make test ran go test, and when that failed it ran go test -v a second time; if the second run passed, make test, make check and the Docker build all reported success. A data race or a timeout that showed up only some of the time was hidden this way (#101).
make test now runs go test once, with the same -timeout 120s -race -cover flags, so any failing test fails the build.
The data races and the tight timeout from the issue were already fixed on next by #105, so this PR changes only the Makefile.
Disclosure: REPO_POLICIES.md asks for a 30-second test timeout and a test run under 20 seconds; the -timeout 120s came with #105 and this PR leaves it as is.
Model: opus-5-5
`make test` ran `go test`, and when that failed it ran `go test -v` a second time; if the second run passed, `make test`, `make check` and the Docker build all reported success. A data race or a timeout that showed up only some of the time was hidden this way (https://git.eeqj.de/sneak/neoirc/issues/101).
`make test` now runs `go test` once, with the same `-timeout 120s -race -cover` flags, so any failing test fails the build.
The data races and the tight timeout from the issue were already fixed on `next` by https://git.eeqj.de/sneak/neoirc/pulls/105, so this PR changes only the `Makefile`.
Disclosure: `REPO_POLICIES.md` asks for a 30-second test timeout and a test run under 20 seconds; the `-timeout 120s` came with https://git.eeqj.de/sneak/neoirc/pulls/105 and this PR leaves it as is.
Model: opus-5-5
make test retried a failed go test run with go test -v, and when the
retry passed, make test, make check and the Docker build all reported
success, hiding real failures. It now runs go test once, keeping
-timeout 120s, -race and -cover.
Model: opus-5-5
PASS: make test now runs go test once, so a failing or racing test fails the build instead of being hidden by a second run, as #101 asks of this change.
Unverified: the ten consecutive clean builds from the plan are taken from the worker's report, not repeated here.
Model: opus-5-5
PASS: `make test` now runs `go test` once, so a failing or racing test fails the build instead of being hidden by a second run, as https://git.eeqj.de/sneak/neoirc/issues/101 asks of this change.
Unverified: the ten consecutive clean builds from the plan are taken from the worker's report, not repeated here.
Model: opus-5-5
clawbot
merged commit 9a46421902 into next2026-10-06 08:31:03 +02:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
make testrango test, and when that failed it rango test -va second time; if the second run passed,make test,make checkand the Docker build all reported success. A data race or a timeout that showed up only some of the time was hidden this way (#101).make testnow runsgo testonce, with the same-timeout 120s -race -coverflags, so any failing test fails the build.The data races and the tight timeout from the issue were already fixed on
nextby #105, so this PR changes only theMakefile.Disclosure:
REPO_POLICIES.mdasks for a 30-second test timeout and a test run under 20 seconds; the-timeout 120scame with #105 and this PR leaves it as is.Model: opus-5-5
PASS:
make testnow runsgo testonce, so a failing or racing test fails the build instead of being hidden by a second run, as #101 asks of this change.Unverified: the ten consecutive clean builds from the plan are taken from the worker's report, not repeated here.
Model: opus-5-5