Repository navigation
fix: preserve build, browser capture, and reporter coverage provenance - #1748
Conversation
The e2e coverage report's region percentages are a V8 artifact: the same commit, suite and fixture builds split into a different number of ranges on a different runtime, so the denominator moves while the code does not. CONTRIBUTING.md measures that drift and tests/e2e-coverage-gate.test.mjs pins the Node major its floors were measured on, but both guard the job rather than the artifact. The report stamped revision -- which code -- and nothing about which runtime chose the denominator, so a report read out of the 90-day upload, or pasted into a pull request from a local render, was indistinguishable from a CI one measured on the pinned major. Record node and v8 on the success envelope and on the failure envelope, and render a Runtime: line beneath Revision: so the step summary and the .txt artifact carry it too. v8 and not just node because V8 is what splits the ranges, and it moves between patch releases of one Node major -- the difference CONTRIBUTING.md warns is runtime drift, not a code change. Additive: no existing key changes meaning, no threshold moves, and the e2e-coverage job already tees the rendered text into its summary. Closes #1747 Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com>
|
Important Held for human review by the hive's ACMM level gate. This PR was opened by the "quality" agent while Hive policy required a human checkpoint for that agent. Agent PRs are held at ACMM L3–L5; an Hive will keep the |
✅ Deploy Preview for endusers ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
![hivecommons-hive[bot]](https://avatars.githubusercontent.com/in/3568013?s=60&v=4){kind=small}
There was a problem hiding this comment.
Read the tree at c3420ae, not just the diff. This looks correct to me.
Verified:
- The
runtimestamp reaches every envelope: the full build path starts fromemptyReport(run)(tests/tools/e2e-coverage-report.mjs:813), which now carriesruntime: reportRuntime()(line 770), and the failure envelope inmain's catch stamps it independently (line 1114). So the normal non-empty success report gets the key too, not just the trivially-empty one. - Render falls back per-field (
report.runtime?.node ?? 'unknown', lines 1034-1035), so old schema-v2 JSON without the key still renders a line instead of throwing — matching theRevision: unknownpattern directly above it. - Ran
node --test tests/e2e-coverage-runtime.test.mjsat this head afternpm ci: 3/3 pass, including the failure-envelope case which asserts against the liveprocess.version/process.versions.v8rather than literals. - Overlap claim checked against the queue: no other open PR touches
tests/tools/e2e-coverage-report.mjs.
Additive, no workflow edit needed, body matches the diff. No findings.
— hive: agent=reviewer backend=copilot model=claude-fable-5 copilot=1.0.88
|
The maintainer has explicitly authorized a temporary hold release for the agreed P2 provenance repairs: distinguish build/capture producer metadata from reporter runtime, preserve capture identity across re-rendering, and validate success/failure behavior. I will restore the hold after updating this PR. No merge is authorized. |
Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com>
Record browser and build identity at production time, seal it with the run and distinguish the current reporter runtime in success and error output. Preserve unknown legacy provenance and explicit capture/sealing diagnostics. Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
mrbobbytables
changed the title
|
Published the maintainer-approved provenance repairs at 9841db1. Build and capture producer identities are now sealed separately from the current reporter runtime, including Chromium/V8 and Playwright identity, legacy unknowns and failure diagnostics. Updated the PR description with precise scope and local evidence. The temporary release is finished: the hold label is restored, and this PR remains unmerged for human review. No coverage thresholds, workflow scope or cross-chunk attribution policy changed. |
Distinguish NoCoverage accounting from assertion gaps and preserve build/capture producer identity independently of the reporter. Keep runner architecture, mutation scope, thresholds and test-policy restrictions unchanged. Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
|
Added the requested dated clarification to ADR 0007: calibrate mutation attribution before interpreting NoCoverage as assertion gaps, and preserve build/capture identities independently of the current reporter across re-renders and failures. Legacy producer metadata remains explicitly unknown. No new runner architecture, scope expansion, threshold change or relaxation of existing test-policy restrictions is approved by this clarification. Updated the PR description; hold remains in place and the PR is unmerged. |
|
The maintainer has now explicitly authorized adding this PR and #1754 to the normal merge queue. Releasing hold for that authorized submission. Required repository-validation and lint checks are green; remaining checks and merge-group validation continue through the normal workflows. No protections or checks are bypassed. |
🔒 Hold-gate integrity: branch changed while hold-gated — fresh review requiredThis PR's branch moved while it sat hold-gated: the head recorded when the hold was applied was Commits not present in the hold-time snapshot (1):
A human should review the FULL diff at the current head — a rebase or force-push renames every commit, so everything above needs eyes even if it looks familiar. Removing the hold label after that review is the fresh approval: the guard has re-pinned its snapshot to the current head, so a clean lift re-opens the merge lanes. |
1 participant
Summary
Correct the runtime-provenance proposal so later report rendering cannot
misidentify the toolchain that produced browser coverage.
and Chromium product/V8 versions obtained through
Browser.getVersion.manifest. Re-rendering reuses those identities, even if the current build
metadata has changed.
buildRuntimes,captureRuntimesand the live
reporterRuntime. Text output labels all three separately.produce error artifacts, malformed capture JSON retains explicit provenance
warnings, and malformed build metadata fails sealing visibly.
process's V8; no single runtime stamp is presented as proof of what caused
region-count drift.
preserving producer identities separately from reporter identity. It selects
no new mutation architecture and changes no scopes, thresholds or existing
test-policy restrictions; pilot measurements remain on [quality] Pilot: callable-core mutation attribution for CLI validators #1752.
This replaces the original reporter-only
runtimeproposal, not an existingmerged public contract. Normal noncoverage production builds, coverage
thresholds, mutation scope, workflows and conservative cross-chunk attribution
are unchanged. This does not resolve #1736.
Verification
harness 100% lines / 99.77% regions, with every per-file floor preserved.
and dependency audit passed.
HeadlessChrome/156.0.8078.4 and browser V8 15.6.75, separately from local Node
v26.11.1 / V8 14.6.202.34-node.37.
not a claim that the full variant-build browser-coverage lane ran locally.
changed-marker re-rendering, main/variant builds and capture failures.
Maintainer disposition
The maintainer approved these bounded repairs and the subsequent ADR
clarification. The
holdlabel is restored and remains in place. Leave this PRunmerged for review.
Current main, including #1753, is merged into this existing branch;
no competing provenance PR was created.
Closes #1747.