[PATCH weston 0/6] Make the headless backend useful again

[Pekka Paalanen]

On Thu,  6 Feb 2014 12:30:30 +0100
Emilio Pozuelo Monfort <pochu27 at gmail.com> wrote:

> From: Emilio Pozuelo Monfort <emilio.pozuelo at collabora.co.uk>
> 
> This fixes a few bugs in the headless backend, to the point where
> it can run the test suite, except for two issues:
> 
> - the buffer-count is skipped because mesa initialization fails,
>   more details in the patch commit message
> - I haven't been able to properly test the xwayland test as it
>   hangs here (with both headless and x11 backends), probably
>   because my xserver doesn't have xwayland support compiled in.
> 
> Finally the headless backend is used by default, with an easy way
> to use other backends if so desired. This will make it easier to
> have `make check' run in a CI environment, or for distro package
> builds.
> 
> Emilio Pozuelo Monfort (6):
>   compositor-headless: create input devices
>   noop-renderer: Set the buffer size on attach requests
>   noop-renderer: Read the shm buffer contents on attach
>   tests: Skip buffer-count if EGL initialization fails
>   tests: use the headless backend to run the test suite
>   tests: Properly report skipped tests
> 
>  src/compositor-headless.c         | 30 ++++++++++++++++++++++++----
>  src/noop-renderer.c               | 27 ++++++++++++++++++++++++++
>  tests/buffer-count-test.c         | 29 +++++++++++++++++----------
>  tests/weston-test-client-helper.c |  8 +++++---
>  tests/weston-test-runner.c        | 41 ++++++++++++++++++++++++++++-----------
>  tests/weston-test.c               |  6 ++++++
>  tests/weston-tests-env            | 10 ++++------
>  7 files changed, 117 insertions(+), 34 deletions(-)
> 

Hi,

I commented on a couple minor things, but otherwise this series looks
good. Didn't test yet, though.

Hmm, the buffer-count test... I think the EGL init in the test client
would not fail, if you built Mesa with --enable-gallium-egl and swrast
gallium driver (softpipe/llvmpipe). What would the test do then is a
good question, since the client would be using egl_gallium, and the
server just wl_shm.


Thanks,
pq