How is the Git AuthorDate determined when pushing changes with jj git push?

[stanleyj-edsn]

Hi Jujutsu team and community,

I’m trying to understand how Jujutsu determines the Git AuthorDate when pushing changes to a Git repository. I’ve encountered a situation where the AuthorDate for my commits is set to a date two weeks ago, even for revisions that were created today (e.g., using jj split). I’m hoping you can help clarify the behavior and whether this is expected or if I might be missing something.

Example:

I worked on a change today and used jj split to create multiple revisions. When I pushed the changes to a Git repository using jj git push, I noticed that all commits in the pushed branch have an AuthorDate of two weeks ago (March 13, 2025). This is confusing for reviewers on GitLab, as it makes it appear as though I started working on this branch two weeks ago, which isn’t accurate.

Here’s an example of the relevant information:

jj evolog output for the revision in question (created today):

surysnyx [email] 2025-03-27 15:07:03 class-hierarchy* git_head() bf81fab7
surysnyx hidden [email] 2025-03-27 14:18:34 class-hierarchy@origin 5bc24578
surysnyx hidden [email] 2025-03-27 14:18:24 a974665d
surysnyx hidden [email] 2025-03-27 14:16:12 afd6767e
surysnyx hidden [email] 2025-03-27 13:58:52 b2a6c934
surysnyx hidden [email] 2025-03-27 13:55:10 be228d6a
surysnyx hidden [email] 2025-03-27 13:53:59 4a70abe1
surysnyx hidden [email] 2025-03-27 13:53:26 4c7fa17d
surysnyx hidden [email] 2025-03-27 13:50:40 50468f79
surysnyx hidden [email] 2025-03-27 13:50:40 e2623d46
surysnyx hidden [email] 2025-03-27 13:50:27 bd44a3f9
surysnyx hidden [email] 2025-03-27 13:50:27 46ff5d60
surysnyx hidden [email] 2025-03-27 13:50:08 757c09a0
surysnyx hidden [email] 2025-03-27 13:43:57 00715a16
surysnyx hidden [email] 2025-03-27 13:43:57 27454b6a
surysnyx hidden [email] 2025-03-27 13:43:57 ee0b8166
surysnyx hidden [email] 2025-03-27 13:42:29 d043767a
surysnyx hidden [email] 2025-03-27 13:41:59 950687e2 conflict
surysnyx hidden [email] 2025-03-27 13:41:09 3119c3ef
surysnyx hidden [email] 2025-03-27 13:41:09 916c9293
surysnyx hidden [email] 2025-03-27 13:38:54 b5020cb0
surysnyx hidden [email] 2025-03-27 13:38:54 8a8b2f01 conflict
surysnyx hidden [email] 2025-03-27 13:38:37 47cbd99a
surysnyx hidden [email] 2025-03-27 13:31:32 569f5ba4

From this, you can see that the revision surysnyx was created today.

Git commit details after pushing to the Git repository:

commit bf81fab70987fd90d910048923f6d48a2545f7f5 (HEAD, class-hierarchy)
Author: [email]
Date:   Thu Mar 13 07:43:10 2025 +0100

Notice that the AuthorDate is set to March 13, 2025—two weeks ago. This is true not only for this commit but for all commits in the pushed branch.

I would expect the AuthorDate in the Git commits is the first revision in the evolog for this change (e.g., 2025-03-27 13:31:32 in this example).

Why is the AuthorDate set to a date two weeks ago, even for revisions created today (using jj split)?

I haven’t filed this as an issue because I suspect there might be a good reason for this behavior, and I’d like to understand it better before assuming it’s a bug. Any insights or explanations would be greatly appreciated!

Thank you for your time and for building such an interesting tool!

[PhilipMetzger]

I worked on a change today and used jj split to create multiple revisions. When I pushed the changes to a Git repository using jj git push, I noticed that all commits in the pushed branch have an AuthorDate of two weeks ago (March 13, 2025). This is confusing for reviewers on GitLab, as it makes it appear as though I started working on this branch two weeks ago, which isn’t accurate.

When did you create the initial revision, so jj newed? Here's a relevant issue from some time ago #2000.

[jennings]

The author date is set when a revision is first created, then persisted as that revision evolves. You can reset the author date of a commit to the current time with jj describe --reset-author.

It works this way because revisions have two dates: the author date and the committer date.

With git, the committer date is generally updated on any modification. But, the author date is not. So, if you rebase or amend a commit, the committer date is updated but the author date is left alone.

This can be a little confusing with jj because you start work by creating a revision rather than end work by creating one. So, you might end up finalizing your work in a revision that was originally created weeks ago.

[stanleyj-edsn]

The author date is set when a revision is first created, then persisted as that revision evolves.

when I use jj split that creates a new revision, is that correct? But also here the author date is from 2 weeks ago. My expectation, without knowing about jj internal details is: jj split sets author date to moment (timestamp) of the split action and does not inherit the author date

[cenviity]

The jj evolog output (and jj in general) shows committer date by default. If you configure jj to show author date instead, you should find that it matches what Git shows. See here: https://jj-vcs.github.io/jj/latest/config/#commit-timestamp

[stanleyj-edsn]

The jj evolog output (and jj in general) shows committer date by default. If you configure jj to show author date instead, you should find that it matches what Git shows. See here: https://jj-vcs.github.io/jj/prerelease/config/#commit-timestamp

Thanks! I was inaccurate in my original post. I am already happy with how jj shows things, to me it is the most logical representation of timestamps etc. My question should have focused only on visualization in git itself after jj git push.