Summary: | Ignore collisions if target symlink owned by same package | ||
---|---|---|---|
Product: | Portage Development | Reporter: | Sam James <sam> |
Component: | Core | Assignee: | Portage team <dev-portage> |
Status: | CONFIRMED --- | ||
Severity: | normal | CC: | flow, gentoo, hlein, hydrapolic, pacho, richard, zx2c4 |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: |
https://bugs.gentoo.org/show_bug.cgi?id=664940 https://bugs.gentoo.org/show_bug.cgi?id=861785 https://bugs.gentoo.org/show_bug.cgi?id=916036 |
||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 564428, 815118, 834503, 957712 |
Description
Sam James
![]() ![]() ![]() ![]() To be clear: Portage currently will reject replacing a symlink to a directory with a directory and vice-versa, but it looks like it'd be valid for it to introspect whether the package owns both and do the replacement itself if so. I also have hit this some times, I see also some ebuilds in the tree workarounding the problem by removing the directory at pkg_preinst phase :/ This also affects coreutils, which is probably more of a worry for everyone. (In reply to Richard Gray from comment #3) > This also affects coreutils, which is probably more of a worry for everyone. I'm not sure how it does. We're not doing any replacements between symlinks and directories there. I'm rather confused myself. I'll post more detail when I'm (more-or-less) up and running. I'm in the middle of a rebuild right now. Since everything seems to be converging on /usr/bin/ these days, I'll probably try removing the symlinks temporarily just to appease portage and have in intact installation tree. I suggest filing a bug (or perhaps better, going on IRC to #gentoo) and showing full output. Sounds like possibly botched merge-usr conversion. |