threads / discuss / 13881

'git format-patch' on older commits, and mailinglist

Subject: 'git format-patch' on older commits, and mailinglist

## tl;dr

2 messages between Jun 10, 2008 and Jun 10, 2008.

replies: 1people: 2as markdown or json

Marius Storm-Olsen· Jun 10, 2008, 10:09 UTC · lore

Hi,

Recently, I sent out patches which were fixed up with --amend on older 
commits. When I sent them out, the patches contained
     Date: <more than 2 days old datestamp>
in the headers. Now, sending these mails is fine, and mail clients 
generally handles it perfectly fine. However, after doing this I got 
an email from postmaster@vger.kernel.org, basically telling me to not 
do this, since they get a lot of bounces where the return is marked with
     Diagnostic Code: smtp; 550 (4.5 DATE_IN_PAST_48_96 Date: is 48 to 
96 hours before Received: date)

This is understandable. The question is, do we fix the tools to handle this, so that emails are always generated with now() date, and the commit content contains a tag for the original commit; or do we simply say, always send patches to the mailing list with a current timestamp?

Maybe my workflow is incorrect too. I don't mind pointers on this.

I guess rebasing before generating the patch series would have fixed this, but I really didn't need to. I simply reset HEAD~2, fixed with --amend, then cherry-picked the other on top again; then created the patch series.

-- 
.marius
- simply wondering what others on the git mailinglist do..

Junio C Hamano· Jun 10, 2008, 22:37 UTC · re: Marius Storm-Olsen · lore

Re: 'git format-patch' on older commits, and mailinglist

Marius Storm-Olsen <marius@trolltech.com> writes:

Show 15 quoted lines

> Recently, I sent out patches which were fixed up with --amend on older
> commits. When I sent them out, the patches contained
>     Date: <more than 2 days old datestamp>
> in the headers. Now, sending these mails is fine, and mail clients
> generally handles it perfectly fine. However, after doing this I got
> an email from postmaster@vger.kernel.org, basically telling me to not
> do this, since they get a lot of bounces where the return is marked
> with
>     Diagnostic Code: smtp; 550 (4.5 DATE_IN_PAST_48_96 Date: is 48 to
> 96 hours before Received: date)
>
> This is understandable. The question is, do we fix the tools to handle
> this, so that emails are always generated with now() date, and the
> commit content contains a tag for the original commit; or do we simply
> say, always send patches to the mailing list with a current timestamp?

I think the right solution would be for send-email to move Date: to the beginning of the body part, just like it adds an extra From: there when the sender is different from the author, and not lie about the mail transmission date.

Then we won't lose the authorship date information in the resulting commit and will still keep the mailpath happy.