Re: git-mailinfo doesn't stop parsing at the end of the header

[Philip Hofstetter <phofstetter@sensational.ch>]

Hello,

On Wed, Nov 18, 2009 at 6:24 PM, Jeff King <peff@peff.net> wrote:

>  1. It creates a bad user experience. You are not unreasonable for
>     wanting to put some specific text in your commit message. Having
>     git come back and say "oops, I might get confused by this later"
>     just seems like an annoyance to the user.

agreed, though it's not that bad: when learning git, you will be
confronted with the fact that the commit message has a few things that
are special (well. it's doesn't break git, but the first line should
be < 56 chars in length for example).

Not being able to have From: lines in them that are not describing an
author would then just be one of them.

>  2. Mailinfo has to deal with data created by older versions of git. So
>     in your case, the rebase was a bomb waiting to go off. If we can
>     fix it so that an existing bomb doesn't go off, rather than not
>     creating the bomb in the first place, then we are better off.

This is a very good point. I didn't quite think about that.

>  4. Commit messages can come from other places than "git commit". What
>     should we do with a commit message like this that is imported from
>     SVN? Reject the import? Munge the message?

I would leave that to the tool that does the import. Probably it would
have to munge it. Yes.

I DO see though that implementing the check at commit time would lead
to problems popping up at other places.

> Of course all of that presupposes that we can correctly handle the
> existing data after the fact. Even with my patch, you still can't write
> "From: foo@example.com" as the first line of your commit body. But that

can't you? IMHO it would just attribute the commit to foo@example.com
which can be an equally bad, if not worse thing (I'm saying that
without the needed knowledge about git internals to really be sure, so
take this with a grain of salt)

I just have a bad feeling about trying out heuristics to see whether
thing thing after from: is an email address or not as email addresses
are notoriously hard to detect.

Typing a commit message and applying a patch from an email should be
separate things and should be handled separately. Currently they are
not and this is what's causing the problem in the first place.

Maybe that --strict thing is actually a good thing in the long run,
even though I don't quite like it either :-)

Interesting problem to have though.

Philip