| Summary: | [Tracker] KEYWORDS must not be set in eclass | ||
|---|---|---|---|
| Product: | Quality Assurance | Reporter: | Ralph Sennhauser (RETIRED) <sera> |
| Component: | Trackers | Assignee: | Gentoo Quality Assurance Team <qa> |
| Status: | CONFIRMED --- | ||
| Severity: | QA | CC: | dev-portage, enlightenment+disabled, pacho, patrick, sam |
| Priority: | High | Keywords: | Tracker |
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| See Also: | https://bugs.gentoo.org/show_bug.cgi?id=170165 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Bug Depends on: | 562464 | ||
| Bug Blocks: | |||
|
Description
Ralph Sennhauser (RETIRED)
2010-10-22 13:15:22 UTC
games-q3mod.eclass lacks a @MAINTAINER - CC'ing games@ tetex.eclass lacks a @MAINTAINER: - CC'ing tex@ vdr-plugin.eclass lacks a @MAINTAINER: - CC'ing vdr@ From PMS, section 8.2: KEYWORDS: A Whitespace separated list of keywords for the ebuild. Each token must be a valid keyword name, as per section 3.1.6. May include -*, which indicates that the package will only work on explicitly listed archs. May include -arch, which indicates that the package will not work on the specified arch. May be empty, which indicates uncertain functionality on any architecture. >>>May be defined in an eclass.<<< So it is valid to set KEYWORDS in the eclass and the devmanual may need to be updated. Where does vdr-plugin.eclass sets KEYWORDS? I only see the commented KEYWORDS="..." as part of an example ebuild. whatever the decision is, feel free to touch tetex.eclass: it's deprecated, hasn't been updated for years, I don't think anyone care about it and it'll be useless when ptex will be removed (ptex has been merged with texlive for the 2010 release) Most of those are false positives (KEYWORDS is part of comment - example ebuild) - please fix reporting tool. Removing myself and autotools-utils from bug. (In reply to comment #5) > Most of those are false positives (KEYWORDS is part of comment - example > ebuild) - please fix reporting tool. > Removing myself and autotools-utils from bug. > I'm sorry for the overhead I created, only games-q3mod, enlightenment and tetex are effectively affected. Please someone remove @vdr as well. (In reply to comment #2) > > So it is valid to set KEYWORDS in the eclass and the devmanual may need to be > updated. > This is certainly less work as it wont break anything. However Gentoo may have stricter QA rules than the PM requires if it makes sense. Removing vdr as we don't set KEYWORDS in the vdr-plugin eclass. Update: games-q3mod and tetex are no longer in gentoo-x86, so only enlightenment.eclass is left doing this now. (would likely be a good idea to remove games from CC, but I will not do this myself) Indeed enlightenment.eclass is still doing it and we have a new one in this list, namely intel-sdp.eclass Adding maintainers to CC. + 14 Feb 2013; Justin Lecher <jlec@gentoo.org> intel-sdp.eclass: + Drop KEYWORDS from eclass, #342185 + Interestingly, the eclass was reviewed exactly with this on g-dev ml. This shows the quality of such an review. (In reply to comment #9) yes, 9 years and counting of enlightenment.eclass doing this and causing horrible QA breakage everywhere. no, wait, that has yet to happen. i see no reason to fix something isn't broken. *** Bug 486392 has been marked as a duplicate of this bug. *** Well, I noticed this bug report some minutes ago due: http://www.gossamer-threads.com/lists/gentoo/dev/289697 that explains we setting keywords outside ebuilds is a problem at stabilization time (In reply to Thomas Sachau from comment #2) > From PMS, section 8.2: > > KEYWORDS: > > A Whitespace separated list of keywords for the ebuild. Each token must be a > valid > keyword name, as per section 3.1.6. May include -*, which indicates that the > package will > only work on explicitly listed archs. May include -arch, which indicates > that the package > will not work on the specified arch. May be empty, which indicates uncertain > functionality on > any architecture. >>>May be defined in an eclass.<<< > > So it is valid to set KEYWORDS in the eclass and the devmanual may need to > be updated. In current PMS I see instead KEYWORDS listed under "Chapter 7", that is about variables defined by ebuilds :/ (maybe it was changed at some time) (In reply to Pacho Ramos from comment #14) > In current PMS I see instead KEYWORDS listed under "Chapter 7", that is > about variables defined by ebuilds :/ (maybe it was changed at some time) We have moved a couple of variables from mandatory to optional at some point: http://git.overlays.gentoo.org/gitweb/?p=proj/pms.git;a=commit;h=d99d0fe7c4fef5b541604020642255de291e4dbd Also, PMS primarily defines package manager behaviour. It is not a problem for the PM if KEYWORDS (or most other variables) are defined in an eclass. That doesn't prevent us from having a policy that is stricter than PMS. See bug 170165. Well, the original issue was that KEYWORDS isn't a "merged" variable like IUSE is, so setting it in eclasses doesn't do what you'd expect. So far, everyone has ignored this QA issue. Can we get a QA vote on whether this will be enforced or not and if there will be exceptions? Maybe you should suggest it for the Council agenda :-D. (In reply to Michał Górny from comment #18) > Maybe you should suggest it for the Council agenda :-D. I don't see how this issue needs the council to step in. It's clearly a QA issue. if QA agrees, we can make this a fatal repoman warning no? Fixed keywording for enlightenment.eclass. It's been more than two years and paludis is still complaining. The only remaining eclass failing is ros-catkin, but bug 562464 wasn't updated since late 2015. Is there anything I can help with to help fixing these? |
