notmuch: 0.34.2 -> 0.34.3

[doronbehar]

Motivation for this change

I noticed the tests fail, with and without this version bump. cc @flokli @puckipedia .

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• 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.

[flokli]

Is there an upstream issue about this? Not a fan of simply disabling tests that quickly.

[doronbehar]

Is there an upstream issue about this? Not a fan of simply disabling tests that quickly.

It seems the list of it is here:

https://nmbug.notmuchmail.org/status/

There are many tests that all of a suddenly fail, couldn't find a common denominator for all of them... I also don't have time right now to bisect this. The issue is that this failure blocks many mail clients that depend on notmuch.

[doronbehar]

Here's the failing build log:

https://gist.github.com/doronbehar/9000c47accead127772a32e5615c9471

[flokli]

So can we revert until the version where tests don't fail, instead of bumping to yet another version?

[doronbehar]

So can we revert until the version where tests don't fail, instead of bumping to yet another version?

It's just that even the tests of 0.34.2 and 0.34.1 fail. I suspect this is due to a dependency of notmuch that got upgraded and it's API slightly changed in a way that failed the tests, but hopefully not notmuch itself. Unfortunately, I don't have the resources and/or the time to bisect this and this issue is blocking neomutt users from upgrading their systems.

[flokli]

I'll try to bisect this later this evening.

[flokli]

I successfully bisected this, the notmuch tests started to fail with the gnupg upgrade:

flokli@ws:~/dev/nixos/nixpkgs|1ee8f771320 ⇒  git bisect log
git bisect start
# bad: [21dc827a948da8302b5b08cfdc81d58053a8e318] solana-testnet: init at 1.9.2 (#152055)
git bisect bad 21dc827a948da8302b5b08cfdc81d58053a8e318
# good: [971c10bb232cefba0840b65da3a61e61255ec8a9] Merge pull request #153466 from bryanasdev000/mdzk
git bisect good 971c10bb232cefba0840b65da3a61e61255ec8a9
# good: [91917122d3c8f22316c2f6128af1cc89988302c6] Merge pull request #126718 from roberth/testEqualDerivation
git bisect good 91917122d3c8f22316c2f6128af1cc89988302c6
# bad: [bfc5fca3426bafa452b466e4b09791563db91870] Merge master into staging-next
git bisect bad bfc5fca3426bafa452b466e4b09791563db91870
# good: [0f85f14d71103bf30c8f33f48261ced0af5e7bd9] python3Packages.s3fs: 2021.11.0 -> 2021.11.1
git bisect good 0f85f14d71103bf30c8f33f48261ced0af5e7bd9
# bad: [88d66d43080776f2cff363278d2d6b6f43ffa16e] Merge pull request #151672 from r-ryantm/auto-update/xwayland
git bisect bad 88d66d43080776f2cff363278d2d6b6f43ffa16e
# good: [3e2d1c1e6578add783f9f15921eadf8b35819246] Merge staging-next into staging
git bisect good 3e2d1c1e6578add783f9f15921eadf8b35819246
# bad: [e8d85c6c8c99cfa96770c143b548c8eb8e245022] kde-frameworks: 5.87 -> 5.89
git bisect bad e8d85c6c8c99cfa96770c143b548c8eb8e245022
# bad: [8d45187fd8b7eaa0e651dfd1ba9036e6e28634e6] Merge staging-next into staging
git bisect bad 8d45187fd8b7eaa0e651dfd1ba9036e6e28634e6
# bad: [7599020b2168d88475677f110d53a96895cb0557] Merge pull request #145434 from colemickens/gpg23
git bisect bad 7599020b2168d88475677f110d53a96895cb0557
# bad: [062de00b5a4d1444f98a3f8ea42b24304075a0d6] libcryptui: use 'gnupg' instead of versioned alias
git bisect bad 062de00b5a4d1444f98a3f8ea42b24304075a0d6
# bad: [55dc828ef658a66fc61128fe05fbdf209d526375] gnupg: fix eval by replaced alias 'libgpg-error'
git bisect bad 55dc828ef658a66fc61128fe05fbdf209d526375
# bad: [1ee8f771320b45651d21e4cdbf15141e1c143887] gnupg: 2.2.27 -> 2.3.3
git bisect bad 1ee8f771320b45651d21e4cdbf15141e1c143887
# first bad commit: [1ee8f771320b45651d21e4cdbf15141e1c143887] gnupg: 2.2.27 -> 2.3.3

commit 1ee8f771320b45651d21e4cdbf15141e1c143887
Author: Cole Mickens <cole.mickens@gmail.com>
Date:   Wed Nov 10 12:47:21 2021 -0800

    gnupg: 2.2.27 -> 2.3.3

 pkgs/tools/security/gnupg/22.nix                   |  97 --------------------
 pkgs/tools/security/gnupg/23.nix                   | 101 +++++++++++++++++++++
 ...-of-previously-known-keys-even-without-UI.patch |  21 +++--
 .../security/gnupg/fix-libusb-include-path.patch   |   2 +-
 pkgs/top-level/all-packages.nix                    |   6 +-
 5 files changed, 116 insertions(+), 111 deletions(-)
 delete mode 100644 pkgs/tools/security/gnupg/22.nix
 create mode 100644 pkgs/tools/security/gnupg/23.nix

[colemickens]

I can't even figure out if this has been reported to notmuch; I looked for a bug report place, I think I found the ML they would end up on and didn't see anything. Maybe I'll try IRC, I guess.

[doronbehar]

We can disable the tests and write a comment that this is related to the gnupg upgrade of 2.2.27 -> 2.3.3 . Then it'd be easy for someone with more time to report this upstream.

[flokli]

We still don't know if the upgrade broke gpg operations with notmuch (and notmuch should be passed an older version of gnupg).

There's a reason these tests fail, we should only disable them as a last resort.

[flokli]

@colemickens would it be possible to resurrect gnupg22 instead, and give this to notmuch?

[stigtsp]

Adding the following to postPatch makes the tests pass

    substituteInPlace test/test-lib.sh \
      --replace '.*fingerprint: ' '.*sha1 fpr: '

gpgsm added both sha1 and sha2 fingerprints with different prefixes to the output of --list-key

https://github.com/notmuch/notmuch/blob/master/test/test-lib.sh#L148

Will send patch to upstream, opening separate nixpkgs PR with fix

[stigtsp]

Fixed in #154547

[flokli]

With #154547 merged and the tests now not being broken anymore, can we put the doCheck back to how it was before? I manually set it to true, and verified the tests now work again.

[doronbehar]

With #154547 merged and the tests now not being broken anymore, can we put the doCheck back to how it was before? I manually set it to true, and verified the tests now work again.

Added again the doCheck line.