On 09/09/2014 12:09 AM, Dylan Cali wrote: > On Mon, Sep 8, 2014 at 10:04 AM, Dylan Cali <calid1...@gmail.com> wrote: >> On Sep 8, 2014 9:27 AM, "Eric Blake" <ebl...@redhat.com> wrote: >>> Does this work for all supported versions of gcc? Or do you need to make >>> it conditional on new enough gcc (it's okay if warnings have to be >>> disabled to compile with older gcc, but not okay if the way to disable >>> warnings for newer gcc causes compilation failure in older gcc). >> >> Yep you're right, looking at some other code of mine I have a conditional >> checking for gcc > 3.7 before doing the push/pop, so it looks like that is >> the min version for this pragma. So wrap the pragma in a conditional, and >> add a conditional in the Makefile turning off warnings altogether for gcc <= >> 3.7? > > Ok cool, it looks like this is mostly already implemented. First, I > was wrong about the version of gcc that supports these pragmas, it's > 4.6 [1]. And gnulib already uses the diagnostic pragmas elsewhere, so > I mimicked the conditional format used in other files. Finally, the > makefile conditional is already done in the coreutils configure.ac > [2]. > > I've attached a patch with the updates, let me know if you see any > further issues. > > Thanks, > Dylan > > [1] https://gcc.gnu.org/ml/gcc-help/2011-01/msg00113.html > [2] https://github.com/coreutils/coreutils/blob/master/configure.ac#L97
I don't think we need to worry about the pragma as it was a valid warning in this case. To tell the compiler we want to discard the result in this context we can use ignore_value() or the more appropriate (void) cast in this case. Also it's best to leave the prototype in the tests. I've adjusted and pushed your patch accordingly at: http://git.sv.gnu.org/gitweb/?p=gnulib.git;a=commitdiff;h=5549ef8 thanks! Pádraig. p.s. It's marked trivial for licensing reasons.
