Re: [docs] [PATCH] ref-manual: document PTEST_RUNNER_TIMEOUT variable

["Antonin Godard" <antonin.godard@bootlin.com>]

Hi,

On Tue Mar 3, 2026 at 9:17 PM CET, Tim Orling wrote:
> On Tue, Mar 3, 2026 at 2:59 AM Antonin Godard <antonin.godard@bootlin.com>
> wrote:
>
>> Hi,
>>
>> On Fri Feb 27, 2026 at 9:02 PM CET, Tim Orling via lists.yoctoproject.org
>> wrote:
>> [...]
>> > diff --git a/documentation/ref-manual/variables.rst
>> b/documentation/ref-manual/variables.rst
>> > index 208b5e43d..587f67cc6 100644
>> > --- a/documentation/ref-manual/variables.rst
>> > +++ b/documentation/ref-manual/variables.rst
>> > @@ -7793,6 +7793,20 @@ system and gives an overview of their function
>> and contents.
>> >        For more information, see
>> >        the :ref:`ref-classes-ptest-python-pytest` class documentation.
>> >
>> > +   :term:`PTEST_RUNNER_TIMEOUT`
>> > +      Specifies the timeout in seconds for each ptest suite executed by
>> > +      ``ptest-runner`` during :ref:`ref-classes-testimage` runtime
>> testing.
>> > +      The default value is ``"450"`` (7.5 minutes).
>> > +
>> > +      This variable is set in the :ref:`ref-classes-testimage` class
>> and can
>> > +      be overridden in image recipes to adjust the timeout for specific
>> > +      images or multiconfig variants. For example, in
>> ``core-image-ptest``::
>> > +
>> > +         PTEST_RUNNER_TIMEOUT:virtclass-mcextend-python3-cffi = "600"
>>
>> Could you explain why you're using the :virtclass-mcextend-python3-cffi
>> override
>> here? Seems like a complicated example for this variable's description,
>> but if
>> it is needed, we might want to explain it?
>>
>
> The reason is simple, all the other variables in the "core-image-ptest.bb"
> recipe were set with  "virtclass-mcextend-". [1] A prior attempt at setting
> with "pn-python3-cffi" in "ptest-packelists.inc" had not worked, although
> that was before I caught the need for the "testimage.bbclass" change.
>
> After further experimentation:
>
> 1. PTEST_RUNNER_TIMEOUT:pn-python3-cffi = "600" does NOT work [2][3]
> 2. PTEST_RUNNER_TIMEOUT:pn-core-image-ptest-python3-cffi = "600" DOES work
> [4][5]
>
> We also know from Trevor Gamblin experiments that
> PTEST_RUNNER_TIMEOUT:riscv-64 = "600" in local.conf works (and similar
> experiments with "conf" changes in "yocto-autobuilder-helper" worked). The
> goal of the "core-image-ptest.bb" changes were intended to avoid REQUIRING
> a change in "conf" only, and instead enable the change to happen in recipe
> scope.
>
> So the conclusion appears to be that "image recipe" and "conf" scope work,
> but an individual ptest enabled recipe doesnot work. The "image" scope
> probably makes sense since PTEST_RUNNER_TIMEOUT is actually enabled by the
> "testimage.bbclass"?
>
> Would it make more sense to use the "more obvious" usage in #2 above (even
> though it differs from other variable overrides in "core-image-ptest.bb"
> recipe)?

I think we need to write this variable's documentation as a point of view of
someone using the testimage class, not the core-image-ptest recipe. The
core-image-ptest recipe appears to use some niche use-cases of overrides with
mcextend (which also isn't documented).

So, if I were a user writing an image that inherits the testimage class, and my
image recipe does not contain any mcextend shenanigans, would I really be able
to use syntax #1 or #2?

I can obviously override the value of PTEST_RUNNER_TIMEOUT for all my underlying
ptests, with something like:

my-image.bb:
  IMAGE_CLASSES += "testimage"
  PTEST_RUNNER_TIMEOUT = "1000"

But that's it, I guess?

Antonin