Resolve desktop artifact version from HEAD semver tags#5
Conversation
|
ClawSweeper status: review started. I am starting a fresh review of this pull request: Resolve desktop artifact version from HEAD semver tags This is item 1/1 in the current shard. Shard 0/1. This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking. Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted. |
|
Codex review: needs changes before merge. Summary Reproducibility: yes. from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR’s new multi-tag path can be checked directly through resolveDesktopArtifactVersion inputs. Next step before merge Security Review findings
Review detailsBest possible solution: Land a tag-aware resolver, but compare candidate tags with real semver precedence and add a regression case for numeric ordering such as 1.10.0 versus 1.2.0. Do we have a high-confidence way to reproduce the issue? Yes, from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR’s new multi-tag path can be checked directly through resolveDesktopArtifactVersion inputs. Is this the best way to solve the issue? No, not as written; the precedence idea is narrow and maintainable, but the ordering implementation should use semver comparison rather than localeCompare. Full review comments:
Overall correctness: patch is incorrect Acceptance criteria:
What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning low; reviewed against 35721d9a08b2. |
1 similar comment
|
Codex review: needs changes before merge. Summary Reproducibility: yes. from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR’s new multi-tag path can be checked directly through resolveDesktopArtifactVersion inputs. Next step before merge Security Review findings
Review detailsBest possible solution: Land a tag-aware resolver, but compare candidate tags with real semver precedence and add a regression case for numeric ordering such as 1.10.0 versus 1.2.0. Do we have a high-confidence way to reproduce the issue? Yes, from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR’s new multi-tag path can be checked directly through resolveDesktopArtifactVersion inputs. Is this the best way to solve the issue? No, not as written; the precedence idea is narrow and maintainable, but the ordering implementation should use semver comparison rather than localeCompare. Full review comments:
Overall correctness: patch is incorrect Acceptance criteria:
What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning low; reviewed against 35721d9a08b2. |
|
ClawSweeper 🐠 reef update Thanks for the contribution here. ClawSweeper pushed the narrow repair here so review, credit, and history stay together. Validation: bun run fmt:check; bun run lint; bun run typecheck; bun --cwd scripts run test build-desktop-artifact.test.ts fish notes: model gpt-5.5, reasoning low; reviewed against 314437c. |
|
Codex review: needs maintainer review before merge. Summary Reproducibility: yes. from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR exposes a pure resolver path that exercises explicit, tag, and fallback precedence. Next step before merge Security Review detailsBest possible solution: Have a maintainer review and merge the repaired tag-aware resolver if CI passes, preserving the narrower scope that excludes unrelated upstream shell/PATH changes. Do we have a high-confidence way to reproduce the issue? Yes, from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR exposes a pure resolver path that exercises explicit, tag, and fallback precedence. Is this the best way to solve the issue? Yes, the latest PR head is a narrow maintainable solution: it keeps the build-script behavior local, uses semver-aware ordering, and adds focused tests for the corrected precedence rules. Acceptance criteria:
What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning low; reviewed against 35721d9a08b2. |
1 similar comment
|
Codex review: needs maintainer review before merge. Summary Reproducibility: yes. from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR exposes a pure resolver path that exercises explicit, tag, and fallback precedence. Next step before merge Security Review detailsBest possible solution: Have a maintainer review and merge the repaired tag-aware resolver if CI passes, preserving the narrower scope that excludes unrelated upstream shell/PATH changes. Do we have a high-confidence way to reproduce the issue? Yes, from source inspection: current main ignores HEAD tags when resolving desktop artifact versions, and the PR exposes a pure resolver path that exercises explicit, tag, and fallback precedence. Is this the best way to solve the issue? Yes, the latest PR head is a narrow maintainable solution: it keeps the build-script behavior local, uses semver-aware ordering, and adds focused tests for the corrected precedence rules. Acceptance criteria:
What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning low; reviewed against 35721d9a08b2. |
Summary
Credit
This carries forward the tagged desktop artifact version portion of upstream PR pingdotgg#337 while intentionally leaving out the unrelated shell/PATH changes that current main no longer needs.
Fixes #3
Validation
ClawSweeper 🐠 replacement reef notes:
fish notes: model gpt-5.5, reasoning low; reviewed against d947550.