The v0.1.0-rc.8 release (run 91) and main's PostgreSQL 16 job (run 90) both failed on one test: DecodeParallelismTests.ADemandLoadWaitsForTheReadAheadOfADevelopingWindow, with Decodes > 4 returning 4.
Cause: the test counted on the window's read-ahead starting on the thread pool before the next frames were requested. On the busy shared runner it had not started, so only the four requested frames were decoded.
Fix: after the first requested frame, the test now waits, within the CI time budget (TestTiming.Budget), until the read-ahead decodes a frame of its own. The real check, at most one development at a time (Peak == 1), is unchanged.
Product code: no changes.
Once this is merged, v0.1.0-rc.8 is tagged again on the new main; its first release run built nothing and published nothing.
The `v0.1.0-rc.8` release (run 91) and `main`'s PostgreSQL 16 job (run 90) both failed on one test: `DecodeParallelismTests.ADemandLoadWaitsForTheReadAheadOfADevelopingWindow`, with `Decodes > 4` returning 4.
- **Cause:** the test counted on the window's read-ahead starting on the thread pool before the next frames were requested. On the busy shared runner it had not started, so only the four requested frames were decoded.
- **Fix:** after the first requested frame, the test now waits, within the CI time budget (`TestTiming.Budget`), until the read-ahead decodes a frame of its own. The real check, at most one development at a time (`Peak == 1`), is unchanged.
- **Product code:** no changes.
Once this is merged, `v0.1.0-rc.8` is tagged again on the new `main`; its first release run built nothing and published nothing.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
ADemandLoadWaitsForTheReadAheadOfADevelopingWindow counted on the
read-ahead starting on the thread pool before the next frames were
asked for. On a busy runner it had not begun, only the four demanded
frames were decoded, and the rc.8 release and main's PostgreSQL 16 job
failed on 'Decodes > 4'. The test now waits, within the CI time budget,
until the read-ahead develops a frame of its own; the one-development
limit it checks is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Alby96
merged commit 30a5b03f37 into main2026-11-04 11:01:21 +01:00
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.
The
v0.1.0-rc.8release (run 91) andmain's PostgreSQL 16 job (run 90) both failed on one test:DecodeParallelismTests.ADemandLoadWaitsForTheReadAheadOfADevelopingWindow, withDecodes > 4returning 4.TestTiming.Budget), until the read-ahead decodes a frame of its own. The real check, at most one development at a time (Peak == 1), is unchanged.Once this is merged,
v0.1.0-rc.8is tagged again on the newmain; its first release run built nothing and published nothing.🤖 Generated with Claude Code