lsm / HyperNeo / 31150867040 / 20

40%

dev: 40%

push

github

fix(github-events): drop inactive deployment statuses + exclude closed PRs (#2387)

* fix(github-events): drop inactive deployment statuses + exclude closed PRs

Follow-up to #2351 fixing two gaps in the deployment/deployment_status
webhook ingestion surfaced in late review.

1. normalizeGitHubDeploymentStatus now returns null for state === 'inactive'
   per the merge-blocking spec (row 4, #2324): deployment_status fires no
   event for inactive states. It previously published
   .deployment_status_inactive, which woke any workflow subscribed to
   pull_request/<id>.* on an ephemeral-env teardown.

2. pickPrNumbersByHeadSha now requires state === 'open' before attributing a
   PR. /commits/{sha}/pulls returns open + closed + merged PRs, so a
   closed/merged PR whose retained head.sha equals the deployed SHA was
   admitted, publishing under a finished PR's topic and waking a stale
   subscription. This also matches resolveDeploymentPrNumbers's "open PR
   numbers" contract.

Tests: inactive status dropped (no publish) at both the normalizer and
handler level; a closed PR sharing the deployed head.sha excluded while the
open PR at the same head publishes.

* fix(github-events): exclude closed PRs from status webhook attribution

Address review on #2387: resolvePullRequestNumbersForCommit (status webhook)
filtered on head.sha only, so a closed/merged PR whose retained head.sha
matched the commit was attributed — the same latent bug the prior commit
fixed for deployments, and a spec row 1 violation ("commit.sha → open PR").
Add the same state === 'open' guard so it genuinely mirrors
pickPrNumbersByHeadSha and the "mirrors" doc comment is accurate again.

Existing status fixtures modeled PR rows without `state` (real /pulls rows
always carry it); added state: 'open' where an attributable open PR is
intended, plus a status test asserting a closed PR at a matching head is
excluded while the open PR at the same head publishes.

1794 of 37677 branches covered (4.76%)

Branch coverage included in aggregate %.

6198 of 46553 relevant lines covered (13.31%)

3.55 hits per line