fix(test): restore the community test build and the doc cap on main #812

Merged

coilysiren merged 1 commit from aos/claude/bk79-test-build-break into main

2026-08-15 16:33:54 +00:00

coilyco-ops commented

2026-08-15 16:29:12 +00:00

Member

main does not pass ward gate right now, on two independent counts. Found while picking up #717, whose pinning test is one of the casualties.

1. The community test package does not compile

vet: internal/community/replyfailure_test.go:229:71: not enough arguments in call to agent.deliverOrReport
	have (context.Context, *rejectingTurn, string)
	want (context.Context, turnIO, string, string)

A semantic merge conflict, 22 minutes wide:

feb86b8 08:38 - added replyfailure_test.go, calling deliverOrReport with three arguments.
1568d7f 09:00 - gave deliverOrReport a fourth parameter, updated undelivered_test.go, and did not update a caller that did not exist when that branch started.

Neither touched the other's lines, so git merged both cleanly and the result does not build. Each PR was green on its own branch.

Why this is worse than a red gate. A test package that fails to compile emits zero --- FAIL lines. This repo has already read that as "caught" once (#653), and sirens-echo#678 cites the same trap. Every pin in internal/community is currently silent, including reasoningomitempty_test.go, which is the only thing tracking the live defect in #717.

Fix is the obvious one: pass nothingWithheld, matching what 1568d7f did to the sibling call in undelivered_test.go. The test is about send-failure verdicts, not overflow, so "the message is all of it" is the correct value.

With the package compiling again, TestAnEmptyReasoningContentLeavesNoKey runs and passes, confirming #717's defect is still live rather than silently fixed.

2. A doc is one line over the cap

FAIL: docs/sirens-echo-delivery-failures.md: 81 lines exceeds the 80-line cap.

Arrived at 81 lines via af5816c (#807). Reflowed one paragraph in "What this does not do" from four lines to three. No wording changed, only the line break.

Verification

ward gate PASS on all six steps: build, policy-check, vet, test, test-skips, pre-commit. Both defects reproduce on a clean clone of main, so this is not a local artifact.

Not in this PR

Why two green PRs can merge into a red main. That is a CI question - the gate runs per branch and nothing re-runs it on the merge result - and it is the reason this class recurs rather than a thing to patch in a test file. Worth its own issue if it has not got one.

**`main` does not pass `ward gate` right now, on two independent counts.** Found while picking up #717, whose pinning test is one of the casualties. ## 1. The community test package does not compile ``` vet: internal/community/replyfailure_test.go:229:71: not enough arguments in call to agent.deliverOrReport have (context.Context, *rejectingTurn, string) want (context.Context, turnIO, string, string) ``` A semantic merge conflict, 22 minutes wide: * `feb86b8` 08:38 - added `replyfailure_test.go`, calling `deliverOrReport` with three arguments. * `1568d7f` 09:00 - gave `deliverOrReport` a fourth parameter, updated `undelivered_test.go`, and did not update a caller that did not exist when that branch started. Neither touched the other's lines, so git merged both cleanly and the result does not build. Each PR was green on its own branch. **Why this is worse than a red gate.** A test package that fails to compile emits zero `--- FAIL` lines. This repo has already read that as "caught" once (#653), and `sirens-echo#678` cites the same trap. Every pin in `internal/community` is currently silent, including `reasoningomitempty_test.go`, which is the only thing tracking the live defect in #717. Fix is the obvious one: pass `nothingWithheld`, matching what `1568d7f` did to the sibling call in `undelivered_test.go`. The test is about send-failure verdicts, not overflow, so "the message is all of it" is the correct value. With the package compiling again, `TestAnEmptyReasoningContentLeavesNoKey` runs and passes, confirming #717's defect is still live rather than silently fixed. ## 2. A doc is one line over the cap ``` FAIL: docs/sirens-echo-delivery-failures.md: 81 lines exceeds the 80-line cap. ``` Arrived at 81 lines via `af5816c` (#807). Reflowed one paragraph in "What this does not do" from four lines to three. **No wording changed**, only the line break. ## Verification `ward gate` PASS on all six steps: build, policy-check, vet, test, test-skips, pre-commit. Both defects reproduce on a clean clone of `main`, so this is not a local artifact. ## Not in this PR Why two green PRs can merge into a red `main`. That is a CI question - the gate runs per branch and nothing re-runs it on the merge result - and it is the reason this class recurs rather than a thing to patch in a test file. Worth its own issue if it has not got one.

coilyco-ops added 1 commit

2026-08-15 16:29:12 +00:00

fix(test): restore the community test build and the doc cap on main

ci / image-build (pull_request) Successful in 27s

Details

ci / test (pull_request) Successful in 42s

Details

ci / publish-echo-image (pull_request) Has been skipped

Details

ci / publish-observed (pull_request) Has been skipped

Details

f62382eccc

Two PRs landed 22 minutes apart and their merge does not compile. feb86b8
added replyfailure_test.go calling deliverOrReport with three arguments,
and 1568d7f gave that function a fourth, updating undelivered_test.go and
not the caller that did not exist when it started. Git merged both
cleanly because neither touched the other's lines.

The cost is larger than a red gate. A test package that does not compile
emits zero `--- FAIL` lines, which this repo has already read once as
"caught" (#653), and it silences every pin in the package including the
one sirens-echo#717 depends on.

Separately, docs/sirens-echo-delivery-failures.md arrived at 81 lines and
the cap is 80, so pre-commit failed for anyone on main. Reflowed one
paragraph, no wording changed.

`ward gate` PASS on all six steps.

Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops requested review from coilysiren

2026-08-15 16:29:12 +00:00