tracexec: Packaging improvements

[nh2]

Original PR title: tracexec: init at 0.1.0

Description of changes

See https://github.com/kxxt/tracexec

❤️ co-maintainers would be very appreciated!

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 24.05 Release Notes (or backporting 23.05 and 23.11 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.

---

Add a 👍 reaction to pull requests you find important.

[kxxt]

Thanks for your packaging effort! I have added some review comments and the test and licenses. Apart from that, there's some extra complexity for handling different architectures: seccomp-bpf feature isn't available for riscv64 architecture due to seccompiler doesn't support it: rust-vmm/seccompiler#72 .

In general, you can take a look at my PKGBUILD at AUR: https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=tracexec

And a kind reminder that in the future tests should be run single threaded, with RUST_TEST_THREADS=1

[nh2]

seccomp-bpf feature isn't available for riscv64 architecture due to seccompiler doesn't support it: rust-vmm/seccompiler#72 .

Done.

And a kind reminder that in the future tests should be run single threaded, with RUST_TEST_THREADS=1

I'm skipping over this as https://github.com/kxxt/tracexec/releases/tag/v0.2.2 mentions this as fixed.

[nh2]

Ah, looks like tracexec was already packaged meanwhile:

• tracexec: init at 0.2.2 #311809

I'll apply the improvements made here onto it.

[nh2]

Good to merge from my side; if ofborg passes, please somebody merge.

[WilliButz]

@ofborg build tracexec

[WilliButz]

Hmm, this seems to introduce a build failure in the check phase on aarch64-linux:

Relevant log:

buildPhase completed in 2 minutes 43 seconds
Running phase: checkPhase
Executing cargoCheckHook
++ cargo test -j 4 --profile release --target aarch64-unknown-linux-gnu --frozen -- --test-threads=4 --skip=cli::test::log_mode_without_args_works --skip=tracer::test::tracer_emits_exec_event
   Compiling proc-macro2 v1.0.81
   Compiling unicode-ident v1.0.12
   Compiling cfg-if v1.0.0
   Compiling libc v0.2.154
   Compiling autocfg v1.3.0
   Compiling memchr v2.7.2
error: linker `aarch64-linux-gnu-gcc` not found
  |
  = note: No such file or directory (os error 2)

error: could not compile `proc-macro2` (build script) due to 1 previous error
warning: build failed, waiting for other jobs to finish...
error: could not compile `libc` (build script) due to 1 previous error

Edit: ofborg logs: https://logs.ofborg.org/?key=nixos/nixpkgs.310158&attempt_id=2fc50c54-2fc3-4592-9531-96631b6355dd

[DontEatOreo]

It would be nice to have updateScript to automate updates

passthru.updateScript = gitUpdater {
  url = "https://github.com/kxxt/tracexec";
  rev-prefix = "v";
};

[fpletz]

Sorry, didn't see your PR. Thanks! ❤️

[nh2]

Hmm, this seems to introduce a build failure in the check phase on aarch64-linux:

error: linker `aarch64-linux-gnu-gcc` not found

@WilliButz No clue about that one. Should Rust aarch64 have a working GCC out of the box?

Edit:

I succeed to build it cross, which suggest to me that maybe something not specific to this package is broken:

NIX_PATH=nixpkgs=. nix-build '<nixpkgs>' --arg crossSystem '(import <nixpkgs/lib>).systems.examples.aarch64-multiplatform' --no-out-link -A tracexec

[kxxt]

Hmm, this seems to introduce a build failure in the check phase on aarch64-linux:

error: linker `aarch64-linux-gnu-gcc` not found

@WilliButz No clue about that one. Should Rust aarch64 have a working GCC out of the box?

I defined it here: https://github.com/kxxt/tracexec/blob/main/.cargo/config.toml . It's because rust defaults to linking with cc and that will fail with a recent gcc(e.g. gcc13/14) when cross-compiling from x86_64 host to aarch64 target.

[nh2]

passthru.updateScript = gitUpdater { url = "https://github.com/kxxt/tracexec"; rev-prefix = "v"; };

@DontEatOreo I used the more generic

passthru.updateScript = nix-update-script { };

mentioned in the Tip of https://nixos.org/manual/nixpkgs/stable/#var-passthru-updateScript which works for this package and does not copy-pasting the Github URL.

Tested with:

NIX_PATH=nixpkgs=. nix-shell maintainers/scripts/update.nix --argstr package tracexec

[nh2]

I defined it here: https://github.com/kxxt/tracexec/blob/main/.cargo/config.toml . It's because rust defaults to linking with cc and that will fail with a recent gcc(e.g. gcc13/14) when cross-compiling from x86_64 host to aarch64 target.

@kxxt What would be a good solution to that?

As in, can that be written such that it does not break native (non-cross) compilation, or should we work around that in nixpkgs somehow?

(The cross-compilation of the package currently works in nixpkgs btw, I just tested it and edited put the invocation into #310158 (comment), in case that is useful for you).

[nh2]

@kxxt

How does this work in the AUR package (which only has 0.1.0 but that already has the relevant code kxxt/tracexec@b8d29bd)? Is the compiler just always called/aliased aarch64-linux-gnu-gcc there?

[nh2]

I suspect aarch64 worked (ofborg log) in the initial packaging because that had tests disabled (doCheck = false;) and the problem only occurs in checkPhase.

I'll disable tests for non-x86_64 archs until a solution is found.

[DontEatOreo]

LGTM!

[kxxt]

@kxxt

How does this work in the AUR package (which only has 0.1.0 but that already has the relevant code kxxt/tracexec@b8d29bd)? Is the compiler just always called/aliased aarch64-linux-gnu-gcc there?

Because when host and target architecture is the same, a compatible symlink is there, at least on Arch Linux(and Ubuntu or GitHub runner image) . (e.g. aarch64-linux-gnu-gcc symlinked to gcc).

[Qyriad]

pname probably shouldn't be reused for the license directory

[nh2]

With disabled tests on aarch64, ofborg passes, merging.