Repository navigation
STF-2455: Release 1.7.0 and 1.7.1 - #299
Conversation
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 14 minutes. View limit details
No actionable comments were generated in the recent review. Walkthrough
Merge Risk:
|
maxmind-db 1.5.0 limits the resources a crafted database can make the decoder use, rejects pointers to pointers, and speeds up lookups. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
0420b07 to
aaa9be3
Compare
7f369e5 to
0279fa5
Compare
rubygems/release-gem signs the gem by running sigstore-cli 0.2.3 with Bundler.unbundled_env. The bundle install step had put json 3.1.0, a RuboCop dependency in Gemfile.lock, into the system gems, so sigstore loaded it and failed: "unknown keywords: allow_blank, create_additions, allow_infinity". The action then pushed 1.7.0 without an attestation. With BUNDLE_PATH set for the job, bundle exec still finds the bundle, and the unbundled sigstore sees only Ruby's default json. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The v1.4.1 "Wait for release to propagate" step runs rubygems-await, which fails on Ruby 4.0 with "undefined method 'dependencies' for Gem::CompactIndexClient::Cache". It retried until the 1.7.0 job timed out. v1.4.2 waits for the release without rubygems-await. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
release-gem catches a signing error, pushes the gem without an attestation, and the job still passes. That is how 1.7.0 shipped unsigned. Ask RubyGems.org for the attestation after the push. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
0279fa5 to
736b702
Compare
Releases 1.7.0 and 1.7.1. 1.7.0 is already on RubyGems.org. 1.7.1 republishes it with a Sigstore attestation.
1.7.0 (released). The only user-facing change is the higher
maxmind-dbfloor,~> 1.5. maxmind-db 1.5.0 also makes lookups faster, so 1.7.0 is a minor release. The commits:MaxMind::DB::InvalidDatabaseError.Gemfile.lockalready resolved 1.5.0, so only its dependency line changes.release.shadded.The 1.7.0 release workflow pushed the gem without an attestation, and then failed. The workflow's
bundle installput json 3.1.0, a RuboCop dependency inGemfile.lock, into the system gems.rubygems/release-gemsigns the gem by running its pinnedsigstore-cli0.2.3 in an unbundled environment. That process loaded json 3.1.0 and failed:unknown keywords: allow_blank, create_additions, allow_infinity. The wait step then failed too, becauserubygems-awaitbreaks on Ruby 4.0. RubyGems.org accepts an attestation only when a gem is pushed, so 1.7.0 cannot get one now.1.7.1 (not released yet). This is a patch release, because the gem's code does not change. The commits:
BUNDLE_PATH: vendor/bundle.bundle exec rake releasestill finds the bundle, and the unbundled sigstore process sees only Ruby's own json. Ruby 4.0's built-in json is 2.18.0. I tested this locally on Ruby 4.0.6 with an empty system gem directory: a frozenbundle installintovendor/bundle, then the same unbundledgem exec sigstore-cli:0.2.3 signcall that the action makes. sigstore got past the json parse and stopped only because no OIDC identity token exists locally.vendor/bundleis in.gitignore, and the gemspec packs onlygit ls-files.rubygems/release-gemto v1.4.2. v1.4.2 waits for the release withoutrubygems-await, whose 0.5.4 version raisedundefined method 'dependencies' for Gem::CompactIndexClient::Cacheuntil the 1.7.0 job failed.The same failure can affect minfraud-api-ruby and MaxMind-DB-Reader-ruby, which still use release-gem v1.4.1 with no
BUNDLE_PATH. minfraud-api-ruby's release PR in this round will get the same changes.mise.lockdoes not change. mise 2026.11.9 refuses to lock Ruby 4.0.9, because its linux-arm64-musl entry would lose the GitHub attestation that the current Ruby 4.0.6 entry records.Part of STF-2455.
🤖 Generated with Claude Code