bazel: 0.26.1 -> 0.27.0

[guibou]

• Upgrade to bazel 0.27
• Fixs for newly introduced bin/bash hardcoded reference
• Bazel now references remote_java_tools_xxx which contains prebuilt
  binaries. We prefetch them, fix them, and force bazel to use the
  fixed repository.

It also closes #63096

Motivation for this change

Bazel upgrade and bug fixing.

Note: instead of fixing #63096 and upgrading to bazel in two separates commits, I've done everything in once. I think it reduce the pull request pressure.

There is no test which proves that I really fixed stuffs, but I'm now able to build rules_haskell with bazel 0.27, which was not possible due to #63096. See tweag/rules_haskell#956

Things done

• Tested using sandboxing (nix.useSandbox on NixOS, or option sandbox in nix.conf on non-NixOS)
• 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 pkgs that depend on this change using nix-shell -p nix-review --run "nix-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• Determined the impact on package closure size (by running nix path-info -S before and after)
• Assured whether relevant documentation is up to date
• Fits CONTRIBUTING.md.

---

[guibou]

Please, wait before merging. I'm having a regression in rules haskell and I'd like to understand if it comes from this change or not.

That's OK. I screwed stuff by a last minute refactoring, but that's OK now.

[FRidh]

Duplicate of #63451.

[Profpatsch]

Ah, can you cherry pick the first commit of #63451?

[guibou]

Ah, can you cherry pick the first commit of #63451?

I rather prefer fixing the test instead. I'll push that in a few minutes.

[guibou]

@Profpatsch I've pushed the repair of the installCheck and the python3 update. I'm currently running the bazel.passthru.tests I was unaware of.

Offline you suggested that I refactor the installCheck to move it to the baze.passthru.tests. I suggest to postpone this refactor, because I think that the important task is to get a working bazel back in nixpkgs.

But I totally agree that later we may move all the tests in bazel.passthru.tests, set it as an always run / fixed output derivation. This will enable network access and allows us to get ride of most hacks.

[Profpatsch]

I suggest to postpone this refactor, because I think that the important task is to get a working bazel back in nixpkgs.

Yeah, but for that we need to turn the installCheck phase off completely. I just don’t want to have bazel builds that fail at the last possible moment anymore.

[Profpatsch]

@GrahamcOfBorg build bazel.tests

[guibou]

Yeah, but for that we need to turn the installCheck phase off completely. I just don’t want to have bazel builds that fail at the last possible moment anymore.

I'm not sure I understand this sentence. Is there something I need to do to close this PR?

[guibou]

@GrahamcOfBorg build bazel.tests

I've rebased on master and fixed the deps sha256 to fix the darwin build failure.

[guibou]

@GrahamcOfBorg build bazel.tests

[guibou]

I amended some style only changes to match nixpkgs style guide, based on recommendations by @flokli.

[guibou]

@kalbasit

I suggest moving the bazelrc into $out/etc or $out/share. It feels a bit odd to have it at the top-level of $out.

Done. You are right. I used $out/etc, to match with default bazel which looks for the file in /etc.

Also, aren't you missing test and run?

The bazel docs says that test and run options "Inherits all options from build.", so I think we are good. I actually tested with bazel test and it works (but not tested with bazel run).

[flokli]

Do we run all tests? I see the hello_test function, but not sure if that's everything we could/should do…

[kalbasit]

@guibou can you also include a Java test, similar to the existing Python test? Having a test will ensure Bazel will not be updated broken again.

[guibou]

@guibou can you also include a Java test, similar to the existing Python test? Having a test will ensure Bazel will not be updated broken again.

@kalbasit There is already a java test, it was in the disabled (but now re-enabled) checkPhase. It is simple, but it shows the ijar issue we previously had.

[guibou]

Do we run all tests? I see the hello_test function, but not sure if that's everything we could/should do…

We can do better, for sure.

The point of this PR was to fix the bazel currently included in nixpkgs. I did the upgrade to bazel 0.27 at the same time because I see no point duplicating effort here. There is a lot of other stuffs we can do to get a perfect bazel nix derivation, but I think that's not the priority now considering that the bazel which is currently provided by nixpkgs is broken.

[guibou]

@GrahamcOfBorg build bazel.tests

[flokli]

The derivation ran

//examples/cpp:hello-success_test                                        PASSED in 0.0s
//examples/java-native/src/test/java/com/example/myproject:hello         PASSED in 0.2s

I guess this is enough coverage to see at a quick glimpse if anything is utterly broken.

We can probably add more tests in a later PR, nothing that should block this bump, given it also fixes the bazel build currently being entirely broken.