go, buildGoModule, buildGoPackage: default to 1.21#254878
go, buildGoModule, buildGoPackage: default to 1.21#254878zowoq merged 1 commit intoNixOS:stagingfrom qowoz:go120121
Conversation
yu-re-ka
left a comment
There was a problem hiding this comment.
just tested the gccgo based bootstrap in pkgsMusl still works
|
added 'closes #255907' to the description |
|
I started building all affected packages on my Darwin M2. It's been running all day, I'll post the result once it's done. So far 34 have failed. |
|
@zowoq do you mind if I push force your branch to match my base branch? It's been building for over 24h right now and I'd like to avoid having to rebuild for the list of broken ones, but I have started testing with my own commit not noticing your PR. |
I'd need to invite you to the org that owns the fork and give you permissions on the repo. If you give me a commit to rebase onto I'll do it for you. |
The base commit is 6273c3e which I pulled yesterday (master branch) right before I started a rebuild. |
|
Thank you for rebasing! In addition to the in-progress nix-review on my M2, I started one on my x86_64 NixOS machine. They should both post results here when they're done. |
|
Result of 314 packages marked as broken and skipped:
220 packages failed to build:
2154 packages built:
|
|
Result of 57 packages marked as broken and skipped:
136 packages failed to build:
2675 packages built:
|
I haven't done any pinning yet as I've been a bit sick but seems like a lot of the failed builds were spurious as they are now succeeding on staging-next, I might just wait and pin the stuff that doesn't build on hydra. |
|
What else has to happen to make the "go" package available in the "unstable" channel use this new version? |
|
This PR reaching the channel, I believe. |
|
I've never seen that PR Tracker before. Thank you for sharing the reference. What governs that progress? Do humans review and decide when to move a given patch forward, or is it something like positive results of tests that allow it to graduate automatically? |
|
Machines performing rebuilds, humans fixing regressions, etc. You can have a look at PR #257792 (well, anyone is welcome to help). |
|
This broke |
|
I think several others as well. At a quick glance I see traefik, cloudflared, erigon and maybe most of other go-related regressions. It will get more visibility now that it's newly in nixpkgs master. There's an interesting question about how exactly we deal with such regressions, in particular who is supposed to discover and resolve them – and at which stage. (Fortunately, trivial pinning to 1.20 should work basically always, I think.) |
|
Thank you for the change and pushing go within nixpkgs forward! As a heads up, this also broke k3s_1_26 and k3s_1_27 (fix PR #263780). It still compiles, so the build tests above didn't catch it, but the NixOS tests fail since the binary panics on startup when built with go1.21. |
|
Fix for the cloudflared: #263841 |
Description of changes
Things done
sandbox = trueset innix.conf? (See Nix manual)nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage./result/bin/)https://go.dev/doc/go1.21
On master for testing, will go to staging when ready.
closes #255907