[Bug time/30200] time sometimes appears to go backwards

["adhemerval.zanella at linaro dot org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=30200

Adhemerval Zanella <adhemerval.zanella at linaro dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |adhemerval.zanella at linaro dot o
                   |                            |rg

--- Comment #6 from Adhemerval Zanella <adhemerval.zanella at linaro dot org> ---
(In reply to Bruno Haible from comment #5)
> For comparison, the test program succeeds (even 100000 consecutive runs) on
> - musl libc (Alpine Linux 3.17)
> - FreeBSD 13.1
> - NetBSD 9.0
> - OpenBSD 7.2
> - macOS 12.5
> - Solaris 11.4
> - GNU/Hurd

If I recall correctly, we used CLOCK_REALTIME_COARSE for time to mimic Linux
time syscall kernel behavior (which is implemented in a couple of old ABI and
used historically).  Recently on 64 time_t time, we changed time to use
clock_gettime and kept the same clock for compatibility.

I agree that it makes sense to use the same clock for time and timespec_get,
however, using CLOCK_REALTIME would make the vDSO optimization for x86_64 and
powerpc64 unusable (not a big deal IMHO). And another question is whether we
will need to add a compact symbol to time (and I also do not think it would
make sense either).

-- 
You are receiving this mail because:
You are on the CC list for the bug.