Split out of issue #44, whose own comment thread flags this as "still open after that PR", and out of the review of PR #45, which declined to widen scope to cover it.
PR #45 fixes the canonical Go test target. It does not fix the repos that already copied the broken one. Every Go repo that vendored the pre-fix target has the false green live right now: go test replays cached successful results, so the target can exit 0 having executed no test.
Required change per repo, in the test target, both invocations:
test:
@go test -count=1 -timeout 90s -race -cover ./... ||\
{echo"--- Rerunning with -v for details ---";\
go test -count=1 -timeout 90s -race -v ./...;exit 1;}
The rerun needs the flag too: without it a failing run replays from cache instead of reproducing the failure, which is exactly when the output matters.
Same shape as issue #35 and issue #39: ONE commit per repo, one PR per repo, not a fleet-wide sweep in a single change.
Blocked on PR #45 merging first — the canonical text has to be right before it is copied outward.
Note for whoever does the enumeration: build the repo list from "repos vendoring any canonical revision", not just the current one. The review of PR #55 found that enumerating only the newest canonical revision silently misses repos one revision behind.
Split out of [issue #44](https://git.eeqj.de/sneak/prompts/issues/44), whose own comment thread flags this as "still open after that PR", and out of the review of [PR #45](https://git.eeqj.de/sneak/prompts/pulls/45), which declined to widen scope to cover it.
[PR #45](https://git.eeqj.de/sneak/prompts/pulls/45) fixes the **canonical** Go `test` target. It does not fix the repos that already copied the broken one. Every Go repo that vendored the pre-fix target has the false green live right now: `go test` replays cached successful results, so the target can exit 0 having executed no test.
Required change per repo, in the `test` target, **both** invocations:
```makefile
test:
@go test -count=1 -timeout 90s -race -cover ./... || \
{ echo "--- Rerunning with -v for details ---"; \
go test -count=1 -timeout 90s -race -v ./...; exit 1; }
```
The rerun needs the flag too: without it a failing run replays from cache instead of reproducing the failure, which is exactly when the output matters.
Already done: [sneak/cattbox](https://git.eeqj.de/sneak/cattbox) (see [cattbox issue #36](https://git.eeqj.de/sneak/cattbox/issues/36)).
Same shape as [issue #35](https://git.eeqj.de/sneak/prompts/issues/35) and [issue #39](https://git.eeqj.de/sneak/prompts/issues/39): **ONE commit per repo**, one PR per repo, not a fleet-wide sweep in a single change.
Blocked on [PR #45](https://git.eeqj.de/sneak/prompts/pulls/45) merging first — the canonical text has to be right before it is copied outward.
Note for whoever does the enumeration: build the repo list from "repos vendoring **any** canonical revision", not just the current one. The review of [PR #55](https://git.eeqj.de/sneak/prompts/pulls/55) found that enumerating only the newest canonical revision silently misses repos one revision behind.
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.
Split out of issue #44, whose own comment thread flags this as "still open after that PR", and out of the review of PR #45, which declined to widen scope to cover it.
PR #45 fixes the canonical Go
testtarget. It does not fix the repos that already copied the broken one. Every Go repo that vendored the pre-fix target has the false green live right now:go testreplays cached successful results, so the target can exit 0 having executed no test.Required change per repo, in the
testtarget, both invocations:The rerun needs the flag too: without it a failing run replays from cache instead of reproducing the failure, which is exactly when the output matters.
Already done: sneak/cattbox (see cattbox issue #36).
Same shape as issue #35 and issue #39: ONE commit per repo, one PR per repo, not a fleet-wide sweep in a single change.
Blocked on PR #45 merging first — the canonical text has to be right before it is copied outward.
Note for whoever does the enumeration: build the repo list from "repos vendoring any canonical revision", not just the current one. The review of PR #55 found that enumerating only the newest canonical revision silently misses repos one revision behind.
clawbot referenced this issue2026-09-03 23:07:18 +02:00
clawbot referenced this issue2026-09-03 23:12:36 +02:00
Closed: filed by the agent system outside the managed fleet that was shut down on 2026-09-06.
Model: fable-5-1