Bug 16112

Recently Qi Chen sent this patch (1):
  insane.bbclass: avoid unnecessary rerun of do_patch
and while this is a good insane check, there is more that can an should be done.


I'm creating this ticket from memory of the YP patch review discussion
so if you work on this you may have to ask for additional guidance that I
may have missed.

The basic idea is to do two different builds, with and without a 
given DISTRO_FEATURE set and run a series of tests on the resulting outputs
to ensure that  "things make sense". In the case of the ptest DISTRO_FEATURE,
recipes that do not have ptest should not rebuild.

There was some discussion of how a test such as this one for ptests is
similar to the sstate oeqa tests, which can be helpful to run in the AB. 
This likely refers to:
   meta/lib/oeqa/selftest/cases/sstatetests.py and other changes,

Qi has these tests in mind:

   1. All target DISTRO_FEATURES should not affect native recipes.
   2. Some of the target DISTRO_FEATURES should not affect cross recipes.

Richard suggests that a for simple recipes that don't depend on other recipes:
   3. if a target recipe doesn't use ptests, ideally changing
     DISTRO_FEATURES to add ptest shouldn't rebuild that recipe since 
     it should be unchanged.

 

1) all versions:
https://lore.kernel.org/openembedded-core/?q=insane.bbclass%3A%20avoid%20unnecessary%20rerun%20of%20do_patch

v3: 
https://lore.kernel.org/openembedded-core/20251218025351.427853-1-Qi.Chen@windriver.com/