Re: Problems with format-patch UTF-8 and a missing second empty line

[Jeff King <peff@peff.net>]

On Thu, Sep 15, 2011 at 11:45:15AM +0200, Ingo Ruhnke wrote:

> Creating a patch of a commit including UTF-8 and no empty second line,
> like this:

I already responded about the bug with utf8-encoded subjects, but let me
address the second half of your mail, too:

> Here the newline between ABC\nABC gets stripped out and replaced with
> a space when transferring the commit with format-patch from one
> repository to another.

This is by design. Git commit messages are intended to have a
single-line subject, followed by a blank line, followed by more
elaboration. A multi-line subject is treated as a single line that has
been line-broken, and is subject to being reflowed onto a single line.
This is done to help with commits imported from other version control
systems which don't follow this pattern (the other option is truncating
the subject and putting the other lines into the "body", but that often
ends up quite unreadable).

If you really want to retain the newlines across "format-patch | am",
use the "-k" option of both to preserve the subject (I don't recall the
details, but I think you need a more recent version of git for
format-patch to correctly encode this, but "am" can be from any
version).

> Another small issue is that the filename of the patch will strip out
> any UTF-8 characters, Thus a commit message of "123Äöü456" will result
> in "0001-123-456.patch".

Yes, it's an attempt to strip out characters that some filesystems might
not support well. We could probably enable high-bit characters with a
config option (maybe even just using core.quotepath).

-Peff