-----BEGIN PGP SIGNED MESSAGE-----
=?iso-8859-1?Q?Jerker_B=E4ck?= <jerker.back@xxxxxxxxx> writes:
> Well, fopen is one of the few I/O functions which is not part of the NT
> POSIX API (closed Windows source), instead it's implemented by the Interix
> LIBC. So, a replacement is very likely to work. I doubt however I can do
> better than the Interix developers.
> FILE * fopen(char *path, const char *mode);
> FOPEN_BINARY_READ macro seems to be consistent with my docs.
> So, apparently I have to start a debug session to solve this.
> You wouldn't have a suggestion of a shell command that triggers the bug
> directly, so I don't have to click my way through the whole test?
my normal process is to hack a copy of sanity.sh so that it does an exit
right before it runs the command for the teset which is at fault. So, in
this case, hack in the exit command right before the 'dotest' with the
label which is failing. For your case, I might try something like this:
... elided start of src/sanity.sh file copy...
ls -l fileX
# Writes actually cause symlinks to be resolved.
dotest rcslib-long-symlink-3 "$testcvs -q ci -mwrite-it" \
"Checking in fileX;
new revision: 1\.5; previous revision: 1\.4
... elided rest of src/sanity.sh file copy...
This will print out the environment and the current working directory so that
I can more easily set things up to do what I need to do when I run the
debugger on the 'cvs ci -mwrite-it' command.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (FreeBSD)
-----END PGP SIGNATURE-----