Re: commit -- build fix


Subject: Re: commit -- build fix
From: Aaron Lehmann (aaronl@vitelus.com)
Date: Sun Apr 29 2001 - 00:21:31 CDT


On Sat, Apr 28, 2001 at 07:41:57PM -0500, Sam TH wrote:
> That shouldn't be neccessary. @GNOME_CFLAGS@ shouldn't ever be used
> if the build isn't a gnome one. Can you point out where it was being
> used? That's the real bug.

You're right that it's a bug.

GNOME_CFLAGS is always included in ABI_CFLAGS...

        if WITH_UNIX
                PLATFORM_LIBS=
                if WITH_GNOME
                        PLATFORM_CFLAGS=@GNOME_CFLAGS@ @GMODULE_CFLAGS@ @GTK_CFLAGS@
                else
                        PLATFORM_CFLAGS=@GMODULE_CFLAGS@ @GTK_CFLAGS@
                endif
        endif

So WITH_GNOME is being set unconditionally .... weird. According to
configure.in it hinges on $gnome, which is in turn set by a command
line arg and defaults to false. The real question is whether the
value of WITH_GNOME is actually being saved anywhere. A grep through
the top level of my build dir only gives:

        config.status:s%@WITH_GNOME_TRUE@%#%g
        config.status:s%@WITH_GNOME_FALSE@%%g

Sorry for the harsh remarks; it seemed as though the scripts weren't
set up to only include GNOME_CFLAGS if Gnome was enabled. Still, some
more testing would be appreciated. :)

Once we figure this out, please back out my change. It's a simple hack
that prevents WITH_GNOME from being defined is gnome-config doesn't
exist, but won't help if a user specifies --disable-gnome to configure
(or runs configure without --enable-gnome).

P.S. I HATE autoconf (and especially automake) and am currently
designing a minimalist replacement for them.



This archive was generated by hypermail 2b25 : Sun Apr 29 2001 - 00:21:39 CDT