Zack Weinberg <invalid.nore...@gnu.org> writes: > Follow-up Comment #1, sr #110740 (project autoconf): > Version checks are always fragile, but writing a functional probe for > properties related to race conditions can be extremely difficult. Do you have > any ideas for how we should go about it?
I think autoconf is being overambitious by promising to detect race-free-ness. You can't really. Having said that, it appears to me that POSIX has required race-free-ness since the nineties, so do you really need to? Just check that the -p option is accepted, and call it good. regards, tom lane