I. CHECK YOUR INPUT DATA FOR AUTOMATA. I.1. Problems with sorting (for fsa_build - fsa_ubuild should work anyway). Check the value of your environment variables LANG, and LC_* (i.e. LC_ followed by any string). LANG should be set to POSIX, all LC_* - probably undefined. In bash, you see all variables with the command export with no arguments. You can check whether your sort does the job, by giving it data like e.g.: A a Abakus abakus über Uffer If `über' is neither the first, not the last on the list produced by sort, or if the rest of words is not in the order `A', `Abakus', `Uffer', `a', `abakus', then your sort is broken - check your variables again. The solution to this problem is to precede invocation of sort with LANG=C, e.g.: LANG=C sort -u mylist | fsa_build > mylist.fsa I.2. Duplicates (for fsa_build - fsa_ubuild should work anyway) Use sort -u to get rid of duplicates. If you use sort without options and uniq, use sort first, and then uniq, not the other way round. I.3. Empty lines. Some software you use can put them e.g. at the end of your file, so that you don't see them. Use e.g. grep to find them and to filter them out. I.4. Additional invisible characters in your data. Files copied from Windows or Mac may have different/additional characters at the end of a line. Use appropriate software for the conversion. Also note that some scripts expect horizontal tabulation characters (Ctrl-I or HT) as default separators. If you give them spaces instead, it looks the same, but it is not the same. If you copy and paste parts of data directly from the screen, your HTs will be converted to spaces. I.5. Special characters in your data. Some characters are used by default for special purposes. These are "_" (the filler, an empty character that is ignored by the software), and "+" (annotation separator). They can be changed with run-time arguments. If your data contains such characters, either convert then to something else, or specify different fillers and annotation separators for appropriate programs and scripts. I.6. Check the format of your data. Read the file README to learn how to prepare the data. The package contains a number of scripts to help you, but the input to the scripts also has to be in some (very simple) format. I.7. Isolate data that causes trouble. If any of the above does not work, try to find the smallest amount of input data that reproduces the error. Use binary search to cutt off some lines from the end of input file, from the beginning, and after each line. If you have bash, you can use the script ie1 provided in the package. Change the contents of the function f1() to reflect the kind of errors you are looking for. II. Check the compilation. Always issue: make clean before you recompile the package after having changed some compile options. If you don't, you're asking for trouble. Also, some of the programs in the package require some compile options with particular combinations of arguments. E.g. you cannot use fsa_hash at all if the package is not compiled with NUMBERS. Some versions of gcclib contain bugs. Use LOOSING_RPM compile option. If this does not work, turn optimization off (delete -O2 from CPPFLAGS). III. Check the invocation of programs and scripts. Programs have their man pages, scripts are described in the file README. IV. Send a bug report. If it is a bug, I'd like to hear about it. Send it to jandac at eti.pg.gda.pl.
