From: Jeff King <peff@peff.net>
Date: Thu, 24 Feb 2011 07:15:35 GMT
Subject: Re: [PATCH 3/3] format-patch: rfc2047-encode newlines in headers
Message-ID: <20110224071534.GC16550@sigill.intra.peff.net>
In-Reply-To: <7vhbbu7792.fsf@alter.siamese.dyndns.org>

On Wed, Feb 23, 2011 at 01:47:53PM -0800, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> > These should generally never happen, as we already
> > concatenate multiples in subjects into a single line. But
> > let's be defensive, since not encoding them means we will
> > output malformed headers.
> 
> In this particular case, wouldn't it be more conservative and defensive to
> produce malformed headers so that the patch won't leave the
> originator?

No. If you go back to xzer's original mail, the malformed headers didn't
cause messages not to be sent. They just resulted in corrupted and
missing data on the receiver side. I don't think we can rely on any MUA
or MTA having a particular behavior for malformed mail. Some of them may
complain, but many won't.

> I have a suspicion that mailinfo would choke on the output of this
> one, even though I didn't try.

Actually, it does quite well. Without "-k", mailinfo turns it into a
single line, which is what I would expect. With "-k", the info file
contains:

  Author: Jeff King
  Email: peff@peff.net
  Subject: this is a long
  Subject: subject line with
  Subject: many lines in it
  Date: Wed, 23 Feb 2011 11:30:43 -0500

which "git am" turns back into the original multi-line subject.

So I think it's definitely the right thing to do. Not only does it avoid
us generating malformed mail, but because existing mailinfo handles it
sanely, it makes it easy to do a "preserve-newlines" patch on top (which
I'm still not sure is a great idea, but I can see the use in certain
circumstances).

-Peff