RFC: patch size analysis prompt
Ian Rogers <[email protected]> Wed, 10 Jun 2026 15:40:27 -0700
| Newsgroups | dev.linux.lists.sashiko |
|---|---|
| Message-ID | <CAP-5=fWmn9+LT-4VeAm6JwTM=FB071w-Oa2QXDROLK7VDXcfow@mail.gmail.com> |
Hi, I sent an RFC PR to add patch size analysis to Chris Mason's prompts: https://github.com/masoncl/review-prompts/pull/73 The proposed prompts aren't subsystem specific but maybe they should be? The prompt itself: 1) Shows what a decomposition analysis for a patch should look like. 2) Categorizes the patch size, ignoring small ones. 3) Recommends decomposition when multiple subsystems are involved, or when bug fixes, features and refactorings are mixed together. 4) Establishes an 80% rule: if any of the decomposed patches is sized at 80% of the original patch's size, then the original patch was appropriately sized. Using the prompt locally I find the feedback useful and the approach doesn't feel specific to the perf subsystem I largely work in. I could imagine that if someone is say dropping a new device driver, getting feedback to break that apart may be unwelcome. Perhaps people have extra advice on how to determine if a patch is appropriately sized. I think the feedback would be useful for both reviewers and contributors. For reviewers it would let them know if asking for decomposition was reasonable. For contributors it would give them quick feedback that their code should be broken apart to make it easier to review. However, these are just my thoughts and the RFC and this email is intended for others to provide theirs. Thanks, Ian