4grayandClaude Opus 5 3300ed13ea ci(release): only a confirmed 404 excuses a failed draft delete ...

Greptile and Codex both flagged the same hole in the sweep's error
handling: `gh api` exits 1 for every failure alike, so a rate limit,
permission loss or outage that hit the DELETE would hit the follow-up GET
too — and the GET's nonzero exit was read as proof the draft was already
gone. `failed` stayed 0 and the sweep went green over a draft it never
deleted.

Read the response status instead of the exit code. `gh api -i` prints the
status line even for an error status, so a genuine 404 (the event job
racing the sweep to the same draft) is distinguishable from everything
else; a request that never got a response leaves the status empty, which
is not 404 and so stays a failure.

Verified against the live API: a deleted release re-checks as 404 and is
excused, an existing release re-checks as 200 and fails the job, and an
unreachable host yields no status and fails the job.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

2026-09-19 20:37:56 +02:00

..

codeql

test(stalker): enforce portal auth in the mock and cover the full-portal flow (#1324)

2026-08-01 21:52:09 +02:00

ISSUE_TEMPLATE

chore: back to electron + nx

2025-10-07 18:02:13 +02:00

workflows

ci(release): only a confirmed 404 excuses a failed draft delete

2026-09-19 20:37:56 +02:00

actionlint.yaml

ci: cut PR runner waste and harden workflow permissions (#1226)

2026-07-25 14:37:40 +02:00

dependabot.yml

ci(deps): split sensitive dependency updates (#1409)

2026-08-11 02:56:50 +02:00

FUNDING.yml

Update FUNDING.yml

2024-06-30 18:54:51 +02:00

pull_request_template.md

chore(release): author release notes in .changes instead of reconstructing them (#1256)

2026-07-25 18:22:43 +02:00