Skip to content

Evidence for #2959 and #2794: duplicated findings and inline comment accumulation on conforma/cli PR #3136 #3423

Description

@fullsend-ai-retro

What happened

The fullsend review agent posted the same [high] logic-error finding (unconditional caching of transient failures) as separate inline comments on Jun 16 (line 152) and Jun 26 (line 145) of sigstore.go. 7 total inline comments accumulated across 2 runs with substantive duplicates. The finding had been raised by qodo (Feb 28) and coderabbit (May 14) months earlier, and the author had explicitly acknowledged it as an intentional tradeoff. The maintainer eventually closed all threads.

What could go better

This matches the dedup gap in #2959 (findings re-raised across review iterations) and the inline accumulation pattern in #2794 (inline comments not deduplicated across runs). Implementing #2959 would have prevented the duplicate logic-error inline comments. Implementing #2794 would have reduced 7 inline comments to 4-5.

Proposed change

This is evidence for existing issues #2959 and #2794. No new proposal needed — the existing issues accurately describe the dedup mechanism that would have prevented this noise.

Validation criteria

After #2959 and #2794 are implemented, re-review runs should not post new inline comments for findings substantively identical to ones already posted by the review agent. On PRs with 2+ review runs, duplicated inline findings should appear in under 5% of cases.


Generated by retro agent from conforma/cli#3136

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions