patchelf: fix golang support #57656
patchelf: fix golang support #57656
Conversation
Ideally, we would have the fix released in an official tarball and we would pick it up in nixpkgs but since there was no new release of the tool in the last three years, and the issue that asks to release a new version is open for at least two years now, it's probably better to not wait for this to happen and cherry-pick the fix.
|
What specifically does this fix in regards to Go? |
|
@infinisil The upstream bug is: NixOS/patchelf#66 |
|
I see. If we're going to cause a mass rebuild by updating patchelf, I think it's be a better idea to update it to the master version instead (since there's no new release). Can you open a PR for that instead? @edolstra Alternatively, could you release a new patchelf version? |
|
@infinisil I was trying to stay low profile with the change. :) I can probably propose something for master but I don't know if I need to do any extensive testing for the change when doing it. It's 3 years of development in a core component. I would not be surprised if things break on the way. |
|
nixpkgs is most likely the biggest user of patchelf, so if there's problems with patchelf, the sooner we use the latest version, the sooner we can fix those problems, if there are any. On the other hand, we probably wouldn't break anything with this simple patch, so we might as well just merge this.. I'd really like to have a newer version of patchelf though. Am conflicted. |
|
After talking about it on IRC, @edolstra agreed to make a new patchelf release :) Can you make a PR for that update? |
|
@infinisil I will update with a link to PR that bumps the version. Thanks for checking with upstream. |
|
That being said, I don't think we should necessarily switch to latest for release branches, that may make use of a fix for golang. So if you think a backport would be acceptable in release branches, please tell me so, I will then propose separate PRs. |
|
@infinisil here is the PR to bump patchelf version: #58518 Again, if you think it's acceptable to go with a backport of a particular patch for release branches to avoid unexpected breakages due to three years of development, please tell me. |
Ideally, we would have the fix released in an official tarball and we
would pick it up in nixpkgs but since there was no new release of the
tool in the last three years, and the issue that asks to release a new
version is open for at least two years now, it's probably better to not
wait for this to happen and cherry-pick the fix.