Python people would eventually like to be able to use python-2.6 or something like that in their USE_EXPAND names. PMS currently states [A-Za-z0-9+_@-]. Current situation seems to be that Portage acceps them but repoman doesn't: [ebuild R ] dev-java/ant-core-1.8.0-r2 USE="-doc -foo.bar% -source" 0 kB IUSE.invalid 1 dev-java/ant-core/ant-core-1.8.0-r2.ebuild: foo.bar
(In reply to comment #0) > Current situation seems to be that Portage acceps them but repoman doesn't: It think repoman is just complaining that there's no description in metadata.xml. Anyway, an EAPI bump for that isn't a bad idea.
(In reply to comment #1) > (In reply to comment #0) > > Current situation seems to be that Portage acceps them but repoman doesn't: > > It think repoman is just complaining that there's no description in > metadata.xml. Anyway, an EAPI bump for that isn't a bad idea. I'd be a bit curious how folk plan on dealing w/ parsing metadata.xml (which is eapi unversioned), while varying the EAPI rules for USE parsing of local entries w/in there...
(In reply to comment #2) > > I'd be a bit curious how folk plan on dealing w/ parsing metadata.xml (which is > eapi unversioned), while varying the EAPI rules for USE parsing of local > entries w/in there... > Parsing metadata.xml strictly even now is a bad option as they DTD changes every once in a while.
(In reply to comment #3) > Parsing metadata.xml strictly even now is a bad option as they DTD changes > every once in a while. Someone might want to point that out since we moved our use local descriptions into metadata.xml ;) Regardless, dtd changes, sure, but for any PM that does validation of the USE flags that come out of the spec, my point still stands.
(In reply to comment #4) > (In reply to comment #3) > > Parsing metadata.xml strictly even now is a bad option as they DTD changes > > every once in a while. > > Someone might want to point that out since we moved our use local descriptions > into metadata.xml ;) > Yeah we could think about coming up with a solution to the metadata.xml problem as a dependency to this one.
(In reply to comment #5) > Yeah we could think about coming up with a solution to the metadata.xml problem > as a dependency to this one. Just noting it on the ticket from the irc discussion, profiles/use.desc has the exact same issue... Might be time to revisit adding a repository version format file (would also be a way to address portage enabling package.mask as a directory for profiles cleanly).
No progress. Closing.