brew test-bot --only-tap-syntax failing on Linux due to “Invalid byte sequence in utf-8”

[nwhetsell]

Output of brew config

HOMEBREW_VERSION: 4.5.13-6-g00c528b
ORIGIN: https://github.com/Homebrew/brew
HEAD: 00c528bc540fecf08f922e03d98507ba5f5bbea0
Last commit: 5 hours ago
Branch: main
Core tap JSON: 28 Jul 20:55 UTC
HOMEBREW_PREFIX: /home/linuxbrew/.linuxbrew
HOMEBREW_CACHE: /home/runner/.cache/Homebrew
HOMEBREW_CASK_OPTS: []
HOMEBREW_CLEANUP_PERIODIC_FULL_DAYS: 3650
HOMEBREW_COLOR: set
HOMEBREW_CURL_PATH: /usr/bin/curl
HOMEBREW_DEVELOPER: set
HOMEBREW_FAIL_LOG_LINES: 150
HOMEBREW_GITHUB_API_TOKEN: set
HOMEBREW_GIT_EMAIL: [email protected]
HOMEBREW_GIT_NAME: BrewTestBot
HOMEBREW_GIT_PATH: /usr/bin/git
HOMEBREW_LOGS: /home/runner/work/homebrew-lilypond/homebrew-lilypond/logs
HOMEBREW_MAKE_JOBS: 4
HOMEBREW_NO_AUTO_UPDATE: set
HOMEBREW_NO_EMOJI: set
HOMEBREW_NO_ENV_HINTS: set
HOMEBREW_NO_INSTALLED_DEPENDENTS_CHECK: set
HOMEBREW_SORBET_RUNTIME: set
Homebrew Ruby: 3.4.5 => /home/linuxbrew/.linuxbrew/Homebrew/Library/Homebrew/vendor/portable-ruby/3.4.5/bin/ruby
CPU: quad-core 64-bit zen3
Clang: 14.0.0
Git: 2.50.1 => /usr/bin/git
Curl: 7.81.0 => /usr/bin/curl
Kernel: Linux 6.8.0-1031-azure x86_64 GNU/Linux
OS: Ubuntu 22.04.5 LTS (jammy)
Host glibc: 2.35
/usr/bin/gcc: 11.4.0
/usr/bin/ruby: 3.0.2
glibc: N/A
gcc@11: N/A
gcc: N/A
xorg: N/A

Output of brew doctor

All steps passed!

Description of issue

(Please note that I actually think brew doctor is producing no output; I’m getting the apparent output from a GitHub Actions run here.)

Recently (sometime after July 20), running brew test-bot --only-tap-syntax seems to have started including files in a Ruby gem cache when linting, which leads to errors like this. (Or perhaps this is a malformed gemspec file?) I’ve never seen this error before, and I’m not sure if I’ve missed something on my end, so I thought I should post this as a discussion rather than open an issue.

[PatTheMav]

Thank you for opening this post, was about to post about the same issue (and Mike pointed out in the brew test bot repo that discussions are the better place to raise issues like this).

Just like you reported this issue only occurred recently, the last formula update in my custom tap took place in mid-May, which succeeded without issue.

I also noticed that the brew style command finishes without issue on macOS runners, but on Linux runners it installs action lint which then fails.

Here's the more detailed output:

Taps/patthemav/***/home/.cache/gem/specs/index.rubygems.org%443/quick/Marshal.4.8/bundler-2.6.8.gemspec:1:1: F: Lint/Syntax: Invalid byte sequence in utf-8.
  
  13 files inspected, 1 offense detected
  ==> Installing `actionlint` for GitHub Actions checks...
  ==> Fetching downloads for: actionlint
  ==> Downloading https://ghcr.io/v2/homebrew/core/actionlint/manifests/1.7.7
  ==> Fetching dependencies for actionlint: gmp, libffi and shellcheck
  ==> Downloading https://ghcr.io/v2/homebrew/core/gmp/manifests/6.3.0
  ==> Fetching gmp
  ==> Downloading https://ghcr.io/v2/homebrew/core/gmp/blobs/sha256:3dca3544faca889c7389a5fdbd2b5b00582c34a4e14607033573ad3b06ca7882
  ==> Downloading https://ghcr.io/v2/homebrew/core/libffi/manifests/3.5.1
  ==> Fetching libffi
  ==> Downloading https://ghcr.io/v2/homebrew/core/libffi/blobs/sha256:2109450cd961d89edfdf8b0fce916dd9c5c63a15f52a97c443412a62b2ca8264
  ==> Downloading https://ghcr.io/v2/homebrew/core/shellcheck/manifests/0.10.0
  ==> Fetching shellcheck
  ==> Downloading https://ghcr.io/v2/homebrew/core/shellcheck/blobs/sha256:6d0867f144686a5caa025cb15ecac49286654b78e7b89979a54eedc9a0cc9b6b
  ==> Fetching actionlint
  ==> Downloading https://ghcr.io/v2/homebrew/core/actionlint/blobs/sha256:04830cca2357b8cb8cf30a297bbbe5e1e347f5c0320ca67f621805e6f4843a87
  ==> Installing dependencies for actionlint: gmp, libffi and shellcheck
  ==> Installing actionlint dependency: gmp
  ==> Downloading https://ghcr.io/v2/homebrew/core/gmp/manifests/6.3.0
  Already downloaded: /home/runner/.cache/Homebrew/downloads/70a72a71216843d66a953c06ff6337445ce9bc94fae9f0e301e2f59005274a8e--gmp-6.3.0.bottle_manifest.json
  ==> Pouring gmp--6.3.0.x86_64_linux.bottle.tar.gz
  /home/linuxbrew/.linuxbrew/Cellar/gmp/6.3.0: 24 files, 3.9MB
  ==> Installing actionlint dependency: libffi
  ==> Downloading https://ghcr.io/v2/homebrew/core/libffi/manifests/3.5.1
  Already downloaded: /home/runner/.cache/Homebrew/downloads/2a0a00cd44d587830113cc55dfda470280f6667255105214e0e6941d4c8b3802--libffi-3.5.1.bottle_manifest.json
  ==> Pouring libffi--3.5.1.x86_64_linux.bottle.tar.gz
  /home/linuxbrew/.linuxbrew/Cellar/libffi/3.5.1: 19 files, 811.5KB
  ==> Installing actionlint dependency: shellcheck
  ==> Downloading https://ghcr.io/v2/homebrew/core/shellcheck/manifests/0.10.0
  Already downloaded: /home/runner/.cache/Homebrew/downloads/939bfe6f0c1f854f0f44eb595c468ce80fe4741964900f4a22292a34eb148723--shellcheck-0.10.0.bottle_manifest.json
  ==> Pouring shellcheck--0.10.0.x86_64_linux.bottle.tar.gz
  /home/linuxbrew/.linuxbrew/Cellar/shellcheck/0.10.0: 8 files, 33MB
  ==> Installing actionlint
  ==> Pouring actionlint--1.7.7.x86_64_linux.bottle.tar.gz
  /home/linuxbrew/.linuxbrew/Cellar/actionlint/1.7.7: 8 files, 5MB
  ==> Running `brew cleanup actionlint`...
  ==> No outdated dependents to upgrade!
  Error: Lint/Syntax: Invalid byte sequence in utf-8.

[PatTheMav]

So I downloaded the bundler gemspec file (https://rubygems.org/quick/Marshal.4.8/bundler-2.6.8.gemspec.rz) and tried to inspect its contents. IIRC those rz files are compressed using deflate in ruby, so I used Zlib in ruby to inflate the file and tried to inspect it, but it indeed contains bytes that aren't valid utf-8:

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc0 in position 100: invalid start byte

Note that the same file, when trying to read it as plain 'utf-8' also fails (though I would find that to be expected as it's a deflated binary stream):

UnicodeDecodeError: 'utf-8' codec can't decode byte 0x9c in position 1: invalid start byte

But I don't think this is an error, those files aren't supposed to be "plain" 'utf-8', so the question is why Lint/Syntax picks up the gemspec file?

[jasonkarns]

Similar failure here: https://github.com/nodenv/homebrew-nodenv/actions/runs/16630365669/job/47057914670

[jasonkarns]

I'll share that my last successful test-bot job ran on Jul 25, 2025 at 11:03 AM EDT. First failure was Jul 25, 2025 at 2:00 PM EDT - https://github.com/nodenv/homebrew-nodenv/actions/workflows/tests.yml

My test-bot workflow file pins to precise git-sha (Homebrew/actions/setup-homebrew@6c79ae0677a9b0b8046a936b4ada8980885eafb8) for the homebrew action, so I presume nothing in the action itself is responsible. And assuming that the action sets up the latest homebrew available, then I'd wager we're looking at something in here:

$ git log --since="Jul 25, 2025 at 11:03 AM EDT" --until="Jul 25, 2025 at 2:00 PM EDT" --oneline --first-parent
4d14be89e9 (HEAD -> stable, tag: 4.5.13) Merge pull request #20309 from Homebrew/dependabot/all-f8dbd9efe6
4d58e746da Merge pull request #20313 from Homebrew/download_queue_no_tty
bd3461d4b0 Merge pull request #20312 from Homebrew/development_tools_env
0c969c2f82 Merge pull request #20311 from Homebrew/quieter_debug
3562adb376 Merge pull request #20310 from Homebrew/attestation_debug

[PatTheMav]

So I added a step to run brew style --verbose --debug directly and indeed the gemfile is included in the style checks:

Running parallel inspection
Scanning /home/linuxbrew/.linuxbrew/Homebrew/Library/Taps/***/***/Formula/SomeFormula.rb
Scanning /home/linuxbrew/.linuxbrew/Homebrew/Library/Taps/***/***/README.md
Scanning /home/linuxbrew/.linuxbrew/Homebrew/Library/Taps/***/***/home/.cache/gem/specs/index.rubygems.org%443/quick/Marshal.4.8/bundler-2.6.8.gemspec

I very well could be wrong, but doesn't this line imply that any file not matching the suffix rb, sh, or yml is added to ruby_files to be checked by rubocop?

https://github.com/Homebrew/brew/blob/ff4ee5d0b5c214d67a15abbf46722bc375c5f137/Library/Homebrew/style.rb#L60

[PatTheMav]

Looks like this was added on purpose to just add the tap path itself to rubocop for checking (Homebrew/brew#17482).

[jasonkarns]

Interesting this comes down to rubocop. I have a semi-related discussion over at rubocop for how to allow taps to use github's superlinter (with rubocop) but using homebrew's rubocop config.

It's a bummer that the homebrew rubocop config isn't published as a gem that tap authors could just require and use themselves, directly.

[PatTheMav]

I haven't been able to pin this down yet, but for some reason the gemspec cache directory ends up being inside the custom tap directory: /home/linuxbrew/.linuxbrew/Homebrew/Library/Taps/***/***/home/.cache/gem/specs/index.rubygems.org%443/quick/Marshal.4.8/bundler-2.6.8.gemspec

I would have expected that directory to be put much closer to the local user's home directory where it also wouldn't be pulled in to the files checked by rubocop.

[carlocab]

Probably worth filing an issue at Homebrew/brew.

[PatTheMav]

@carlocab and I are Homebrew maintainers. You are good to submit an issue despite not being able to check all the boxes. Just link to this discussion please.

Got it. As I'm often on the receiving end of incomplete issues or pull requests I'm rather careful to be respectful and actually provide the required information and follow the required steps.

[PatTheMav]

Reported as Homebrew/brew#20339

[p-linnane]

Got it. As I'm often on the receiving end of incomplete issues or pull requests I'm rather careful to be respectful and actually provide the required information and follow the required steps.

We get plenty of these too, but when you have a legitimate issue, are friendly, and have already put in some effort...we're more than happy to do what we can! 😄

[jasonkarns]

fixed by Homebrew/brew#20340 ! (for me at least)