I have last rited PyCXX and PySVN, with the following rationale: # Michał Górny <mgorny@gentoo.org> (2025-03-26) # PyCXX and PySVN have no Gentoo maintainer since 2019, and they have # been effectively abandoned years earlier. PyCXX no longer installs # correctly, and the hacks to workaround upstream bugs are piling up. # PySVN ebuild has faked PEP517 support. The packages don't have # working tests nor Python 3.13 support. # Removal on 2025-04-25. Bug #751541. ulm has subsequently reverted the p.mask after adding another hack to workaround one of the listed issues. However, none of the remaining problems were addressed: 1. The packages still have no maintainer. 2. PySVN still doesn't do PEP517 correctly. 3. All affected packages don't support Python 3.13. On top of that, the PyCXX issue was hacked around locally without even bothering to report the issue upstream. All this considered, I find the revert unacceptable.
Log of #gentoo-python from 2025-03-26. Times are in CET (UTC+0100). <+ulm> mgorny: do you expect a quick resolution of https://discuss.python.org/t/installing-headers-from-wheel-should-the-directory-be-normalized-or-not/85990 ? [08:52] <+ulm> if not, I'd suggest adjusting the install path in the pycxx ebuild (maybe just doing a manual move in src_install?) [08:53] <@mgorny> suggest all you want, that package is maintainer-needed and i'm not touching it [08:57] <+ulm> not helpful :( <@mgorny> it also still doesn't support 3.13, and i doubt it ever will [08:58] <@mgorny> it is maintainer-needed since 2019 [08:59] <+ulm> dev-vcs/svneverever needs it as reverse dependency, so last-riting isn't an option <+ulm> and AFAICS it runs perfectly well <@mgorny> ...as the linked bug report proves [09:00] <+ulm> well, the setuptools change broke it <@mgorny> the package relied on undocumented, nonstandard behavior that only happened to work because of how things were accidentally implemented [09:01] <+ulm> then fix it? <@mgorny> and because we already hacked other bugs in the ebuild <@mgorny> ulm: no. i won't waste my time on some random dead NIH package just because someone decided to use it for their tool [09:02] <+ulm> again, not helpful :( [09:03] <@mgorny> oh great, pysvn doesn't even use PEP517 build, it just shoved a DISTUTILS_USE_PEP517 on top to workaround the check [09:05] <+ionen> :/ [09:06] <+ulm> the problem is that these things have become so complicated that it's practically impossible for a python non-expert to maintain them [09:07] <+ulm> it start with names like "pep517" that are totally undescriptive <+ulm> *starts <+ionen> well, the python guide exists for a reason [09:08] <+ulm> that doesn't contradict my statement [09:09] <+ulm> arguably, that one would need a guide is an indicator for its complicatedness <@mgorny> because python ecosystem wasn't ever designed for packages like pycxx [09:10]
So I see a breakage in dev-python/pysvn, spend some time investigating the reason, and report in (existing) bug 950381. Then I ask in #gentoo-python and don't get any helpful answer, only a "that package is maintainer-needed and i'm not touching it" from mgorny. From that I conclude that I can commit an ad-hoc fix (especially since the situation seems not all that clear, see the upstream discussion linked in bug 950381), and I spend further time. Only to discover before pushing that the packages have been masked. Thank you very much, Sir. Also, the package mask had bug 751541 as removal bug, which seems unrelated to any of the packages listed in it.
I don't have a problem with an ad-hoc fix. What I have a problem with, you unilaterally deciding to revert last rites without addressing any of the remaining problems with the packages. Which means that they remain first-grade last rite candidates, except they're not masked now, because masking them is likely to provoke a revert war with an entitled developer who believes that it's fine to keep semi-broken maintainer-needed packages in the repository because he wants to use them, but doesn't want to take responsibility for them. And yes, I have accidentally typed the wrong bug number. However, the mask was reverted so swiftly, I didn't even manage to correct that.
(In reply to Michał Górny from comment #3) > ... entitled developer ... I am out of here, because I am unwilling to continue the discussion on this level.
