[PATCH] patchbomb: make --git imply --plain

[Henning Schild]

Am Thu, 17 Nov 2016 16:39:05 -0500
schrieb Augie Fackler <raf at durin42.com>:

> On Thu, Nov 17, 2016 at 4:34 PM, Sean Farley <sean at farley.io> wrote:
> > Henning Schild <henning at hennsch.de> writes:
> >  
> >> Am Wed, 16 Nov 2016 17:41:03 -0500
> >> schrieb Augie Fackler <raf at durin42.com>:
> >>  
> >>> > On Nov 16, 2016, at 15:41, Henning Schild <henning at hennsch.de>
> >>> > wrote:
> >>> >
> >>> > # HG changeset patch
> >>> > # User Henning Schild <henning at hennsch.de>
> >>> > # Date 1479327376 -3600
> >>> > #      Wed Nov 16 21:16:16 2016 +0100
> >>> > # Node ID 181aaf02c1542387651e66c22063bb276271ced2
> >>> > # Parent  db8637a8a7464aca668e5aaa2fb9610edaf6f286
> >>> > patchbomb: make --git imply --plain  
> >>>
> >>> No thanks. --git here is about using the git-diff format, which is
> >>> still extremely valuable for hg users.  
> >>
> >> Ok. I guess that consequently also means no to the diffstat patch
> >> (2 of 2 out of this thread)?
> >>
> >> Let me see how to fit that change into hg-git. I did not have a
> >> look at how that hooks into hg, but i can imagine a similar patch
> >> would be much harder there - like parsing the mbox. I need that
> >> change working with hg-git for contributing code back to upstream
> >> mailing lists. I can imagine that similar patches would be
> >> required for the reverse operation (hg mimport), but i did not try
> >> that way yet.  
> >
> > Since a git mailing list would be only for git clients, it'd be ok
> > to have both import / export in the git style for hg-git. Maybe
> > others feel different than me, though.  
> 
> It's probably Not Okay to have hg-git break hg email for users that
> have hg-git turned on globally, rather than per-repo?
> 
> >  
> >> How about a new option --git-format that implies --git and --plain
> >> and takes care of the diffstat?  
> >
> > We usually shy away from new arguments and instead suggest the
> > improvement of templates which could be used here (maybe they
> > already can, I haven't checked).  
> 
> I'm not sure that applies entirely to email tho - the email formats
> aren't templated, though I suppose they could be...

The whole template-story sounds like a good idea but kind of long-term.
For the last series i have sent the question is whether the new
argument would contradict such a change (if hg-git brought its own
template and the new argument would not be required anymore).

Today you can already send valid mails to git-based projects. Including
the diffstat is just a style thing, if you skip it git-am wont care.
Setting --plain and --git automatically is pure convenience.

That being said, sending emails without diffstat - and maybe other
additional information - might result in git upstream to "dislike" the
mails and rejecting the contributions.

While this is not an hg issue it could mean that hg-git is not a
viable option if you are planning to contribute back to upstream git.

> Now you've got me thinking.
>