Looks like all packages using new python*-r1 eclasses need to have PYTHON_REQUIRED_USE set, I think would be interesting to get warnings when ebuild is completely missing a PYTHON_REQUIRED_USE reference Thanks Reproducible: Always
I'm not quite sure what that check would look like. Also, it is possible for the ebuild developer to come up with his own REQUIRED_USE constraint which would serve the same purpose. Does repoman have access to the resulting value in REQUIRED_USE? It may be better to check for python_targets_* than PYTHON_REQUIRED_USE specifically.
(In reply to Mike Gilbert from comment #1) > I'm not quite sure what that check would look like. Me neither. Patches welcome. :) > Also, it is possible for the ebuild developer to come up with his own > REQUIRED_USE constraint which would serve the same purpose. > > Does repoman have access to the resulting value in REQUIRED_USE? Yes, REQUIRED_USE is part of one of the cached metadata variables (like IUSE, *DEPEND, and stuff).
And, what about checking for that variable in eclass? That way, if PYTHON_REQUIRED_USE is not being exported, eclass could show a warning :/
You can't reference REQUIRED_USE in eclass safely. However, we could do a simple repoman check that would scan REQUIRED_USE for anything python_targets*. That should be good enough IMO.
Moving to the new bug where more discussion happened. *** This bug has been marked as a duplicate of bug 530092 ***