Guard the coverage map's blocked reasons too, and record the durable blocker - #20
Merged
Conversation
…blocker The guard added for UNPROBEABLE has a twin here, because the same thing had already happened in the coverage map. analysis.defects.corrections.kumagai said pip could not place pydefect because scikit-image had no wheel for this glibc. scikit-image arrived as somebody else's dependency, that clause stopped being true, and nothing noticed - the install is still blocked, but by a pyarrow version reached through vise and mp-api, which is a different fact. A reason that records the wrong cause is worse than one that records none: it is the reason somebody will act on when they try to unblock it. Writing the fix taught the more useful half. My first correction named pyarrow, and py3.10 promptly failed the new guard because pyarrow is installed there. The install situation is not stable enough to record at all - it has moved twice and differs between interpreters on this one machine. The reason now leads with what does not move, which is the data: the correction maps each defect site onto a perfect one and reads a site potential for both, and those come from the OUTCARs of two real runs. The install is described as a nuisance that varies rather than asserted as a fact. The check needed the same lesson. Its first version searched a whole reason for any phrase asserting absence, and flagged a correct reason that named one package as present and another as missing. It now looks only within ninety characters of the package it is checking. 973 tests pass on py3.12, 853 on py3.10 — both, which is the point, since the first version of this passed on one and failed on the other. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016wh5NzNE9Wn49nuUm7zg2a
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.
The guard added for
UNPROBEABLEin #19 has a twin here, because the same thing had already happened in the coverage map.analysis.defects.corrections.kumagaisaid pip could not place pydefect because scikit-image had no wheel for this glibc. scikit-image later arrived as somebody else's dependency, that clause stopped being true, and nothing noticed. The install is still blocked — but by a pyarrow version reached through vise and mp-api, which is a different fact.A reason that records the wrong cause is worse than one that records none: it is the reason somebody will act on when they try to unblock it.
Writing the fix taught the more useful half
My first correction named pyarrow. py3.10 promptly failed the new guard, because pyarrow is installed there.
The install situation is not stable enough to record at all — it has moved twice and differs between interpreters on this one machine. So the reason now leads with what does not move:
The install is described as a nuisance that varies, rather than asserted as a fact.
The check needed the same lesson
Its first version searched a whole reason for any phrase asserting absence, and flagged a correct reason that named one package as present and another as missing. It now looks only within ninety characters of the package it is checking.
Verification
973 pass on py3.12, 853 on py3.10 — both, which is the point, since the first version of this passed on one and failed on the other. Verified the guard fails by writing a stale reason deliberately and reverting.
🤖 Generated with Claude Code
https://claude.ai/code/session_016wh5NzNE9Wn49nuUm7zg2a