Re: utimens and non-standardized futimesat [was: coreutils-6.11-1 in release-2 area]

[Eric Blake]

Bruno Haible <bruno <at> clisp.org> writes:

> Two nits (I'm really only nitpicking):
> 
> - You introduce two #ifs that have to be the same condition:
>     #if HAVE_FUTIMESAT || HAVE_WORKING_UTIMES
>     ...
>     #if HAVE_FUTIMESAT || HAVE_WORKING_UTIMES

And only so that C89 compilation will work when you request
HAVE_BUGGY_NFS_TIME_STAMPS.  On the other hand, if you have a new enough kernel
that supports futimens, you probably don't have to worry about buggy NFS time
stamps.

>   It is preferrable, for maintenance (I speak from experience with
vasnprintf.c...)
>   to have this #if only at one place. In this case, I would simply add a
>   brace group to the contents and reindent:

Done.

> 
> - Is it conceivable that a platform has
>     HAVE_FUTIMENS && (HAVE_FUTIMESAT || HAVE_WORKING_UTIMES)

Absolutely.  Cygwin 1.7.0 is one of these; I suspect newer glibc falls in this
group too.

>   ? In this case, gcc will warn about dead/unreached code. I would change the

Also done.  The resulting patch is bigger (mostly indentation), but functionally
equivalent.  I pushed:
http://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commitdiff;h=faeb3e6

-- 
Eric Blake