On Wed, May 16, 2007 at 03:32:04PM +0200, Frank Lichtenheld wrote:
> This seems to be a compiler bug.
>
> Compiling with -O0 lets the test succeed
> (-O1 doesn't help, tested that too).
>
> As does adding some code in the "true" case in
> merge-recursive.c, ll_ext_merge:
> if (status < -ERR_RUN_COMMAND_FORK)
> ; /* failure in run-command */
> else
> status = -status;
>
> Compiling with gcc-4.2 from experimental didn't help.
Thank you Frank. Trying to reproduce on paer.d.o to track it down gives
random test failures due to an 'error while loading shared libraries',
e.g.:
* expecting success:
rm -f tmp one dir/two &&
git repo-config core.autocrlf true &&
git read-tree --reset -u HEAD &&
git apply patch.file &&
test "$patched" = "`remove_cr one | git hash-object --stdin`" ||
{
echo "Eh? apply without index"
false
}
sed: error while loading shared libraries: internal error: symidx out of
range of fptr table
Eh? apply without index
* FAIL 9: apply patch (autocrlf=true)
rm -f tmp one dir/two &&
git repo-config core.autocrlf true &&
git read-tree --reset -u HEAD &&
git apply patch.file &&
test "$patched" = "`remove_cr one | git hash-object --stdin`"
|| {
echo "Eh? apply without index"
false
}
Regards, Gerrit.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
