This is the mail archive of the
gsl-discuss@sourceware.cygnus.com
mailing list for the GSL project.
Re: Testing GSL with DJGPP
- To: pavenis at lanet dot lv
- Subject: Re: Testing GSL with DJGPP
- From: Brian Gough <bjg at network-theory dot co dot uk>
- Date: Sun, 28 May 2000 12:16:25 +0100 (BST)
- Cc: gsl-discuss at sourceware dot cygnus dot com
- References: <39300444.23726.77E304@localhost>
Thanks for the bug report.
I'll add the "b" option to fopen() for binary files and substitute
$(LN_S) for ln -s in gsl/Makefile.am. 8+3 filenames don't seem
feasible with our naming scheme, so we'll have to let that slide. If
you discover any more facts about the crashes of specfunc/test.exe let
us know.
regards
Brian Gough
pavenis@lanet.lv writes:
> Tried to test current CVS version of GSL with DJGPP
> (target=i586-pc-msdosdjgpp).
>
> At first there were some problems to build it at all and to run tests:
>
> 1) Makefile.in and gsl/Makefile.in uses 'ln' directly, so one cannot replace it on
> systems where are no symbolic links. Editted files to workaround problem
>
> 2) Many tests opens binary files without specifying binary mode.
> So tests breaks for MS-DOS and perhaps Windows related targets.
> Editted files and put fopen ("test.dat" , "[rw]b") were needed.
>
> Here is list of changes:
>
> sh-2.03$ grep -n fopen\ \(\"test[.]dat `find . -name '*.c'`
> ./block/test_complex_source.c:76: FILE *f = fopen ("test.dat", "wb");
> ./block/test_complex_source.c:94: FILE *f = fopen ("test.dat", "rb");
> ./block/test_source.c:78: FILE *f = fopen ("test.dat", "wb");
> ./block/test_source.c:95: FILE *f = fopen ("test.dat", "rb");
> ./histogram/test.c:425: FILE *f = fopen ("test.dat", "wb");
> ./histogram/test.c:431: FILE *f = fopen ("test.dat", "rb");
> ./histogram/test2d.c:573: FILE *f = fopen ("test.dat", "wb");
> ./histogram/test2d.c:579: FILE *f = fopen ("test.dat", "rb");
> ./matrix/test_complex_source.c:149: FILE *f = fopen ("test.dat", "wb");
> ./matrix/test_complex_source.c:169: FILE *f = fopen ("test.dat", "rb");
> ./matrix/test_source.c:356: FILE *f = fopen ("test.dat", "wb");
> ./matrix/test_source.c:372: FILE *f = fopen ("test.dat", "rb");
> ./vector/test_complex_source.c:161: FILE *f = fopen ("test.dat", "wb");
> ./vector/test_complex_source.c:177: FILE *f = fopen ("test.dat", "rb");
> ./vector/test_source.c:294: FILE *f = fopen ("test.dat", "wb");
> ./vector/test_source.c:307: FILE *f = fopen ("test.dat", "rb");
> sh-2.03$
>
> 3) There are huge number filename conflicts for DOS. As I run tests
> in DOS box under Win98 SE without setting NameNumericTails=0
> then I didn't get related failures
>
> 4) Only testsuite problems I met after that was in directory
> specfunc:
> test.exe crashes in different places and different ways (SIGSEGV,
> SIGILL, ...) so I suspect some memory corruption (or use of uninitialized
> data). Unfortunatelly (according "Murphys law") it doesn't crash at all
> when I'm trying to run it under gdb (in this case all tests suceeds)
>
> Andris
>
>