ninja_1_11: init at 1.11.1 #354367
ninja_1_11: init at 1.11.1 #354367
Conversation
pkgs/by-name/ni/ninja/package.nix
Outdated
| hash = "sha256-LvV/Fi2ARXBkfyA1paCRmLUwCh/rTyz+tGMg2/qEepI="; | ||
| }; | ||
|
|
||
| "1.12" = fetchFromGitHub { |
There was a problem hiding this comment.
| "1.12" = fetchFromGitHub { | |
| "latest" = fetchFromGitHub { |
this one should be the latest version, we're not trying to pin it to any specific version.
There was a problem hiding this comment.
Fair enough; pushed. Hopefully this is gone by the time 1.13 comes around either way.
emilazy
force-pushed
the
push-ppkxzlxnprsm
branch
from
382907c to
5f0b2a5
Compare
|
De jure Ninja maintainers don’t seem to be actively involved; cc @JohnRTitor @vcunat @NickCao @AndersonTorres who worked on this derivation since 2023. |
| @@ -11,18 +11,35 @@ | |||
| , buildPackages | |||
| , buildDocs ? true | |||
| , nix-update-script | |||
| , ninjaRelease ? "latest" | |||
There was a problem hiding this comment.
| , ninjaRelease ? "latest" | |
| , use_1_11 ? false |
nit. a bool rather than a string would eliminate the error case of an invalid string and make the below assert unnecessary
pkgs/by-name/ni/ninja/package.nix
Outdated
| rev = "v1.12.1"; | ||
| hash = "sha256-RT5u+TDvWxG5EVQEYj931EZyrHUSAqK73OKDAascAwA="; | ||
| }; | ||
| }; |
There was a problem hiding this comment.
| }; | |
| } or (throw "Unsupported Ninja release."); |
pkgs/by-name/ni/ninja/package.nix
Outdated
|
|
||
| assert lib.assertMsg (releases ? ${ninjaRelease}) "Unsupported Ninja release: ${ninjaRelease}"; | ||
|
|
There was a problem hiding this comment.
| assert lib.assertMsg (releases ? ${ninjaRelease}) "Unsupported Ninja release: ${ninjaRelease}"; |
Swift 5.8 doesn’t reliably build with Ninja 1.12, and apparently neither do old versions of ROCm, due to a build scheduling change breaking builds with undeclared dependencies. Upstream Swift have pinned 1.11 for now. Reintroducing it isn’t a good long‐term solution, but seems like the easiest way to avoid shipping 24.11 with a broken Swift on Darwin. The patch for 32‐bit systems has not been restored, as Swift doesn’t support them anyway.
emilazy
force-pushed
the
push-ppkxzlxnprsm
branch
from
5f0b2a5 to
05ccba8
Compare
|
Implemented a variant of @azuwis’s suggestion. I prefer not using a boolean here and it seems pretty compact now. |
|
Going to take the thumbs up as the closest thing we’ll get to Ninja maintainer approval here :) |
|
Result of 1 package built:
|
|
I want to study that patch from in memoriam @amjoseph that enables multiple versions. IIRC it populates passthru.versions with the various derivations. |
|
This change unfortunately breaks $ nix-prefetch -I nixpkgs=. ninja
error:
… while evaluating attribute 'log'
… while calling the 'head' builtin
at /home/n/git/nixpkgs/lib/attrsets.nix:1574:11:
1573| || pred here (elemAt values 1) (head values) then
1574| head values
| ^
1575| else
… while evaluating the attribute 'name'
at /home/n/git/nixpkgs/pkgs/stdenv/generic/make-derivation.nix:336:7:
335| // (optionalAttrs (attrs ? name || (attrs ? pname && attrs ? version)) {
336| name =
| ^
337| let
(stack trace truncated; use '--show-trace' to show the full, detailed trace)
error: attribute 'rev' missing
at /home/n/git/nixpkgs/pkgs/by-name/ni/ninja/package.nix:19:34:
18| pname = "ninja";
19| version = lib.removePrefix "v" finalAttrs.src.rev;
| ^
20| |
|
I suspect |
|
I think the issue here is that the version depends on the rev, unlike most of the nixpkgs where it does not. And the tools were written in assumption to that. |
|
It would be easy to |
The changes introduced with NixOS#354367 lead currently to an evaluation error when prefetching vendored go dependencies with `nix-prefetch`. Although this is probably a bug in `nix-prefetch`, I prefer to use `nurl` instead as it does the same thing, seems better maintained and doesn't block k3s updates even longer.
The changes introduced with NixOS#354367 lead currently to an evaluation error when prefetching vendored go dependencies with `nix-prefetch`. Although this is probably a bug in `nix-prefetch`, I prefer to use `nurl` instead as it does the same thing, seems better maintained and doesn't block k3s updates even longer.
The changes introduced with NixOS#354367 lead currently to an evaluation error when prefetching vendored go dependencies with `nix-prefetch`. Although this is probably a bug in `nix-prefetch`, I prefer to use `nurl` instead as it does the same thing, seems better maintained and doesn't block k3s updates even longer.
Swift 5.8 doesn’t reliably build with Ninja 1.12, and apparently neither do old versions of ROCm, due to a build scheduling change breaking builds with undeclared dependencies. Upstream Swift have pinned 1.11 for now.
Reintroducing it isn’t a good long‐term solution, but seems like the easiest way to avoid shipping 24.11 with a broken Swift on Darwin. The patch for 32‐bit systems has not been restored, as Swift doesn’t support them anyway.
cc @azuwis @JohnRTitor. I think this is a better approach than
overrideAttrs. I confirmed thatnix-updatecan still update the package in this form.Things done
nix.conf? (See Nix manual)sandbox = relaxedsandbox = truenix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage./result/bin/)Add a 👍 reaction to pull requests you find important.