From: Junio C Hamano <gitster@pobox.com>
Date: Tue, 04 May 2010 17:11:13 GMT
Subject: Re: [PATCH 2/2] fast-import: validate entire ident string
Message-ID: <7vk4rjmw26.fsf@alter.siamese.dyndns.org>
In-Reply-To: <20100426160247.GD7502@spearce.org>

"Shawn O. Pearce" <spearce@spearce.org> writes:

> Jonathan Nieder <jrnieder@gmail.com> wrote:
>> The author, committer, and tagger name and email should not include
>> any embedded <, >, or newline characters.  The format of the
>> identification string is
>> 
>>   ('author'|'committer'|'tagger') sp name sp < email > sp date
>> 
>> If an object has no name attached, then git expects to find two spaces
>> in a row.
>
> This is going to be a problem I think.  Some importers are probably
> writing "committer <bob> ...." when pulling from systems that don't
> have a concept of name vs. email (e.g. CVS or SVN).  I highly suspect
> that requiring two spaces here will cause a lot of importers to fail.
>
> If we really need to require two spaces, I think we need to honor
> the documented input format but rewrite the line inside of the
> import process to match the two space convention.

It probably should document [sp name] (or [name sp]) an "zero or one"
item, if we want to be lenient.

I also think it may not be such a bad idea to allow fast-import to add a
phoney "name" by taking everything in e-mail before the first '@', just
like how the git-cvsimport and git-svn does, when the input stream does
not have a "name".  I am not sure how usable "shortlog" output would be
otherwise, without any "name" field.