Simon Glass <s...@chromium.org> schrieb am Sa., 3. Mai 2025, 23:27:
> Hi Heinrich,
>
> On Sat, 3 May 2025 at 07:32, Heinrich Schuchardt
> <heinrich.schucha...@canonical.com> wrote:
> >
> > In some cases we have alternative configuration options that supply the
> > same functionality, e.g CONFIG_NET and CONFIG_NET_LWIP.
> >
> > Allow to specify all of them as arguments for buildconfigspec() and
> execute
> > the text if any of these is fulfilled, e.g.
> >
> > @pytest.mark.buildconfigspec('net', 'net_lwip')
> >
> > Update the documentation.
> >
> > Signed-off-by: Heinrich Schuchardt <heinrich.schucha...@canonical.com>
> > ---
> > doc/develop/py_testing.rst | 24 ++++++++++++++++++++++++
> > test/py/conftest.py | 10 +++++++---
> > 2 files changed, 31 insertions(+), 3 deletions(-)
> >
> > diff --git a/doc/develop/py_testing.rst b/doc/develop/py_testing.rst
> > index 217ae447035..7dbed8833bf 100644
> > --- a/doc/develop/py_testing.rst
> > +++ b/doc/develop/py_testing.rst
> > @@ -514,3 +514,27 @@ of the `ubman.config` object, for example
> > Build configuration values (from `.config`) may be accessed via the
> dictionary
> > `ubman.config.buildconfig`, with keys equal to the Kconfig variable
> > names.
> > +
> > +A required configuration setting can be defined via a buildconfigspec()
> > +annotation. The name of the configuration option is specified in lower
> case. The
> > +following annotation for a test requires CONFIG_EFI_LOADER=y:
> > +
> > +.. code-block:: python
> > +
> > + @pytest.mark.buildconfigspec('efi_loader')
> > +
> > +Sometimes multiple configuration option supply the same functionality.
> If
> > +multiple arguments are passed to buildconfigspec(), only one of the
> > +configuration options needs to be set. The following annotation
> requires that
> > +either of CONFIG_NET or CONFIG_NET_LWIP is set:
> > +
> > +.. code-block:: python
> > +
> > + @pytest.mark.buildconfigspec('net', 'net lwip')
> > +
> > +The notbuildconfigspec() annotation can be used to require a
> configuration
> > +option not to be set. The following annotation requires CONFIG_RISCV=n:
> > +
> > +.. code-block:: python
> > +
> > + @pytest.mark.notbuildconfigspec('riscv')
> > diff --git a/test/py/conftest.py b/test/py/conftest.py
> > index 5aea85647af..6c3ac67979a 100644
> > --- a/test/py/conftest.py
> > +++ b/test/py/conftest.py
> > @@ -711,9 +711,13 @@ def setup_buildconfigspec(item):
> > """
> >
> > for options in item.iter_markers('buildconfigspec'):
> > - option = options.args[0]
> > - if not ubconfig.buildconfig.get('config_' + option.lower(),
> None):
> > - pytest.skip('.config feature "%s" not enabled' %
> option.lower())
> > + nomatch = True
> > + for arg in options.args:
> > + if ubconfig.buildconfig.get('config_' + arg.lower(), None):
> > + nomatch = False
> > + if nomatch:
> > + argsString = ', '.join(options.args)
> > + pytest.skip(f'.config features "{argsString}" not enabled')
> > for options in item.iter_markers('notbuildconfigspec'):
> > option = options.args[0]
> > if ubconfig.buildconfig.get('config_' + option.lower(), None):
> > --
> > 2.48.1
> >
>
> I ran into this problem too. Would it be possible to have CONFIG_NET
> mean that we have networking (either option) and CONFIG_LWIP mean that
> it is lwip? The current Kconfig seems a bit broken to me....
>
> Regards,
> Simon
>
Hello Simon,
Yes, we can think about reworking the network configuration in a future
series.
E.g. have as choice
* CONFIG_NET_LEGACY
* CONFIG_NET_LWIP
* CONFIG_NET_NONE
and CONFIG_NET := !CONFIG_NONE
But I think it is still valuable in its own right to be able to express a
logical OR when defining test dependencies.
Best regards
Heinrich
