On Tue, Mar 17, 2020 at 09:40:00AM -0500, Eric Blake wrote:
> On 3/17/20 9:00 AM, Peter Maydell wrote:
> > On Tue, 17 Mar 2020 at 04:38, John Snow <jsnow@xxxxxxxxxx> wrote:
>
> > >
> > > block/qcow2-bitmap: use bdrv_dirty_bitmap_next_dirty (2020-03-12 16:36:46 -0400)
> > >
> > > ----------------------------------------------------------------
> > > Pull request
> > >
> > > ---------------------------------------------------------------
> >
> > Hi; this fails to compile with clang:
> >
> > /home/petmay01/linaro/qemu-for-merges/nbd/server.c:1937:1: error:
> > unused function 'glib_listautoptr_cleanup_NBDExtentArray'
> > [-Werror,-Wunused-function]
> > G_DEFINE_AUTOPTR_CLEANUP_FUNC(NBDExtentArray, nbd_extent_array_free);
> > ^
> > /usr/include/glib-2.0/glib/gmacros.h:462:22: note: expanded from macro
> > 'G_DEFINE_AUTOPTR_CLEANUP_FUNC'
> > static inline void _GLIB_AUTOPTR_LIST_FUNC_NAME(TypeName) (GList
> > **_l) { g_list_free_full (*_l, (GDestroyNotify) func); } \
> > ^
> > /usr/include/glib-2.0/glib/gmacros.h:443:48: note: expanded from macro
> > '_GLIB_AUTOPTR_LIST_FUNC_NAME'
> > #define _GLIB_AUTOPTR_LIST_FUNC_NAME(TypeName)
> > glib_listautoptr_cleanup_##TypeName
> > ^
> > <scratch space>:49:1: note: expanded from here
> > glib_listautoptr_cleanup_NBDExtentArray
> > ^
>
> Should we add -Wno-unused-function to CFLAGS when dealing with a version of
> clang that complains about that version of glib's headers? Is it fixed in a
> newer version of glib, where we could just backport a newer definition of
> G_DEFINE_AUTOPTR_CLEANUP_FUNC() that adds whatever annotations are needed to
> shut the compiler up?
>
> On IRC, danpb pointed me to libvirt's solution:
> https://libvirt.org/git/?p=libvirt.git;a=commit;h=44e7f029
>
> Maybe we just write our own macro wrapper around
> G_DEFINE_AUTOPTR_CLEANUP_FUNC which takes care of adding necessary
> annotations and use that instead (and our macro name might be shorter...)
My preference is to stick with regular glib functions/macros whereever
practical, rather than inventing QEMU replacements which add a knowledge
burden for contributors. That's why we moved to -Wno-unused-function in
libvirt.
I don't feel like -Wno-unused-function looses anything significant, as
the GCC builds will still be reporting unused functions which will
catch majority of cases.
Possibly we could figure out a patch for glib upstream that uses pragma
push/pop to squelch the warning ? They are quite receptive to patches
IME.
Regards,
Daniel
--
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org -o- https://fstop138.berrange.com :|
|: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
