On Thu, 19 Aug 2010, Josh Triplett wrote:
> On Thu, Aug 19, 2010 at 05:28:57PM -0500, Jonathan Nieder wrote:
> > (+cc: some format-patch --thread people for input)
> >
> > Andreas Metzler wrote:
> >
> > > It is very easy to shoot yourself into the respective foot when
> > > combining git format-patch with git send-email:
> > >
> > > ametz...@argenau:/tmp/WMAKER/wmaker-crm.git$ git format-patch -o
> > > /tmp/gittest -M --to=ametz...@downhill.at.eu.org --thread=shallow
> > > 8b6f37d4bb5245d5debcf3aad30e9d98cceb3db9
> > > /tmp/gittest/0001-actually-add-bug-number.patch
> > > /tmp/gittest/0002-SelectWindowsMouseButton-MouseLeftButtonAction.patch
> > > ametz...@argenau:/tmp/WMAKER/wmaker-crm.git$ git send-email
> > > --to=ametz...@downhill.at.eu.org --smtp-server=localhost
> > > --suppress-cc=self /tmp/gittest/
> > > [...]
> > >
> > > With these options git send-email sends out mails that have:
> > > * Two To: headers
> > > * Two In-Reply-To: headers
> > > * Two References: headers
> >
> > Yep, that's true. There is this odd tension between format-patch and
> > send-email options, since format-patch --to and so on came later and were
> > not really meant for use with send-email as far as I can tell.
> >
> > Probably "git send-email" should override and warn about existing headers
> > when it adds its own.
>
> Certainly the result of having duplicate headers seems wrong. Ideally,
> git send-email should augment existing addressing headers (avoiding
> duplicate addresses), and override conflicting message-id/threading
> headers.
>
> I do wonder, though, how often git send-email sends patches not
> generated by git format-patch. I'd guess not often. The reverse seems
> somewhat more likely. Ideally that would allow deduplicating this
> functionality, but any change to git send-email, even if phased in
> slowly, would probably break a lot of scripts.
I certainly always used format-patch to generate messages which I would
then send with my regular mailer (thereby getting into my regular
sent-mail folder &c). I'd guess that send-email, at least when sending
patches that have already been written out, ought to not add identical
headers and add its arguments to the existing headers instead of adding
new ones; I'd expect that existing scripts would be fine with this
behavior, since it only affects a combination of circumstances whose
current behavior is invalid.
-Daniel
*This .sig left intentionally blank*
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
