[10:17:17] <few_> zmedico: could it be that you broke p.mask/p.unmask handling? [10:18:05] <zmedico> few_: haven't touched that lately afaik [10:18:32] <few_> the mysql overlay has a profile/p.unmask and it seems to always override /etc/portage/p.mask [10:18:44] <few_> I'd say that's wrong [10:19:48] <zmedico> well, package.unmask sucks :) [10:20:03] <zmedico> I tend to use negative atoms in package.mask [10:20:12] <zmedico> it's a lot safter [10:20:18] <zmedico> *safer [10:20:52] <few_> then you should not support p.unmask in profile/, no? [10:21:39] <zmedico> I can't think of a reason for it to be used [10:21:55] <zmedico> they can just use negative package.mask atoms [10:22:44] <few_> jmbsvicetto: please migrate the profile/p.unmask to a p.mask with negative atoms in the mysql overlay. see above. [10:22:54] <few_> zmedico: then please remove the support [10:23:33] <zmedico> I guess we should also trigger a warning [10:23:40] <few_> yre you doing that right now or should I file a bug? [10:23:44] <few_> fine with me [10:23:55] <zmedico> file a bug please [10:23:58] <few_> k
To make clear what the actual problem is: If an overlay has profiles/package.unmask, which unmasks cat/pkg from this overlay, then it is not possible to mask cat/pkg from this overlay with /etc/portage/package.mask. This means that the overlays settings would take precedence over user settings, which is wrong. (There is a hack to get around this but the average user shouldn't have to know this.)
Any updates on this?
I see that the PMS 5.2.8 package.mask section documents negative package.mask atoms: https://dev.gentoo.org/~ulm/pms/head/pms.html#x1-510005.2.8 > Note that the -spec syntax can be used to remove a mask in > a parent profile, but not necessarily a global mask (from > profiles/package.mask, section 4.4). Maybe we should add a boolean flag to enable/disable package.unmask via repos.conf. I don't see any mention of package.unmask in PMS, so it seems like we could disable it by default.