Re: [PATCH] meson: generate qemu-version.h with a custom target

["Daniel P. Berrangé" <berrange@redhat.com>]

On Thu, Nov 05, 2026 at 06:34:31PM +0000, Peter Maydell wrote:
> On Thu, 5 Nov 2026 at 18:15, Yonggang Luo <luoyonggang@gmail.com> wrote:
> >
> > configure_file() writes qemu-version.h at setup time. That result is
> > not a ninja edge, so a missing header is never rebuilt and compiles
> > that include it do not order on it. Generate the header with
> > custom_target and list it in genh. scripts/qemu-version.py reads
> > scripts/qemu-version.h.in and substitutes -version. An empty
> > pkgversion drops the QEMU_PKGVERSION line.
> >
> > Link: https://lore.kernel.org/qemu-devel/CAFEAcA8bGODxRfRzrimhGN1v2HtcWvApRsyucemZbMUUBVkYwA@mail.gmail.com/
> >
> > Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
> 
> If we're going back to having an external script, the other
> way to fix this would be to just revert ab78ac76f44a12 and
> maybe fc35f2438770, since their commit messages suggest the
> main aim was to not have an external script.

I'm not clear what problem there was with an external script that these
were solving ? These also seem to have appeared in a pull request
without any prior posting for review, which is rather shortcutting our
normal dev process.

Personally I'd suggest we revert and then simplify the version handling
in a different way:

 * Remove the VERSION file and put the number in meson.build directly
 * Emit QEMU_PKGVERSION into build/config-host.h
 * Remove qemu/version.h and build/qemu-version.h

This avoids any need for rules to generate build/qemu-version.h
and config-host.h already has logic to regenerate itself when
meson.build changes.

With regards,
Daniel
-- 
|: https://berrange.com       ~~        https://hachyderm.io/@berrange :|
|: https://libvirt.org          ~~          https://entangle-photo.org :|
|: https://pixelfed.art/berrange   ~~    https://fstop138.berrange.com :|