Dne 8.11.10 18:33, Eric Blake napsal(a):
On 11/08/2010 10:27 AM, Joerg Schilling wrote:
Petr Sumbera<[email protected]> wrote:
What should HAVE_PARTLY_WORKING_GETCWD be?
If the pointer to the first argument of getcwd() is a NULL pointer, the result
is unspecified. Purple smoke could therefore be a "valid" result ;-)
The proposed patch is to a gnulib file, so the report should be sent to
bug-gnulib. However, I fail to see why the patch is necessary - gnulib
should already be arranging to supply a getcwd() replacement that
automatically mallocs when the second argument is NULL, per glibc
behavior, even on platforms where that behavior is normally undefined as
per POSIX. So the real bug is determining why gnulib's getcwd()
replacement is not properly kicking in for your platform.
I didn't investigate it at this moment any further. I just know that
with the patch test 56 passes. Without it doesn't.
But I'm more interested about others failures. They happen only on sparc
and I don't think there were any problems with version 1.24.
Thanks,
Petr
