[RFC] [tests] Running the wayland tests against compositor-headless

[Eoff, Ullysses A]

Changing the tests to run headless is probably not necessary...

You can create your own custom "check" script if weston-test-env
does not suite your needs.  To use it with "make check", just set/export the
LOG_COMPILER environment variable to refer to that script before running
"make check".  "make check" just invokes your LOG_COMPILER script with the
current test as an argument and then uses your scripts exitcode to determine
pass/fail status.

See <http://www.gnu.org/software/automake/manual/html_node/Parallel-Test-Harness.html>

I use a custom python script in this manner so that I can run the tests on both
Weston drm and x11 backends per test with only one invocation of "make check".
My custom script also allows me to parse the test output/results into a standard, normalized
Xml format that is recognized by our continuous integration server.

U. Artie Eoff

> -----Original Message-----
> From: wayland-devel-bounces+ullysses.a.eoff=intel.com at lists.freedesktop.org
> [mailto:wayland-devel-
> bounces+ullysses.a.eoff=intel.com at lists.freedesktop.org] On Behalf Of Sam
> Spilsbury
> Sent: Monday, April 22, 2013 11:50 AM
> To: wayland
> Subject: [RFC] [tests] Running the wayland tests against compositor-headless
> 
> Hi,
> 
> I'm not sure if this is a work-item that's already in-progress by
> somebody. I noticed that weston has a "headless" compositor backend
> and a "no-op" renderer, which effectively do not depend on having
> OpenGL or *a* windowing system available at the time that weston is
> run.
> 
> It would be really great if we could run the tests against this
> backend, because then it means that they can be run on foreign
> machines like continuous integration servers with ease.
> 
> I've tried quickly changing the backend over in weston-test-env -
> currently most of the tests are failing because client->pointer is not
> initialized. Upon initializing that, most of them pass except those
> that depend on client->keyboard . I suspect that this will involve
> having to create a wl_keyboard with a set of dummy xkb data for
> layouts, though I'm not sure how to do that yet.
> 
> My question to the list is - does it make sense to submit a patch to
> change the tests over to use the headless backend? Or is the
> preference to keep them using the wayland or x11 backend for now.
> 
> Best,
> 
> Sam.
> 
> --
> Sam Spilsbury
> _______________________________________________
> wayland-devel mailing list
> wayland-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/wayland-devel