From: Junio C Hamano <gitster@pobox.com>
Date: Sun, 23 Aug 2009 02:09:57 GMT
Subject: Re: [RFC] gitweb.perl t9500 t9501
Message-ID: <7v3a7j2scq.fsf@alter.siamese.dyndns.org>
In-Reply-To: <4C812C4E-815B-478D-AEB1-3F1DEC5D3641@mailservices.uwaterloo.ca>

Mark A Rada <marada@uwaterloo.ca> writes:

> I have not been telling git-format-patch to use a specific width for
> the diffstat, is the standard width supposed to be 70?

Perhaps there is some misconception here.

We try to keep our _code_ to fit in standard 80-column terminal.  We also
keep lines in the _log message_ shorter than 70-something so that after
indentation by various programs (e.g. shortlog) and a few levels of
quoting in the e-mail are added, the lines will still stay below
80-columns.  These are standard practices.

And you may need to find a good place to break lines in your _code_, and
your _log message_, to stay below that limit.

But that all happens inside your editor, before you touch format-patch nor
your MUA.

The patch contents may contain long lines, such as

    diff --git a/filename-that-is-long b/filename-that-is-long

or

    @@ -old,lines +new,lines @@ original contents that is long

that is neither _code_ nor _log message_.  They could easily go beyond
80-column, but that is normal, and they do not have any 70-column limit
for you to worry about.  They follow a different rule, namely, if
format-patch (or "diff" in general) outputs something on a single line,
that has to stay on a single line.  "patch" expects each of these on a
single line.  Do not break them.

So make sure that you do not let your MUA perform any line-wrapping at
random place in the message after you formatted your commits via
format-patch, and you will be Ok.