To head off some patterns we've seen recently, it may be worth adding some additional instructions to CONTRIBUTING.md. Some topics to write about:
-
Using the CI pipeline to test code rather than testing locally. Some people submit PRs with untested code with the expectation that the CI workflow will alert them to problems, and they then iterate on changes repeatedly instead of performing local testing. This is a waste of compute resources because the CI pipelines run more than is usually needed for testing during code development.
-
Using AIs to generate entire issue descriptions, pull request descriptions, or comments. It is one thing to use a tool to help (e.g., as a proof-reader, or a writing aid when one's native language is not English); it is another to just dump LLM output into the text box. The latter disrespects the human maintainers, usually wastes their time, and gives no indication that there is a human on the other side.
To head off some patterns we've seen recently, it may be worth adding some additional instructions to CONTRIBUTING.md. Some topics to write about:
Using the CI pipeline to test code rather than testing locally. Some people submit PRs with untested code with the expectation that the CI workflow will alert them to problems, and they then iterate on changes repeatedly instead of performing local testing. This is a waste of compute resources because the CI pipelines run more than is usually needed for testing during code development.
Using AIs to generate entire issue descriptions, pull request descriptions, or comments. It is one thing to use a tool to help (e.g., as a proof-reader, or a writing aid when one's native language is not English); it is another to just dump LLM output into the text box. The latter disrespects the human maintainers, usually wastes their time, and gives no indication that there is a human on the other side.