Repository navigation
R7 Q7: a per-leaf question on a whole-repository review - #98
Merged
Merged
Conversation
R7 Q7 asks for a mutation run, or a hand check, on each leaf. Under a whole-repository hunter's budget that never fits, so every --all run ended with "0 of 50 hits judged" and nothing else. On a whole-repository review the question is now not judged: its receipt reads "per-leaf, not judged", R7's tally takes the partial form, and the not-reached line names the three leaf directories with the most hits as the scoped reviews that judge them. On a scoped review the question is judged in full, as before. The lead skips what the Detection already says to skip: .tsx components, components/, hooks/ and context/ in ts-react; cmd/, main and handlers/ in Go and Python. On the fixtures that moves the count from 50 to 44 (ts-react), 35 to 33 (py) and 36 to 36 (go); every Q7 plant keeps its hits. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner
Author
|
Eval case: buzzdan/ldd-evals#47 |
In case G, no Python or ts-react report found the tenant mutant. The tests match only the word "tenant", which every tenant error shares. In ts-react, a removed empty check survives: the empty input reaches the character check, and its error also contains "tenant". In Python, a changed message survives. The hand check asked for rows at comparisons only, so it did not look at the assertions. The hand check now opens the leaf's test file, also when the file is outside the review's scope, and reads each error assertion. An assertion on part of the message is a violation. Go and generic do not change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Python R7 hunter judged Q7 one hit at a time. The only hit in tenant.py is the length comparison, and the table has rows for it, so the hunter cleared the file and did not read its error assertion. The hand check now reads the error assertions of each leaf file in scope, also where no hit row points. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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
1 participant
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.
What
R7 Q7: <hits> hit(s) → per-leaf, not judged, R7's tally takes the partial form, and thenot reached:line names the three leaf directories with the most hits as/<prefix>-review <dir>. On a scoped review the question is judged in full, as before.Detect-grepskips the files its own Detection says to skip. In ts-react that is.tsxfiles,components/,hooks/andcontext/; in Go and Python it iscmd/,mainandhandlers/.toThrow('tenant'),match="tenant") is a violation. A removed check or a changed message survives such an assertion. Go and generic do not change.Why
Every whole-repository run ended with
R7 Q7: 0 of 50 hits judged. A mutation run or a hand check over every leaf never fits a hunter's budget, so the question produced nothing. Both plugins also missed the three R7.Q7 plants on substance in the A/B, even though their basename graders passed.Measured on the fixtures (detection pass only, no model)
The narrower lead helps little. The main change is the routing.
Checks
task check,task lint-core,task docs:check: green.task test-gate: 2 macOS-only failures, fixed in test-gate: the repo-brain tests run on macOS #97.claude-sonnet-5, $28.08 in total over three rounds):tenant.pyon its length rows. 2/2 after the item "once per leaf file, not only where a hit row points".review-fullfor the new receipt and thenot reached:line.🤖 Generated with Claude Code