Re: [Qemu-devel] [PATCH] qemu-ga: do not run qga test when guest agent disabled

[Michael Roth]

Quoting Peter Maydell (2016-04-19 17:08:03)
> On 19 April 2016 at 23:01, Michael Roth <address@hidden> wrote:
> > Quoting Yang Hongyang (2016-04-19 02:39:13)
> >> When configure with --disable-guest-agent, make check will fail with:
> >> ERROR:tests/test-qga.c:74:fixture_setup: assertion failed (error == NULL):
> >>  Failed to execute child process "/home/xx/qemu/qemu-ga" (No such file or
> >> directory) (g-exec-error-quark, 8)
> >> make: *** [check-tests/test-qga] Error 1
> >>
> >> This check was commented out by bab47d9a75a. I think that was by
> >> mistake, because the commit message of that commit didn't mention
> >> this change.
> >>
> >> Signed-off-by: Yang Hongyang <address@hidden>
> >> Cc: Gerd Hoffmann <address@hidden>
> >> Cc: Michael S. Tsirkin <address@hidden>
> >> Cc: Michael Roth <address@hidden>
> >> Cc: Paolo Bonzini <address@hidden>
> >
> > Thanks, applied to qga tree:
> >   https://github.com/mdroth/qemu/commits/qga
> >
> > As much as I'd like to get this fixed for 2.6, the net effect, thanks to
> > the inadvertant commenting out of qga test case in bab47d9a75a, is that
> > the qemu-ga unit test currently gets skipped during make check. Given
> > RC3 is going to be tagged soon, and afaik is the last RC, I'm not
> > sure I would consider this enough of a blocker to send a last-minute
> > pull.
> >
> > Peter: if you think there's still a window to get this in let me know
> > and I'll send the pull immediately. But for now I'll queue this for
> > 2.7 and for stable.
> 
> Well, I'm not planning to tag RC3 til Thursday, so you have time
> in that sense. Whether it's worth putting into RC3 I leave to
> your judgement (it sounds like the only effect of not having it
> is "there's a test case we could be running that we don't run" ?)

Yes. Although the patch fixes a more serious build/make check issue
with --disable-guest-agent, the inadvertant commenting in bab47d9a75a
(which was probably to work around that bug) masks the build failures
by unconditionally disabling the test.

Since disabling the unit test is a late regression, I think it's
probably worthwhile to try to fix as long as it isn't holding up
the release. Will send a pull shortly.

> 
> thanks
> -- PMM
>