Hello Jindrich,
On Tue, 2006-11-28 at 13:21 +0100, Jindrich Novy wrote:
IMO only removal of the fallback will prevent
the infinite loop in any case as it shouldn't call mc_mkstemps() at all.
That cure seems worse than the disease. Isn't the real problem the fact
that mc_mkstemps() blindly concats tmpdir to the prefix instead of
testing if mc_tmpdir() succeeded? Why not abort mc_mkstemps() if
mc_tmpdir() returns "/dev/null/"?
What about the attached (untested) patch?
By the way, could you please add the -p option to your diffs?