Summary: | packages in package.provided are not considered for blocks | ||
---|---|---|---|
Product: | Portage Development | Reporter: | Petteri Räty (RETIRED) <betelgeuse> |
Component: | Core - Dependencies | Assignee: | Portage team <dev-portage> |
Status: | RESOLVED INVALID | ||
Severity: | normal | CC: | anakin.skyw, blubb, jakub |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Petteri Räty (RETIRED)
![]() I'd argue that this is not a bug: Most blockers are there because 2 packages install the same file in the same location. If you install a packages by hand (and put it in package.provided) you usually install it in a different location. (In reply to comment #1) > I'd argue that this is not a bug: Most blockers are there because 2 packages > install the same file in the same location. If you install a packages by hand > (and put it in package.provided) you usually install it in a different > location. So - you install the packages in other locations so that other ebuilds from portage wouldn't find them? Hmmm. I don't. you can put 'it' (e.g. being a binary or a shared lib) in other locations and it will be found nevertheless, that's what we have PATH and LDPATH for. Yeah, you can put 'it' in any location and set PATH and LDPATH accordingly and break the stuff that blocks a particular version of something or blocks even the package completely b/c it breaks when it gets installed at the same time. ;) The current behaviour is inconsistent, that's about it. Quick example? cryptsetup/cryptsetup-luks. I'm sure there are quite a few of them. i didn't say this isn't a bug - i just said your reasoning sucks :P I don't consider this a bug, if you put something in p.provided you take full responsibility for it. You can still use collision-protect to avoid file clashes. And if p.provided packages would block in-tree packages without any real reason (like using a different prefix) people will get angry. Basically portage shouldn't act like it knows things it doesn't know (in this case package identity). (In reply to comment #6) > I don't consider this a bug, [...] hugh, the chief has spoken. |