Summary: | [Future EAPI] pkg_failed phase (formalization of death hooks) | ||
---|---|---|---|
Product: | Gentoo Hosted Projects | Reporter: | Andreas K. Hüttel <dilfridge> |
Component: | PMS/EAPI | Assignee: | PMS/EAPI <pms> |
Status: | CONFIRMED --- | ||
Severity: | normal | CC: | esigra, mgorny, sam |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: | https://bugs.gentoo.org/show_bug.cgi?id=465192 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 174380 |
Description
Andreas K. Hüttel
2022-07-23 21:46:06 UTC
At a first glance, this seems to make sense. I mean, certainly more sense than the API provided for death hooks right now. Agreed. Plus we already have e.g. pkg_info which feels similar in spirit. My only hesitation here is: ... surely this has been proposed before? If so, why didn't we do it then? Precedent seems to suggest pkg_die as the name. There are chromium_pkg_die and java-pkg_die in eclasses, and die_hooks in Portage. (In reply to Sam James from comment #2) > Agreed. Plus we already have e.g. pkg_info which feels similar in spirit. pkg_info is used by exactly one package in the tree (app-emulation/qemu). My main worry is that the new phase function would also have little use. > My only hesitation here is: ... surely this has been proposed before? If so, > why didn't we do it then? I don't remember. Maybe there's little demand for something like this? I'm not sure whether a good proposal was ever made. It might just have skipped between "death hooks are too ugly to put in PMS" and "death hooks just work, and this is a corner case anyway". |