Re: [PATCH 1/1] os_find_datadir: search as in version 4.2

[Brian Norris]

Hello!

On Wed, Jul 15, 2020 at 11:57:14PM +0400, Marc-André Lureau wrote:
> Hi
> 
> On Wed, Jul 15, 2020 at 11:37 PM Peter Maydell <peter.maydell@linaro.org> 
> wrote:
> >
> > On Tue, 16 Jun 2020 at 10:19, Peter Maydell <peter.maydell@linaro.org> 
> > wrote:
> > >
> > > On Tue, 16 Jun 2020 at 00:00, Joe Slater <joe.slater@windriver.com> wrote:
> > > >
> > > > Always look for ../share/qemu then ../pc-bios when looking for datadir.
> > >
> > > Could you provide some more context, please? Why is this
> > > change useful; presumably we broke some setup in 5.0, but
> > > what exactly ?
> > >
> > > I'm guessing this might be a regression introduced by commit
> > > 6dd2dacedd83d12328 so I'm ccing the relevant people.
> >
> > Marco, Paolo: ping? Another user has just asked me the status
> > of this as they also ran into this regression in what directories
> > we search...
> 
> Thanks for the heads-up, I didn't see that bug/mail. Indeed, that
> commit assumed that either we run from a build directory or from an
> installed qemu. It seems this is hybrid approach, which I didn't know
> we supported. I'll check it.

Add one more to the pile! Chrome OS noticed this when upgrading to
5.0.0:

https://bugs.chromium.org/p/chromium/issues/detail?id=1114204#c8

I'd love to see this applied to a release.

I actually wrote basically this exact same patch and was about to submit
it, when I saw that this was already here. I've tested Joe's variant:

Tested-by: Brian Norris <briannorris@chromium.org>
Reviewed-by: Brian Norris <briannorris@chromium.org>

The Chromium bug report is public, so you can see details there, but
it's basically the same -- we sometimes run qemu from a path that's not
the same absolute path noted in ./configure. This is because we build
qemu to run within our SDK (a semi-containerized chroot), but we also
support running that same QEMU binary from outside the container, which
then may be at some arbitrary hierarchy on a developer's machine.

It might be wise to include a tiny bit more verbose of a code comment,
to prevent oversights like this in the future. I'm sure that could be
spliced in when the patch is applied though.

Regards,
Brian