test: push coverage across collect/jobrunner/gitea
collect: 83.0% -> 90.0% via error medium injection (write/list/read/ensureDir
failures), rate limiter blocking tests, HTTP mock servers for papers/market/
bitcointalk write errors, processor context cancellation, state round-trip,
and GitHub collector gh-auth-failure paths.
gitea: 89.2% -> 94.0% via paginated org/user repo servers, PR meta with
comment counting and nil dates, GetCommentBodies with nil poster, ListPullRequests
state mapping, and NewFromConfig flag override tests.
jobrunner: 86.4% -> 94.4% via journal error paths (empty baseDir, readonly dir,
path sanitisation), poller with error-returning source/handler/report, journal
integration (JSONL file verification), multiple sources, immediate cancellation,
and cycle counter tests.
handlers: 83.8% -> 89.2% via dispatch mock servers (invalid repo, EnsureLabel
failure, GetIssue not found, AssignIssue failure, AddIssueLabels error, issue
with no special labels), completion label errors, EnableAutoMerge pending checks,
PublishDraft merged state, SendFixCommand merge conflict, DismissReviews stale
review, TickParent checkbox ticking, and dual-run mode. Remaining 10.8% is in
SSH-dependent dispatch code (secureTransfer/runRemote/ticketExists) that cannot
be tested without modifying production code.
Co-Authored-By: Charon <charon@lethean.io>
8d1d7fce01