You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 523fed7
Browse filesBrowse the repository at this point in the historyBrowse files
- Installed and exercised: `impact-gate==0.4.1`, `lizard==1.23.0`. Lizard 1.24.0 is excluded by upstream ImpactGate package metadata; it is not an upgrade candidate for this rollout.
10
10
- Local measurement runtime: Python 3.13 on macOS arm64. Dependencies resolved during calibration: PyYAML 6.0.3, pygments 2.21.0, pathspec 1.1.1. Runtime timings are local evidence, not hosted-runner guarantees.
11
+
- CI installs `.github/impact-gate/requirements.txt`, which pins those transitive dependencies at the same versions with hashes for every distribution, using `pip install --require-hashes --no-deps`. Its header records the regeneration command; Dependabot proposes pip updates for that directory.
11
12
- Official release source: [CLI](https://github.com/officefloor/ImpactGate/blob/v0.4.1/impact_gate/cli.py), [baseline](https://github.com/officefloor/ImpactGate/blob/v0.4.1/impact_gate/baseline.py), [measurement scope](https://github.com/officefloor/ImpactGate/blob/v0.4.1/impact_gate/core/config.py). Flags and baseline serialization were also inspected in the installed wheel.
12
13
13
14
## Parser checks
@@ -35,7 +36,7 @@ The upstream baseline serializes `_meta` (`tool`, `n`, `head`, `base_ref`) plus
35
36
36
37
## Actions verification still required
37
38
38
-
Local calibration and tests cannot verify GitHub cache scoping, main-run cache saves, later PR restores, or the rendered job summary. After the workflow is available in Actions, record links to a cold run, a main run saving the pair, and a PR run restoring it. Verify both report sections, policy/version cache invalidation, malformed-pair rebuilding, and explicit seed-only disclosure. Do not mark these checks complete based on local cache simulation.
39
+
Local calibration and tests cannot verify GitHub cache scoping, cache saves, later PR restores, or the rendered job summary. Any run that rebuilds the pair saves it, pull requests included; a PR's cache is scoped to its own ref, so it cannot replace main's, and it spares later pushes to that PR a full rebuild. Main pushes run in per-SHA concurrency groups so a quick second merge cannot cancel the first before it saves. After the workflow is available in Actions, record links to a cold run, a main run saving the pair, a PR run restoring it, and a PR run reusing its own saved pair. Verify both report sections, policy/version cache invalidation, malformed-pair rebuilding, and explicit seed-only disclosure. Do not mark these checks complete based on local cache simulation.
0 commit comments