Repository navigation
pr-review skill: say how to verify that a test proves its claim - #7465
Merged
Merged
Conversation
Pass 9 told the reviewer what a good test looks like but not how to check it. Add the techniques that found the test findings in recent reviews: run the tests (also under another Go toolchain), revert the fix for regression tests, restore scripts from BASE for variant-parity changes, read the PR's Integration Tests run before asking about a real workspace, and check Repls for host-dependent URL tokens. Co-authored-by: Isaac <no-reply@databricks.com>
andrewnester
requested review from
janniklasrose,
pietern and
shreyas-goenka
November 11, 2026 14:18
Collaborator
✅ Integration test reportCommit: bac4986
Top 6 slowest tests (at least 2 minutes):
|
Two review threads on pass 9 asked why the reviewer should run tests locally when CI runs them, and to lean on the regression report instead of reverting the fix by hand. The toolchain bullet now says what a green `gh pr checks` already proves: the `task test (linux, ...)` jobs run the whole suite and the windows and macos cells always keep the tests the PR changed, all under the one toolchain pinned in go.mod. The local run stays only for another Go version, which no job runs; a golden that encodes standard-library output passes CI and fails on the next `Bump Go toolchain` PR, as acceptance/cmd/api/request-handling/output.txt did in #7052. The "revert the fix" sentence now reads the Regression test report comment first: `regression=` and `unreleased=` count the changed acceptance tests that fail on main or on the latest release, so a test under `coverage=` passes without the fix unless the fix is already released. Reverting locally is kept for what the report cannot answer: unit tests, tests beyond its selection limit, and released fixes. Co-authored-by: Isaac <no-reply@databricks.com>
andrewnester
enabled auto-merge
November 11, 2026 18:08
janniklasrose
approved these changes
Nov 11, 2026
Collaborator
❌ Integration test reportCommit: 4eeacdb
201 interesting tests: 106 FAIL, 83 MISS, 9 RECOVERED, 3 KNOWN
Top 50 slowest tests (at least 2 minutes):
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Changes
Pass 9 now says how to check a test: run it (also under another Go toolchain), revert the fix for regression tests, restore scripts from
BASEfor variant-parity changes, readgh pr checksbefore asking about a real workspace, checkReplsfor host-dependent URL tokens, and skip the split ask when sections share expensive cloud setup.Why
Each check found a real test gap in a recent review that the skill did not prompt for.
Tests
Docs only.
This pull request and its description were written by Isaac.