Re: [PULL 0/6] Egl 20211105 patches

[Philippe Mathieu-Daudé]

On 11/8/21 10:22, Daniel P. Berrangé wrote:
> On Mon, Nov 08, 2021 at 09:17:19AM +0100, Philippe Mathieu-Daudé wrote:
>> +Thomas & Daniel for Travis-CI
>>
>> On 11/8/21 09:12, Mark Cave-Ayland wrote:
>>> On 05/11/2021 18:49, Philippe Mathieu-Daudé wrote:
>>>> On 11/5/21 19:26, Philippe Mathieu-Daudé wrote:
>>>>> On 11/5/21 18:13, Richard Henderson wrote:
>>>>>> On 11/5/21 7:30 AM, Gerd Hoffmann wrote:
>>>>>>> The following changes since commit
>>>>>>> b1fd92137e4d485adeec8e9f292f928ff335b76c:
>>>>>>>
>>>>>>>     Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream'
>>>>>>> into staging (2021-11-03 13:07:30 -0400)
>>>>>>>
>>>>>>> are available in the Git repository at:
>>>>>>>
>>>>>>>     git://git.kraxel.org/qemu tags/egl-20211105-pull-request
>>>>>>>
>>>>>>> for you to fetch changes up to
>>>>>>> 1350ff156b25be65c599ecca9957ce6726c6d383:
>>>>>>>
>>>>>>>     ui/gtk-egl: blitting partial guest fb to the proper scanout
>>>>>>> surface
>>>>>>> (2021-11-05 12:29:44 +0100)
>>>>>>>
>>>>>>> ----------------------------------------------------------------
>>>>>>> gtk: a collection of egl fixes.
>>>>>>>
>>>>>>> ----------------------------------------------------------------
>>>>>>>
>>>>>>> Dongwon Kim (6):
>>>>>>>     virtio-gpu: splitting one extended mode guest fb into n-scanouts
>>>>>>>     ui/gtk-egl: un-tab and re-tab should destroy egl surface and
>>>>>>> context
>>>>>>>     ui/gtk-egl: make sure the right context is set as the current
>>>>>>>     ui/gtk-egl: guest fb texture needs to be regenerated when
>>>>>>>       reinitializing egl
>>>>>>>     ui/gtk: gd_draw_event returns FALSE when no cairo surface is bound
>>>>>>>     ui/gtk-egl: blitting partial guest fb to the proper scanout
>>>>>>> surface
>>>>>>>
>>>>>>>    include/hw/virtio/virtio-gpu.h        |  5 +++--
>>>>>>>    include/ui/console.h                  |  4 ++++
>>>>>>>    hw/display/virtio-gpu-udmabuf-stubs.c |  3 ++-
>>>>>>>    hw/display/virtio-gpu-udmabuf.c       | 22 ++++++++++++++--------
>>>>>>>    hw/display/virtio-gpu.c               |  4 ++--
>>>>>>>    ui/egl-helpers.c                      | 25
>>>>>>> +++++++++++++++++++++----
>>>>>>>    ui/gtk-egl.c                          | 10 ++++++++++
>>>>>>>    ui/gtk.c                              | 23 +++++++++++++++++++++++
>>>>>>>    8 files changed, 79 insertions(+), 17 deletions(-)
>>>>>>
>>>>>> Applied, thanks.
>>>>>
>>>>> Ubuntu 18.04.4 LTS:
>>>>>
>>>>> ui/gtk-egl.c:159:13: error: implicit declaration of function
>>>>> 'egl_dmabuf_release_texture' is invalid in C99
>>>>> [-Werror,-Wimplicit-function-declaration]
>>>>>              egl_dmabuf_release_texture(vc->gfx.guest_fb.dmabuf);
>>>>>              ^
>>>>> ui/gtk-egl.c:159:13: error: this function declaration is not a prototype
>>>>> [-Werror,-Wstrict-prototypes]
>>>>> 2 errors generated.
>>>>>
>>>>> https://app.travis-ci.com/gitlab/qemu-project/qemu/builds/241272737
>>>>>
>>>>
>>>> This seems to fix but I have no clue whether it is correct:
>>>>
>>>> -- >8 --
>>>> diff --git a/ui/gtk-egl.c b/ui/gtk-egl.c
>>>> index f2026e4b5c9..45cb67712df 100644
>>>> --- a/ui/gtk-egl.c
>>>> +++ b/ui/gtk-egl.c
>>>> @@ -156,8 +156,10 @@ void gd_egl_refresh(DisplayChangeListener *dcl)
>>>>               surface_gl_create_texture(vc->gfx.gls, vc->gfx.ds);
>>>>           }
>>>> +#ifdef CONFIG_GBM
>>>>           if (vc->gfx.guest_fb.dmabuf) {
>>>>               egl_dmabuf_release_texture(vc->gfx.guest_fb.dmabuf);
>>>>               gd_egl_scanout_dmabuf(dcl, vc->gfx.guest_fb.dmabuf);
>>>>           }
>>>> +#endif
>>>>       }
>>>>
>>>> ---
>>>
>>> I see the same error here trying to build QEMU git master on Debian
>>> Buster (oldstable). The fix looks reasonable to me in that it matches
>>> the CONFIG_GBM guards around other similar functions and the resulting
>>> binary appears to work, so:
>>>
>>> Reviewed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
>>
>> Thank you, I'll post a formal patch then.
>>
>> I wonder why this got merged while this configuration is covered in
>> Travis-CI. Is it that we have a too high failure rate than we don't
>> use it anymore?
> 
> I've not looked at travis in ages what matters is GitLab CI, and there

I Cc'ed you because I thought it was you who suggested to add the
Travis-CI - GitLab bridge (external job). Maybe it was Thomas.

> the problem is that we're building ubuntu 20.04 not 18.04 so didn't
> catch the older problem.
> 
> I don't know what Mark sees a problem on Debian Buster though, as we
> successfully built there AFAICT
> 
>    https://gitlab.com/qemu-project/qemu/-/jobs/1756202449

But here libgbm is installed:

Run-time dependency gbm found: YES 18.3.6

The issue is on hosts without it.