glibc: 2.33-108 -> 2.34-115

[Ma27]

Motivation for this change

Announcement: https://sourceware.org/pipermail/libc-alpha/2021-August/129718.html

Hydra jobset: https://hydra.nixos.org/jobset/nixpkgs/glibc-2.34

Note for reviewers: this update is slightly more nasty than the previous ones, especially the following commits should be reviewed closely:

• 97acaf6 (glibc: 2.33-108 -> 2.34-115)
• da905d4 (nixos/stage-1: fix modprobe in initial ramdisk on systems w/glibc-2.34)
• 748faeb (glibc: don't hide symbol __nss_files_fopen

Todos:

• Await current jobset (https://hydra.nixos.org/eval/1741006)
• @Ma27 will try to rebuild their workstation with this PR to check if anything notable broke
• Get reviews :)

Things done

• Tested using sandboxing (nix.useSandbox on NixOS, or option sandbox in nix.conf on non-NixOS linux)
• Built on platform(s)
  • NixOS
  • macOS
  • other Linux distributions
• Tested via one or more NixOS test(s) if existing and applicable for the change (look inside nixos/tests)
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• 21.11 Release Notes (or backporting 21.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

[Ma27]

@vcunat @lovesegfault I can build on this branch pkgs.hello now, but some of the changes are slightly messy. So before I continue here, I'd have two requests:

• It would be nice to get some feedback to the current changes I made (I tried to document the reasons in the corresponding commit messages)
• It would help if https://hydra.nixos.org/jobset/nixpkgs/glibc-2.33 could be reconfigured to be re-used for glibc-2.34.

Thanks! :)

[SuperSandro2000]

Does this require the binutils update?

[Ma27]

Doesn't seem as if that's the case currently :)

[vcunat]

I'm not sure if it's good to test them separately and then merge without testing those two updates together.

[vcunat]

Quick look: the bootstrapping change is OK-ish, I think. The rest seems good.

I'm not sure why diffutils update was included. Nit: 2.33-master.patch.gz should be deleted.

[Ma27]

I'm not sure if it's good to test them separately and then merge without testing those two updates together.

I haven't taken a close look at bintuils, but considering that I even had to do some workarounds to get gcc building (yay, compatibility), I guess this will take a while (no idea if it'll land in 21.11), so it's rather likely that this will be here for a while after a git pull upstream staging --rebase for a while :)

Quick look: the bootstrapping change is OK-ish, I think. The rest seems good.

OK, great! :)

I'm not sure why diffutils update was included

With this change, the package compiled again (but good point, should perhaps add a note to the commit message).

Nit: 2.33-master.patch.gz should be deleted

Oh, good point :)

[vcunat]

I cancelled these builds for now, as too much work is piling up. (also PR #135611)

[Ma27]

@vcunat Hydra's queue status currently says Queue (117/125309). I have to admit that I don't really know the average and I didn't follow the current state of what's going on (with openssl I guess?) very closely (exam phase ^^), but I assume I should wait a few days before starting a new eval? :)

[vcunat]

Yes, most of 21.05 jobs still don't have binaries with the new openssl. And then there's lots of stuff that was cancelled and will also want to continue after this settles, e.g. the current staging-next.

[r-burns]

fetchpatchable patch for catch 1.x: https://src.fedoraproject.org/rpms/catch1/blob/059104ba87494c0b5ebe16844ec190f253e51cac/f/catch1-sigstksz.patch

[jonringer]

@jonringer would you mind restarting hte aborted jobs in https://hydra.jonringer.us/eval/551 ? :)

Sorry I just say this now. Restarted

[Ma27]

TIL to build my entire workstation's NixOS from source, I need a c6a.32xlarge from AWS and ~90GB of free disk space. In other words, I'm currently writing this comment from a NixOS using glibc 2.34.

Since that's also working perfectly fine, I'd say that this PR has reached a reasonably stable state. @jonringer I guess it'd make sense to cancel the jobset again and start a new eval with the current state.

As soon as everything else is sorted out, I'd be fine with merging :)

[jonringer]

I think we can get the remaining regressions in staging-next. The next round should probably be dedicated to glibc, then the round after for systemd 250.

[jonringer]

cc @NixOS/nixos-release-managers

[vcunat]

So the recent hydra.nixos.org evaluation of this should be cancelled? I can't see why it's building on both farms at once.

[Ma27]

I think that was on oversight of mine, cancelled on hydra.nixos.org, sorry!

[jonringer]

LGTM

Result of nixpkgs-review pr 133431 run on x86_64-linux 1

2 packages built:

• python3Packages.numpy
• ripgrep

[trofi]

I think stdenv changes were not completely rolled back. Noticed as a bootstrap failure on glibc-2.35. Proposed the fix as #162197

[Ma27]

fyi the stdenv is still building fine for me, ut I guess that there's another GCC to be built, so this should be merged, but it's not a blocker.

[Ma27]

also, does that mean you're working on glibc-2.35? :)

(in that case the next mass-rebuild project for me would be a GCC major bump %))

[trofi]

I run glibc-2.35 only locally to have a fresh upstream release. I don't think i'll be able to get all of nixpkgs to 2.35. I can only provide one-off fixes for packages I use.