speed...). BUT: VERY DANGEROUS... (hard crash).
- killed outw() argument order bug in Video7 clock setting code.
- a bunch of small changes (mainly to Makefile) by Ron Sommeling
<sommel@sci.kun.nl>
- moved clock setting block from SVGAtextMode.c (and ClockProg.c) into
setclock.c.
- split Run_extern_Prog into separate .c file
- MClk programming on some clockchips now supported (higher text mode clocks
possible).
- merged clockprobe and grabmode into one program
- split up font parsing into parse_font.c
- removed some obsolete code from WDC90CXX chipset unlocking routine.
- safe_wait_vsync created in wait_vsync.c
- used safe_wait_vsync and SCREEN_OFF before programming VGA chip (avoiding
font corruption and system hangs -- hopefully) in ClockProg and STM
- ClockProg can use different config file (/etc/ClockConfig) if it exists
(will use /etc/TextConfig otherwise)
- PERROR now does SCREEN_ON before exiting
- moved some compile-time options from severeal .h files to the Makefile
- changed dependency checking/making in Makefiles (now uses .depend files)
- investigated "unstable Cirrus Clock" problem. Resulted in
contrib/Cirrus_ClockChip.patch
- split setclock.c into setclock.c, std_clock.c & clockchip.c
- max WDC90C3X clock now 50 MHZ (90 was way too much. I should have known
only Tseng could do this...)
- option definitions moved from confdefs.h to parse_opt.h
- added "SyncDisks" option: does a sync() before going to the hardware.
- set/getpalette can input/output in hex now
- DO_SYNCRESET option added to Makefile: this does a synchronous reset of
the timing sequencer before changing the pixel clocks. Was always on by
default, but crashed some ET4000's. now off be default.
- changed patch to S3gendac.c code for ET4000 gendac so that pixel
multiplexing becomes possible (min_n2 changed from 0 to 1)
- changed most Outb_xxxx functions to use outw() instead of 2 outb()'s
- setpalette can now program a "standard" palette without external
parameters or input. This also uncovered a bug in the built-in standard
palette of set80 (data was interpreted as octal...Yuck!)
- "browse" option added to contrib/scripts/STMmenu
- grabmode (in fact modedata.c) has some intelligence built-in to properly
display modes with 1024 lines (1280x1024 modes were shown wrong, because
they used chipset-specific registers)
- probeclocks.c started: XFree86-like clock probe. Not finished yet...
- grabmode _should_ now properly detect ALL CGA-emulated modes (320x200, etc),
even packed pixel modes. Documentation updated/expanded.
- attempt at supporting interlaced modes (for S3 only) Do NOT try this at
home! It doesn't work yet (_can_ it work for text modes ???).
- cleaned up TODO file (who cares...)
- most "Get_..." functions are now inline. Probably not very useful.
- underline position is now also programmed. Some people complained about
missing underline, or strike-through instead of underlining when using
STM. They were right, of course.
Also added "NoUnderline option".
- Added "DoubleScan" attribute for text mode config lines. Anybody needs a
64-pixel high font?
- corrected little bug in START_HBLANK code (set_ and get_): was wrong by 1
char. Need this for correct blanking programming.
- Started correct blanking programming: 1 char before and one after the
active display. This is useful for those who want a nice colored border
around their screen using the overscan color register. If the blanking is
not programmed correctly (as in X), changing the overscan register makes
ALL screen colors change.
This was a VERY nasty thing to do. It seems like most VGA programs don't
bother, and set it to the same position as the sync. It's not easy to get
it right.
- Needed to change the same two a little more for correct blanking
programming: it seems like both start and end of HBLK must be decremented
by 1 before using them. This doesn't show up in any VGA textbook
(Ferraro,...), but the actual display shows otherwise.
- added font dumper/undumper programs in contrib/changefont.
- disabled S3's "high speed text display font fetch mode" until we know what
it does and how to use it. This caused trouble on S3 911 etc.
- added S3 high speed text font writing code. Experimental, and non-functional!
- more tweaking on grabmode to make it mode reliable when detecting CGA modes
and 16/24 bit modes
- manual file for grabmode written. Finally.
- first shot at DOS version of STM, thanks to invaluable help from Stephen
Lee. (Well, in fact, he did all the work). The most important changes are
the ttyresize.c file (DOS resizing code), and some include file changes to
enable DOS compilation.
- Removed argv[0] printing in debug output. Useless, and messy.
- All SVGATextMode code now compiles under DJGPP 2.0 in DOS. I didn't say it
all works perfectly! This is unsupported work.
- added ioctl(</dev/console>, KDSETMODE, KD_TEXT) to SVGATextMode.c as well
(was only in set80.c). Better textmode restoration ???
- more MAN-file updating.
- STM now always does a pre-1.3.3 VT_RESIZE, and if it's a 1.3.3 or newer,
it does a VT_RESIZEX also. Maybe this avoids trouble until those kernel
guys make this a bit less obscure.
- added DOSEMU chapter to doc/PROBLEMS
- fixed makefile so it doesn't re-compile all stuff when binaries are in
source tree, but .o files are not.
- added "backup" target to makefile
- SVGATextMode is now oficially under the GPL (finally -- I read the damn thing)
- final (you must be kidding...) man-file cleanup.
6 JULY 1995 - Version 0.9
- Added "-t" option to use a different TextConfig file (useful e.g. for the
clock probing script in contrib)
- explained "character bandwidth" in doc/PROBLEMS file.
- fixed bug in WDC90cXX code (clock selection bit #2 was wrongly programmed)
- another bug in WDC code: registers were not unlocked: some modes only worked
after X unlocked the VGA registers.
- used CirrusSetClock from XFREE package instead of previous (custom) version.
- added "restore all VGA registers" option to set80. More robust, but not
perfect, since SVGA regs are not restored...
- added "-m" switch. When VT_RESIZE fails due to lack of memory, it doesn't
automatically try via a 1x1 screen anymore. You need to use "-m". This is
to avoid erasing all screens unannounced.
- split up README into README and INDEX.
- updated binary DOS version of grabmode.exe to 0.8.4. (actually Mr
Albanowski updated it...). This needed a few changes to the code here and
there: some double colons in the Makefile (for the grabmode.exe target),
removing "optopt" for the DOS version, and changing the general [][] array
references to more specific ones.
- man-files written for: SVGATextMode(8), TextConfig(5).
- first attempt at a separate clock program using the XFREE clockchip code
(useful for X, where ClockProg DOES restore text mode clock, and
ClockChip doesn't). Also attempt at getting ET4000w32p with GenDAC
working.
- included patches by Mr David Mosberger-Tang <davidm@cs.arizona.edu> for
Linux/Alpha. STM should now compile on Alpha systems without changing a
single byte to the sources nor the Makefile.
- split up grabmode.c into grabmode.c and modedata.c. Should be useful for
scanmode, etc.
- added "DAC" register set to set/getVGAreg.
- patched XFREE/common_hw/S3gendac.c so ET4000W32p with ICS5341 GenDAC now
works, too.
- changed clockchip code to allow clockchips for other-than-S3 chipsets.
- added "DacSpeed" line + default maximum clocks for all chipsets.
- scan-mode now also takes clock limit into account.
- included a patch for ET4000 W32p X-server that solves mouse-garbage
problem on PCI cards.
- replaced 49 (!) misspelled words "usefull" to "useful"... That says
something about the program itself... ;-) (thanks David!).
- separated clock related config file parsing into a separate file
(parse_clock.c). To be used later in a "generic" clock program for XFREE.
Users of XFREE with a ClockChip should appreciate this (correct
restoration of textclock).
- also separated parse_modeline from configfile.c into separate .c file
- moved Hsync/Vsync_range parsing to main(), so it doesn't have to be done
in several places over and over.
- moved mode validation functions into validate.c
- mode validation now also tests if the required clock is in the clocks
line. (useful for "scan" mode).
- fixed bug in grabmode (actually bug in Get_xxx_End() functions in
vga_prg.c) that caused wrong blanking values to be displayed.
- added "-c" option to grabmode: don't probe for pixel clock.
- "slow clock" in grabmode is now non-fatal, 0 MHz is returned as probed
clock value
- all mode parameters (either from mode line or grabmode) use same struct
for data. I think that's called "unification".
- "DefaultMode" command introduced in config file
- mode selection tool using dialog and the enhanced "scan" mode of STM:
STMmenu. This obsoleted the 2 previous ones "grepmodes" and
"textmodeselect".
- added detection for some special features in grabmode (doublescan, ...)
- implemented VT_RESIZEX from kernel version 1.3.3 and up.
- also skipped tty resizing for kernels >= 1.3.3 (should be done by
VT_RESIZE now)
- compilation now automatically selects to allow or disallow resizing,
through checking of availability of VT_RESIZE call.
- changed ClockProg into a generic VGA clock programming tool, that can be
used either in SVGATextMode itself, or in XFREE, to support yet
unsupported clockchips in X (notably ET4000 genDAC), AND to allow for
sub-25 MHz modes in X for ALL cards (DOOM!), and not just those with a
clockchip. This will also smoothe the interaction between X and
SVGATextMode, since they then use the same clock programming code. Also
useful for Clockchip-cards in X, to allow for correct text mode clock
restoration.
- passed CFLAGS from top level to XFREE sub-Makefile
- workaround Cirrus Clockchip bug (unstable clock at 58 MHz --> use 57.9).
2 MAY 1995 - Version 0.8
- changed VT_RESIZE call to use /dev/console instead of stdin. This avoids
having to redirect stdin when running SVGATextMode remotely.
- moved lower-level functions for video timings into vga_prg.c,
SVGATextMode.c now uses more higher-level functions.
- removed "bug" that limited modes to max 1024 lines per screen. Now up to
2048 lines are possible. This is all a standard VGA chip can do.
- made division by 2 feature an OPTION instead of a default. This should
avoid problems with not-so-standard VGA cards. Is now default DISABLED,
instead of enabled in previous versions.
- Amazing... one more dead bug: now the Ti3025 clock chip code was wrong.
(it was passed the desired frequency in MHZ, and it needed KHZ).
AND the XFREE/xfree_compat.c file used the opposite definition of outb(),
so you got a segmentation fault there. Why do people use TWO different
(and completely incompatible) versions of outb() ???
But... it STILL doesn't work.
- Michael Shields contributed a better function to free some RAM to avoid
VT_RESIZE from giving an "out of memory" error. This time it should work a
lot more reliably than what I had done too quickly... But still not
perfect.
- he also fixed a bug that made some font sizes NOT mentionned in the font
selection table be assigned the wrong font.
- changed PDEBUG, PWARNING, etc so they can be used as functions (';'
allowed at end).
- added contrib/loadfont. A better alternative to setfont written by Albert
Cahalan <albert@ccs.neu.edu>
- added two more register unlocking bits to S3 chipset. This COULD have
caused problems (VGA clock and 8/9 pixel wide char selection COULD be
locked from BIOS).
- changed argv[0] processing of set/getVGAreg and set/getpalette so they
select between set/get function correctly.
- put all tty-resizing code in a separate file (ttyresize.c).
- removed the ICD2061 setting programs in the contrib directory once again.
Hopefully nobody minds.
- added a "rescue" program "set80" that programs the VGA 80x25 mode, without
options or configfile. Useful for when you made a mistake using
SVGATextMode...
- moved mode line parsing code from SVGATextMode.c into configfile.c. Needed
for the mode line scan function.
- changed TRUE/FALSE definition, and moved it into misc.h for ALL sources.
also used boolean assignments where possible (instead of "if (...)
bool=TRUE; else bool=FALSE;").
- added '-s' option to scan config file for all possible mode lines.
- added '-v' option to disable mode validation (H/V frequency checking)
- changed sync checking code to allow for sync range validation in "scan" mode.
- changed explicit comparing with "NULL" to implicit.
- Keith Owens (kaoss@singnet.com.sg) suggested a change to make SVGATextMode
show the resulting text mode BEFORE bailing out when the syncs are out of
range.
- Changed TextConfig file clock selection mechanism (again). Now you only
have to define a clocks line when it is actually NEEDED: for non-clockchip
S3 cards, cards needing a Clockprog, and all other cards except Cirrus.
- Fixed a problem with the ATI MACH32 card support. Those cards seem to need
bits 2 and 3 of the clock selection word exchanged. Weird. Now at least it
works for one of them.
- allowed checking of correct number of clocks in 'Clocks' line to be
switched of using #define statement in SVGATextMode.c. It seemed to bother
some people.
- Command-line switch '-c' (don't program clock) produced a warning line.
This too seemed to bother some people. Now it is a debug message (only shown
when debugging is enabled).
- Kris Coryn suggested doing a temporary VT_RESIZE to 1x1 if the new one
fails (because of a lack of memory) at first. This frees (almost) all of
the RAM needed for the current display, thus freeing more for a second
attempt. This is very succesfull! The only downside is that the screen gets
cleared in the process.
- changed resizing code to check for correct kernel version BEFORE doing
ANYTHING. Code hopefully a lot more clean now, too.
- replaced VT_RESIZE.note with a more up-to-date screen-resizing.note.
- added "-b" option to grabmode, to show the H/V-blanking registers.
- changed all absolute VGA register references that were dependent on Mono
or Color VGA to vgaIOBase. If everything went right, SVGATextMode should
now also work on monochrome displays (NOT tested!).
- cleaned out XFREE Makefile and xfree_compat.h so now the clock chip code
compiles cleanly. This is in anticipation of the XFREE 3.1.1 code import.
- XFREE 3.1.1 clock chips now included! But if they work ...?
- get/setVGAreg now also outputs in binary (useful for some bit-level
registers).
- split up configuration defines from vga_prg.h (they have nothing to do
with VGA hardware) into separate confdefs.h file. This also facilitated
the use of the "other" definition of outb() from common/compiler.h for
Ti3025 code in XFREE/xfree_compat.c.
- fixed bug that called random filename for ResetProg when it was not
defined.
27 MAR 1995 - Version 0.7
- You guessed it: I reversed the order of the HISTORY entries. Everybody
else seems to do it that way...
- updated some doc files. Removed obsolete doc/VT_RESIZE.note.
- changed external clock program calling to comply with XFREE3.x method: now
takes MHz as an argument instead of kHz.
- To be more compatible with the XFREE configuration, you must now ALWAYS
define a clocks line. Even when using a Cirrus card, a clock chip or an
external clock program. In these three cases, you can define ANY number of
clocks of any value on the clocks line (as long as they are within range
of what your VGA chip can handle).
This also burries another really idiotic bug that caused the Cirrus Logic
cards to fail with the message "Internal error: No clock programming
defined. Please complain to the author...". Which is what many people did :-(
- started writing the "doc/PROBLEMS" file. Contains some troubleshooting
tips, and some commonly encountered problems.
- added "-x" option to set/getVGAreg, producing pipeable hex output.
- removed tiny bug that produced incorrect error report when combining
command-line switches in one "-abcd..." line, with an incorrect one in
between. All programs had this bug.
- changed Makefiles (again). XFREE/libcommon_hw.a is now NOT deleted in
distribution, so people without an X installation CAN recompile without
missing header file errors. As long as they don't do a "make mrproper"..
- quick hack by Stephen Lee: textmodeselect. A dialog-based mode selector.
- !!!maybe, just maybe, the infamous ICD2061 programming problem is HISTORY!
it used to give a segmentation fault, due to its reading of IO address
0x200, which wasn't given IO permissions by SVGATextMode...
- Reuben Sumner provided a neat function to do AUTOMATIC VT resizing. So
now you don't need the "terminals" line anymore: this function finds out
by itself what VT's are active, and resizes them.
This has the added advantage that putting tty7 and tty8 in the "Terminals"
line doesn't activate them (which TIOCSWINSZ does), so X cannot use them,
and fails when starting up, telling you it cannot open /dev/tty9 (when you
have 8 VT's defined).
Note that this function will be overridden if a "Terminals" line is
specified. This will allow you to override the "intelligent" behaviour of
SVGATextMode, and define which VT's you want resized (and which not).
- Split up the huge "main()" in SVGATextMode.c into smaller functions.
- changed NO_VT_RESIZE into ALLOW_RESIZE. This looks less obfuscated.
- restructured terminal resizing code completely, using more functions.
- changed "open" mode for terminal devices from read/write to write only.
This avoids a permission problem on most machines when running SUID.
- changed check_bounds_realtime(), so it now allows for spaces in the
frequency strings (suggested by Stefan Dalibor).
- put the "diamond" clock program back in. I removed it a bit to early: the
ICD2061a clock code didn't work yet. Maybe now...
- started changing grabmode to make it correctly detect CGA-emulated modes,
and Hi/TrueColor modes. Not there yet.
- VGA IO permission stuff facelifted.
16 MAR 1995 - Version 0.6
- Added support for Video7 (Headland Technologies) chipset.
- changed calling of reset-program to only run when screen actually changes
size.
- more general path option fetching from config file (in preparation for
FontProg et al)
- changed option stuff (code readability improvement)
- fixed a bug in OPT_SWAP_HIBIT for WDC cards.
- !!! added automatic font loading. Major feature improvement.
- fixed a stupid bug that produced an error when using ClockChip ICD2061A.
- fixed ANOTHER stupid bug that found an incomplete clockchip line when
there was none defined, after having found a ClockProg line. I must have
had a bad day there...
- added flags: -r : don't run ResetProg
-f : don't run FontProg
-c : don't program clock (but do the rest)
- cleaned up message printing code. Less overhead now.
- programmable cursor position and size: lets you define what the cursor
looks like.
- added 'HorizSync', 'VertRefresh' parameters to config file, to define
monitor's limits, and avoid loading a mode that the monitor cannot handle.
Same syntax as in XF86Config.
03 MAR 1995 - Version 0.5
- Made clockprobe shut up about interlacing. Was wrong anyway. adapted doc
accordingly
- !!! added mode grabber, which outputs a TextConfig (or XF86Config)
compatible line describing the currently used text mode.
!!! DOS port of grabmode by Kenneth Albanowski. Now you can grab a DOS
text mode and import it into SVGATextMode. (grabmode.exe included)
- removed small bug in set/getVGAreg (setVGAreg didn't work when called with
a path)
- added doc/README.ET4000.AltClockSelect. A technical description of a
rather common problem with ET4000's: the "hibit" option and the different
methods of clock selection.
- added Sigma Legend (ET4000) option.
- added Cirrus DRAM speed options (SLOW_DRAM, MED_DRAM, FAST_DRAM, XFAST_DRAM)
- continued XFREE code importing effort. Now all kinds of clock chips on S3
boards could possibly work. Needs testing though.
Produces lots of warnings during compilation, so it's not sure it will work.
Will somebody test this, please? S3 Gendac tested: OK
- added "ResetProg" line to TextConfig. This calls a program or a script
after SVGATextMode ended succesfully.
- fixed obscure problem with ATI code (needed at least for ATI VGA Wonder V5)
19 FEB 1995 - Version 0.4
This time, lots of user feedback resulted in most of the new features.
Thanks to all those who contributed!
- Increased sync widths in many default TextConfig modes, after TONS of
complaints about many modes just not working. This should fix most
problems (but not all...)
- added a clock probing program "clockprobe". Useful for debugging on
"remote" video cards (i.e. those YOU have, but I don't). Used some fancy
stuff in there. Check the documentation!
- short docs for getpalette, setpalette, getVGAreg, setVGAreg, and
clockprobe added.
- added some more user-contributed text modes. If this goes on, the
TextConfig file will be a MegaByte within a few years...
- added (user contributed) Video 7 mercury external clock program.
- fixed bug in setpalette code (getpalette was wrong).
- now SVGATextMode uses the last matching textmode line, instead of the
first (suggested by Kenneth Albanowski. Thanks). This gives YOUR patched
lines precedence over the original ones, if your lines are at the end of
the TextConfig file.
- changed external clock program calling to comply with XFREE3.1 method
- made option parsing more generic, so now adding options for all cards is
easier.
- added option "ET4000_AltClockSel" for those ET4000 which won't work with
the normal clock selection code.
- "swap_hibit" option for WDC cards added (as in XFREE). This actuall swaps
bit 2 (not 3), so the name is rather misleading...
- added automatic sync polarity selection. Deleted most sync polarities from
default text modes in TextConfig (auto polarity should get most right).
- first attempt at using some XFREE code. This has been suggested by myself
and many others. It's a good idea, but probably not that simple. Let's
see...
Nothing works yet, but it'll come, one day.
- Put VT_RESIZE before VGA programming code instead of after, and added
malloc() to get some more free memory (avoiding "out of memory" error).
- !!! Now does terminal resizing automatically, using the terminals defined
in a "Terminals" line in the TextConfig file. All these terminals get the
equivalent of an "stty rows ... cols ..." sent to them when SVGATextMode
is run. All programs that support the SIGWINCH signal (more/less, joe,
emacs, ...) redraw their screen automatically into the new size. This is a
really nice feature!
- and finally: SVGATexMode now checks for the kernel version, and if it's
older than 1.1.54, it will not allow screen resizing. This seemed to be a
major problem for some people.
This will also make SVGATextMode MUCH less dangerous to use.
31 JAN 1995 - Version 0.3
- more cosmetics. Especially changing small functions into #defines. Is this
wise???
- added setpalette and getpalette, to set/get the color palette. Completely
useless, but is was lying around somewhere else, and this seemed a better
place for it to reside.
- ATI support tested and changed to make use of ALL 128 clocks available!
- multiple clocks lines supported.
- split up README into separate parts in doc directory. It was beginning to
get bloated.
- added conditional compilation for older kernels (#define NO_VT_RESIZE)
- Still 867 bugs undiscovered...
23 JAN 1995 - Version 0.2
- major cosmetic changes. And more will follow...
- more modular code ( = more .c files)
- added options (-n, -d and -h)
- changed define DEBUG into an option. Executable will be always bigger, but
who cares...
- added "setVGAreg" and "getVGAreg" for hacking purposes (formerly not
distributed, but written as lots of separate programs by Wouter Gadeyne)
- added support for Western Digital and ATI cards (but VERY much untested ...)
- removed 8/9 character switching bug.
10 Jan 1995 - Version 0.1 First release
- Xconfig-like configuration
- S3, Et4000, Trident and Cirrus Logic support.
each tested on just ONE of those cards...
- probably a TON of bugs.
