Then gnulib-tests fail: FAIL: test-chown (exit: 134) test-chown.h:106: assertion failed -- FAIL: test-lchown (exit: 134) test-lchown.h:114: assertion failed -- FAIL: test-fchownat (exit: 134) test-chown.h:106: assertion failed -- FAIL: test-utimens (exit: 134) test-utimens.h:43: assertion failed
The problem here is that precision settable by syscalls is not the same as default precision of timestamp. Precision of kernel generated timestamps is given by "sysctl vfs.timestamp_precision". See i.e. http://lists.freebsd.org/pipermail/freebsd-current/2007-June/073865.html It collides with our workaround of kernel didn't set the modification time with O_TRUNC. I have to check whether it is still needed and either drop our workaround (part of libc) or fix it. Please could you in mean time, only for debian, apply extra check bellow for determining needed delay in tests. Petr --- nap.h +++ nap.h @@ -60,8 +60,10 @@ ASSERT (stat (BASE "tmp", &st2) == 0); ASSERT (unlink (BASE "tmp") == 0); } - if (! (st1.st_mtime == st2.st_mtime - && get_stat_mtime_ns (&st1) < get_stat_mtime_ns (&st2))) + if (! ((st1.st_mtime == st2.st_mtime + && get_stat_mtime_ns (&st1) < get_stat_mtime_ns (&st2)) + && (st1.st_ctime == st2.st_ctime + && get_stat_ctime_ns (&st1) < get_stat_ctime_ns (&st2)))) delay = (st1.st_mtime & 1) ? 1000000 : 2000000; } usleep (delay); -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org