sign-offs hung up on old user account

[bndw]

It seems that there may be a situation in which this bot uses the user/email from a previously replaced commit. This seems to be happening over on envoyproxy/envoy#1743

Here's how we got in this state:

1. A commit was made and signed with user a
2. The commit was squashed and resigned with user b
3. The bot is still looking for user a, even though the commit is now signed with user b.

[hiimbex]

Hiya! Just commented in envoyproxy/envoy#1743 (comment) but basically, as long as the git config is set to the correct email, say email a. As long as the commit is signed with email a, the bot should correctly approve the commit.

[bndw]

Hey, thanks for the quick response!

git config --global -l
user.name=bndw
user.email=benjamindwoodward@gmail.com

I just rebased, signed with that account, and force pushed envoyproxy/envoy@75c2374 to envoyproxy/envoy#1768, but it's still reporting:

DCO — Expected "bndw <benwoodward@example.com>", but got "bndw <benjamindwoodward@gmail.com>"

Anything else to try?

[hiimbex]

Ah, you're completely right, I was able to recreate this here: probot/test#24. This is definitely an issue on our end.

This is not behavior I expect, but it appears to be due to our use of the GitHub API here: https://developer.github.com/v3/repos/commits/#compare-two-commits

The API call we're using has this data:

"commit": {
  "author": {
        "name": "Monalisa Octocat",
        "email": "support@github.com",
        "date": "2011-04-14T16:00:49Z"
      },
  "committer": {
        "name": "Monalisa Octocat",
        "email": "support@github.com",
        "date": "2011-04-14T16:00:49Z"
      }
 }

We do our checks using commit.author.email:
https://github.com/probot/dco/blob/3e486021f8333df5c957e9b12c53bca1d9c7ba5f/lib/dco.js#L30-L35

When you rebase, the committer email is probably the one that gets updated, and I think it should be the one we use. So, I think on our end we should swap to commit.comitter.email. Thoughts @bkeepers ?

[bndw]

Nice find! Another test potentially worth running would be to specify the author on the rebased commit and see if that updates the check?

But using commit.comitter.email makes sense to me.

[hiimbex]

Another test potentially worth running would be to specify the author on the rebased commit and see if that updates the check?

It would be my suspicion that that would work! Not super familiar with all the rebase options, so I didn't even realize you could update the author 🙇 let me know if that works!

[bndw]

Just tried this but it didn't appear to update the check. In my commit message, I specified the following but it's still using the old author:

My commit message

Signed-off-by: bndw <benjamindwoodward@gmail.com>
Author:    bndw <benjamindwoodward@gmail.com>

[hiimbex]

Hmm I believe it's because you need to use git commit --amend --author="Author Name <email@address.com>" after an interactive rebase (not my strong suit lol). But this worked here: probot/test#24 for me. I did git rebase -i my_commit and then added x git commit --amend --author="Author Name <email@address.com>" at the top and saved and force pushed.

[bndw]

Excellent, this workaround should unblock the linked PR!

Just to be clear, the steps to change the author were:

1. interactive rebase
2. In the commit message, add this line to the top, above the message? x git commit --amend --author="Author Name <email@address.com>"

[hiimbex]

Yes! To clarify, the x in interactive rebased represents exec, so I did this through vim (not in the commit message itself) and deleted the first line which was noop and replaced it with x git commit --amend --author="Author Name <email@address.com>" and upon saving that and then force pushing, that updated the author for me.

For the purposes of the future, I'm somewhat unsure of whether the DCO should be based on the author or the committer, so I'd be open to input from some of the linux folks who use the bot on what they think.

[bndw]

Ok that makes sense, I was able work around this by amending the author on my commit.

I'll leave this issue open since it sounds like there's some discussion to take place. Thanks again @hiimbex 😄

[bkeepers]

So, I think on our end we should swap to commit.comitter.email. Thoughts @bkeepers ?

👍 Great catch! Sounds good to me.

[stale]

Is this still relevant? If so, please comment with any updates or addition details.

[bndw]

Looks like people are still hitting this issue, but I'm not able to allocate anymore time to it.

[github-actions]

🎉 This issue has been resolved in version 1.0.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀