Repository navigation
fix(hipdnn): close hipdnn-deep-review silent-drop and drafting gaps - #14566
Merged
BrianHarrisonAMD merged 4 commits intoOct 28, 2026
Merged
Conversation
…ALMIOPEN-2945) Fixes found running the skill on #14457: - ledger.py ids fails on marked-up ID: lines (bullet, bold, quote, inline code), not only on partial fences, and on provisional IDs that are not <LANE>-p<n> (follow-ups wrote RUN-f1-p1, which mapped to RUN-f-1). - A markdown heading or --- ends a finding field, so lane notes no longer bleed into Fix and from there into COMMENTS.md. - ledger.py coverage prints why an owner lane's row missed (short path, grouped or globbed row) instead of leaving files pending silently. - ledger.py cluster supersedes duplicates, repoints chains and refuses a superseded keep; it replaces hand edits of ID-MAP.md. - partition.py orient-gate replaces the grep pipeline and accepts a cited file the change adds. - comments.py drafts one General item per SUGGESTION/FUTURE WORK finding. - Docs: Scope: input for deprecated areas; follow-up ID numbering; verify test-gaps from the link graph when no test binary compiles the file.
✅ All Checks Passed — Ready for Review
📖 Need help? See the Policy FAQ for details on every check and how to fix failures. |
|
🎉 All checks passed! This PR is ready for review. |
- ids fails on ID: lines in numbered, plus, heading, table, italic or spaced form, and on an own-lane <LANE>-p<n> cited in Findings that no report of the round parsed. The ID-line pattern is linear time. - A single-# line (shell comment) no longer ends a field; ## headings and --- still do. findings.md says to fence commands and code in fields. - orient-gate strips a leading ./ from cited paths. - RUNBOOK and findings.md wording: one General item per lower-tier finding; VER<batch> provisional prefix.
…OPEN-2945) - ids also counts Tier: lines, so a finding block with no ID: line fails instead of vanishing or overwriting the previous block. - The cited-ID check covers every report prefix, not only the report's own lane, so a merged agent's other-lane findings and table-row findings are caught. A cited ID passes when any report of the round parses it. - The report-lane prefix list lives in partition.py and comments.py reuses it.
BrianHarrisonAMD
marked this pull request as ready for review
October 28, 2026 18:20
adickin-amd
approved these changes
Oct 28, 2026
adickin-amd
left a comment
adickin-amd
left a comment
Contributor
There was a problem hiding this comment.
Ran this against my latest PR and it worked as expected
This was referenced Oct 29, 2026
Contributor
TheRock Submodule Bump ActivityNewest first
|
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.
Summary
Fixes gaps in the
hipdnn-deep-reviewskill that a full review round of #14457 exposed. In that round, findings and coverage rows were dropped without any error, follow-up finding IDs were numbered wrongly, and the drafted comments needed a hand rewrite. The scripts now fail or explain in each of those cases.JIRA ID : ALMIOPEN-2945
Risk Assessment
Risk 1. Changes are limited to the skill's scripts, docs and tests under
projects/hipdnn/tools/ai/. No product code, build or packaging change.Testing Summary
validate-skills.pyvalidates all 16 skills.seed,ids,cluster,coverage,draftandorient-gateon a scratch repository.Testing Checklist
python3 -m pytest projects/hipdnn/tools/ai/tests- Status: Passed (334 passed, 7 skipped)python3 projects/hipdnn/tools/ai/validate-skills.py- Status: Passed (16 skills)Technical Changes
ledger.py idsfails when a report holds a finding block it cannot read: anID:line in bulleted, numbered, bold, quoted, heading or table form, a partly fenced block, or a block with noID:line. It also fails when a Findings section cites a provisional ID that no report of the round parsed, which catches findings written as table rows, including a merged agent's other lanes. Lane notes after a##heading or---no longer leak into the last field of a finding.<LANE>-p<n>. Follow-up reports had writtenRUN-f1-p1, which became the final IDRUN-f-1.ledger.py coverageexplains why an owner lane's row missed its file (short path, grouped or glob row) instead of leaving the filependingwith no reason.ledger.py cluster <kept> <duplicate>...replaces hand edits ofID-MAP.md. It keeps chains of superseded findings valid.partition.py orient-gatereplaces the shell pipeline. It accepts a cited file the change adds.Scope:input records notes such as deprecated areas. Verifiers no longer treat a mutant in code the test binary does not compile as proof of a test gap.