bugbug

bugbug copied to clipboard

> Thought: we might want to mix comments written by developers and comments written by the bot (and accepted by developers). Perhaps it could be a number we can configure...

suhaibmujahid

Currently the script retrieves latest revisions from Phabricator and runs two versions of the code review tool. We should add ability to retrieve latest analyzed revisions from our DB of...

marco-c

On https://phabricator.services.mozilla.com/D224761, the tool generated a comment asking to check if `waitForPdfJSCanvas` was defined, but this function is easily found on Searchfox. We should debug this case and check questions...

marco-c

Should be better than "asking" the LLM

marco-c

Currently the review suggestion can be requested only once per each revision. This works if the update on the revision is minor. But if the revision is significantly modified (e.g....

arai-a

The review helper can sometimes add similar suggestions to multiple places (say, 10 or 20 places). It would be nice to allow dismissing similar suggestions at once when these suggestions...

arai-a

[code_review] The process frequantly stuck on `r.html.find("#file")`

1

https://github.com/mozilla/bugbug/blob/7b29581e03fa630243bc984e26c89017f23647cb/bugbug/code_search/searchfox_api.py#L44

suhaibmujahid

[code_review] Add example of unwanted "Ensure ..." comments to the filtering phase

2

We have noticed this type of comment is either ignored or marked as invalid by users. We should filter it out.

marco-c

See https://platform.openai.com/docs/guides/structured-outputs/

suhaibmujahid

This way we can easily share the evaluator script with developers and ask their feedback about two different approaches.

marco-c