What
#2332 reinstated the forgejo Op generator's findingMode: "issue" refusal
because postOrUpdateGithubIssue's POST and PATCH 401'd against a real
Forgejo instance. #2333 (PR #2350) fixed that credential: ghCredentialEnv
now forwards the resolved token as GH_ENTERPRISE_TOKEN beside GH_TOKEN,
which is the variable gh reads for a host in neither the github.com nor the
ghe.com class.
The refusal's premise is therefore gone. Both halves of
postOrUpdateGithubIssue were driven green during #2333 against a live
codeberg.org/forgejo/forgejo:12 instance reporting 12.0.4+gitea-1.22.0,
from a shell whose GH_CONFIG_DIR held no gh auth login:
first run (expect POST, a new issue): OK -> http://localhost:3000/chantadmin/infra/issues/2
second run (expect PATCH of the same issue): OK -> http://localhost:3000/chantadmin/infra/issues/2
Sticky, not stacking. The paginated read, the .pull_request == null filter
and the marker startswith match were already confirmed correct against the
same image under #2315.
Why it was not done in #2333
#2333 was scoped to the credential. Lifting the refusal is a generator
behavior change with its own surface to settle:
None of that is the credential, and folding it into #2333 would have mixed a
shipped-defect fix with a feature lift.
Do
Lift the findingMode: "issue" refusal in
lexicons/forgejo/src/components/generate-op-pipeline.ts, settle the tests
above, and verify the generated workflow end to end the way #2333 did —
against a real instance, from a shell with no stored gh auth login.
#2333 already corrected that file's module doc comment so it no longer states
the removed premise as fact, and recorded that lifting is filed here.
Refs #2315, #2332, #2333, #2350.
What
#2332 reinstated the forgejo Op generator's
findingMode: "issue"refusalbecause
postOrUpdateGithubIssue's POST and PATCH 401'd against a realForgejo instance. #2333 (PR #2350) fixed that credential:
ghCredentialEnvnow forwards the resolved token as
GH_ENTERPRISE_TOKENbesideGH_TOKEN,which is the variable
ghreads for a host in neither the github.com nor theghe.comclass.The refusal's premise is therefore gone. Both halves of
postOrUpdateGithubIssuewere driven green during #2333 against a livecodeberg.org/forgejo/forgejo:12instance reporting12.0.4+gitea-1.22.0,from a shell whose
GH_CONFIG_DIRheld nogh auth login:Sticky, not stacking. The paginated read, the
.pull_request == nullfilterand the marker
startswithmatch were already confirmed correct against thesame image under #2315.
Why it was not done in #2333
#2333 was scoped to the credential. Lifting the refusal is a generator
behavior change with its own surface to settle:
lexicons/forgejo/src/components/generate-op-pipeline.test.ts("issue finding mode is refused by name (forgejo: the issue finding mode is un-refused and unverified on a real instance #2315)")
findingMode: "issue"to"pull-request"to get them out of the refusal's way, two of which assertexact YAML
cron-triggered issue Op on Forgejo, which is not the same question thepull_request-triggeredcommentmode answeredNone of that is the credential, and folding it into #2333 would have mixed a
shipped-defect fix with a feature lift.
Do
Lift the
findingMode: "issue"refusal inlexicons/forgejo/src/components/generate-op-pipeline.ts, settle the testsabove, and verify the generated workflow end to end the way #2333 did —
against a real instance, from a shell with no stored
gh auth login.#2333 already corrected that file's module doc comment so it no longer states
the removed premise as fact, and recorded that lifting is filed here.
Refs #2315, #2332, #2333, #2350.