Updating commit timestamp on rebase [was: Updating commit timestamp on push.]

[Augie Fackler]

On Fri, Dec 06, 2013 at 12:59:04AM +0000, Kannan Goundan wrote:
> Thanks everyone for the words of caution about my "update timestamp on push"
> idea :)
>
> I went with a different approach: I modified rebase to always use the
> current timestamp when rewriting a patch (instead of using the timestamp of
> the original commit).  This seems simpler and safer than my original idea,
> but still achieves the goal of having the commit timestamp roughly match the
> push timestamp (we rebase before pushing to the central repo).
>
> Unfortunately the implementation is hacky.  I wrote an extension that
> monkey-patches hgext.rebase.concludenode and discards the "date" parameter
> to repo.commit().  Roughly:
>
>    def patched_concludenode(repo, ...)
>        repo = RepoWrapper(repo)
>        orig_concludenode(repo, ...)
>
>    class RepoWrapper(object):
>        def __init__(self, repo):
>            self._original_repo = repo
>        def commit(self, ..., date=None, ...):
>            self._original_repo.commit(..., date=None, ...)
>        ...
>        # Forward all the other methods through unchanged.
>
> I'm a little worried about this breaking in weird corner cases or for new
> Mercurial releases.  Any ideas on how to do this more robustly?
>
> In general, I think having "rebase" update the commit date is useful.  If
> others feel the same way, maybe the rebase extension could be modified to
> support it directly, possibly through an option in hgrc?  Any objections to
> that idea?

I'm open to it. How do other people feel?

>
> _______________________________________________
> Mercurial mailing list
> Mercurial at selenic.com
> http://selenic.com/mailman/listinfo/mercurial