Re: [PATCH] configure.ac: Fix FIONREAD check to work with gcc-14 on Solaris

[Sam James]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Stefan Kangas <stefankangas@gmail.com>
>> Date: Sat, 12 Oct 2024 08:20:52 -0500
>> Cc: Ali Bahrami <Ali.Bahrami@oracle.com>, Eli Zaretskii <eliz@gnu.org>
>> 
>> Alan Coopersmith <alan.coopersmith@oracle.com> writes:
>> 
>> > Before this fix, it would fail to build with gcc-14 with the error:
>> > conftest.c: In function 'main':
>> > conftest.c:265:11: error: implicit declaration of function 'ioctl'
>> >  [-Wimplicit-function-declaration]
>> >   265 | int foo = ioctl (0, FIONREAD, &foo);
>> >       |           ^~~~~
>> >
>> > Solaris documents ioctl() as being defined in <unistd.h>
>> > ---
>> >  configure.ac | 3 +++
>> >  1 file changed, 3 insertions(+)
>> >
>> > diff --git a/configure.ac b/configure.ac
>> > index 8a5ba7db3d1..947c2827b8e 100644
>> > --- a/configure.ac
>> > +++ b/configure.ac
>> > @@ -7117,6 +7117,9 @@ AC_DEFUN
>> >                        #ifdef USG5_4
>> >                        # include <sys/filio.h>
>> >                        #endif
>> > +                      #ifdef HAVE_UNISTD_H
>> > +                      # include <unistd.h> /* defines ioctl() on Solaris 
>> > */
>> > +                      #endif
>> >                      ]],
>> >                      [[int foo = ioctl (0, FIONREAD, &foo);]])],
>> >     [emacs_cv_usable_FIONREAD=yes],
>> > --
>> > 2.45.2
>> 
>> I guess this should be installed on emacs-30.  Eli, WDYT?
>
> I'd prefer on master.  Who knows what addition of unistd.h could cause
> to this test?

Note that the absence of the include means the configure check gives the
wrong result, it doesn't mean Emacs necessarily fails to build as a
result. It may be built with the wrong configuration.