I forgot about this thread until @MaskRay pointed me at it again after I merged a new contributor’s PR (they didn’t have commit access). I remember now when I clicked the squash and merge button that it gave me a preview of the email details that would be included. I should have paid more attention to it and flagged it up, but I didn’t, so the commit landed with bad email addresses.
I think there’s nearly full consensus that we should be mandating that proper email addresses are included in the author information: @j2kun seems to be the sole exception from a quick skim, with, I acknowledge a legitimate concern, but I feel like the health of the organisation outweighs this concern. As a workaround (and I admit it’s not a great one), you could always disable the private address option temporarily whilst you land your changes and then re-enable it afterwards.
Adding a comment to PRs when a commit is added without a correct email address seems like a good idea to me (it seems to be having the desired effect for clang-format, in my experience, for example), but I’d also suggest a few other things:
- Make sure the Getting Started docs for LLVM explain that a real email address is required for all contributions (and point out how to make sure this happens in GitHub).
- Make a broader announcement to the LLVM community to advise people of this policy, and how reviewers can make sure it is enforced.
- Add a comment after the squash and merge has happened if the issue is detected again (to resolve the “email address was overwritten” issue).
- I think you can add something that GitHub will report back to a user who does a direct push with the “wrong” email address style. It won’t stop them but at least it might encourage them to fix things (especially if it happens when they push to their original branch for the PR). Obviously, this won’t detect fake emails though.