Publish evidence-backed QA configuration matrix - #40
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d064befd8a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if not isinstance(environments, list) or not environments: | ||
| errors.append(f"{path.relative_to(ROOT)}: receipt has no environments") | ||
| return errors | ||
| for environment in environments: |
There was a problem hiding this comment.
When a repository receipt backs the two green post-deployment rows, this loop validates only the environments that happen to be present and never requires the expected Paper 1.20.6 and 1.21.11 entries, or even two distinct environments. A truncated receipt with a single passing environment would still pass CI and leave both matrix rows marked verified, overstating public-artifact coverage.
Useful? React with 👍 / 👎.
| name: Post-deployment Paper runtime | ||
| needs: [build-and-smoke, publish] | ||
| runs-on: ubuntu-24.04 | ||
| timeout-minutes: 15 |
There was a problem hiding this comment.
This 15-minute cap covers verify-published-release-runtime.sh, which runs both Paper endpoints in one job; each underlying smoke can legitimately wait up to 300s for startup, 60s for selftest, and 90s for shutdown. If either endpoint is slow but still within its own allowed window, GitHub can cancel the job before a receipt or useful failure evidence is produced; the manual post-release workflow uses the same cap.
Useful? React with 👍 / 👎.
| "size_bytes": 180060 | ||
| }, | ||
| "harness": { | ||
| "commit": "be5807aef8a10fac2fa6ea312f063032b3b98245", |
There was a problem hiding this comment.
The committed receipt points at harness commit be5807a..., but that commit is not an ancestor of the squashed commit under review, so if only this reviewed commit is published reviewers cannot retrieve the exact harness by the recorded SHA. Regenerate the receipt after the final commit, or validate the receipt against the committed script hashes, so the evidence remains durable.
Useful? React with 👍 / 👎.
1 participant
Summary
v1.0.9release artifact after deployment on Paper 1.20.6 and latest stable 1.21Follow-up QA hardening for #28.
Representative post-deployment evidence
v1.0.9435f3855fcf121c6bfef6b558615b22af12ecec9dc398d13f99b15c88b500324fef3282f2374c28452b13f47f49fd8d36fdc9e6c78/78, clean shutdown78/78, clean shutdowndocs/evidence/configuration-matrix/v1.0.9-post-release-runtime.jsonDeliberately visible gaps
Gameplay E2E, intermediate 1.21 patches, Purpur/Pufferfish runtime coverage, WorldGuard/Towny runtime combinations, non-default configuration runtime coverage, and Java >21 remain explicit gaps rather than inferred coverage.
Local validation
mvn -B -ntp clean verify— 93 tests + 38 P0, JaCoCo gate, PIT 85% mutation / 98% line coverage for mutated classes