nixVersions.nix_2_31: init at 2.31.0#437584
Conversation
|
This pull request has been mentioned on NixOS Discourse. There might be relevant details there: |
pkgs/tools/package-management/nix/modular/src/libstore-tests/package.nix
Outdated
Show resolved
Hide resolved
pkgs/tools/package-management/nix/modular/src/libfetchers-tests/package.nix
Outdated
Show resolved
Hide resolved
|
We should bring over the Boehm GC patch from the parallel marking PR: NixOS/nix@0d559f0. I don't see a problem with making it unconditional though. That would have to go into |
This comment was marked as outdated.
This comment was marked as outdated.
Also, don't try to access cache.nixos.org in the libstore unit tests. We have to disable the test in versions prior to 2.31; that's OK, since it was being skipped when run under Nix anyway before that. (cherry picked from commit 01388b3e78582553cc30ee772db5b5aba8d89edd) Co-authored-by: Philip Taron <philip.taron@gmail.com>
(cherry picked from commit f29acd5bbc0505ea6a78a51e4828a22630dd0ff1)
…-tests This failed on macOS: nix-store-tests-run> C++ exception with description "../nix_api_store.cc:33: nix_err_code(ctx) != NIX_OK, message: error: getting status of '/nix/var/nix/daemon-socket/socket': Operation not permitted" thrown in the test body. (cherry picked from commit 778156072444fdd06812b89e86169bf45f2425df)
If the mark stack size is too small, it greatly inhibits parallel marking, which is very bad for performance on multi-core systems. (cherry picked from commit 0d559f0c1301af9bd2c6751761bff6490962d0ee)
…xecution environment (cherry picked from commit b21304fe4c491b871682fc292d9c31410cbb418a)
d75b88e to
42e6ff4
Compare
|
I cherry-picked upstream commits for all the changes made to the modular packaging. Now I'm running the tests for every packaged version of Nix. |
|
Here's the test results you asked for, @Mic92.
|
|
Known issue |
|
In the past I've prioritized getting the new package in over getting everything using |
I'm itching to press the merge button as soon as I get a ✅. The testing I've done is enough to convince me it's good to go. I had hoped that the Nix team's confidence in making a release would be enough to quickly provide that ✅ but as you note it is slow to arrive. C'est la vie I suppose. I have no problems bumping |
|
Thanks for the very thorough job here, @philiptaron! |
| # small, run Nix with GC_PRINT_STATS=1 and look for messages | ||
| # such as `Mark stack overflow`, `No room to copy back mark | ||
| # stack`, and `Grew mark stack to ... frames`. | ||
| NIX_CFLAGS_COMPILE = "-DINITIAL_MARK_STACK_SIZE=1048576"; |
There was a problem hiding this comment.
It looks like this is breaking anyone who sets env.NIX_CFLAGS_COMPILE globally. Previously there were efforts to ensure we only use env.NIX_CFLAGS_COMPILE and not the attrs.NIX_CFLAGS_COMPILE (see #217206).
Is there any reason this needs to be directly in attrs or was this just a small oversight and we can move to using env?
There was a problem hiding this comment.
Yeah that was an oversight.
There was a problem hiding this comment.
I cherry-picked the upstream commit but I agree that we should not use overrideAttrs (and ideally not use overrides) here.
There was a problem hiding this comment.
Thanks for the quick response! I don't have time at the moment to create a PR, but will do so tonight (just to use env. for now, not avoiding overrideAttrs overall) if no one else gets to it first.
There was a problem hiding this comment.
Nvm, it was such a quick fix I just made the PR now: #439910
There was a problem hiding this comment.
By not using env the current code causes problems for anyone globally overriding compile flags. This fix repairs the expected behavior. Break appears in NixOS#437584.
Diff: NixOS/nix@2.30.2...2.31.0
Changelog: https://discourse.nixos.org/t/nix-2-31-0-released/68465
Previous release PR:
Things done
passthru.tests.nixpkgs-reviewon this PR. See nixpkgs-review usage../result/bin/.