Re: [Bug-gnulib] utimes.m4 -> utimes-null.m4; new utimes.m4 (merge from coreutils)

[Paul Eggert]

Bruno Haible <address@hidden> writes:

> And what is the recommended action in the code when HAVE_UTIMES && !
> HAVE_WORKING_UTIMES ? I'm asking because of lib/utime.c and
> lib/copy-file.c.

utime.c should be OK, since it depends only on HAVE_UTIMES_NULL, and I
utimes (file, 0) works even if the bugs are present.

copy-file.c prefers utime to utimes, so it should be fine too.
I don't know of any host that lacks utime and has a broken utimes.

In the general case, if HAVE_WORKING_UTIMES is not set, I'd suggest
avoindg utimes and using utime instead.  That is what
coreutils/lib/utimens.c does.

Perhaps we should make utimens a gnulib module. tar will need
something like it once tar adds support for subsecond timestamps.