Re: FYI, new, rare failure on rawhide

[Paul Eggert]

Possibly a bug in the file system or kernel?
An earlier part of the code does this:

    /* lines 57 and 58 of test-utimens.h */
    ASSERT (func (BASE "file", ts) == 0);
    ASSERT (stat (BASE "file", &st1) == 0);

and this relies on st_ctime being marked for update in the first
line, and being actually updated before the second line takes effect.
This is pretty easy to get wrong, and many file systems do get similar
things wrong, particularly under high load.  If my guess is right,
it should be easy to verify by instrumenting those lines.

There is a similar issue here:

    /* lines 98 and 99 (/
    ASSERT (func (BASE "file", ts) == 0);
    ASSERT (stat (BASE "file", &st2) == 0);

Of course this is just a long-distance guess....