From: Wincent Colaiuta <win@wincent.com>
Date: Sat, 24 Nov 2007 21:50:10 GMT
Subject: Re: [PATCH 2/3] Move pathspec validation into interactive_add
Message-ID: <4D81F973-8951-458A-958D-E24C826CA548@wincent.com>
In-Reply-To: <7vir3rqwts.fsf@gitster.siamese.dyndns.org>

El 24/11/2007, a las 20:15, Junio C Hamano escribió:

> Wincent Colaiuta <win@wincent.com> writes:
>
>> Instead of throwing away the return status of pathspec_match() I am
>> keeping it, and any successful match breaks out of the loop early.
>
> Leaving it early before checking if all the given pathspecs are
> used defeats the whole "error-unmatch" business, doesn't it?

No, I probably didn't explain that clearly enough... If you look at  
the patch, I break out of the *inner* loop the first time a particular  
pattern matches. Then I move on to the next pattern, and any single  
invalid pattern will be enough to provide the "error-unmatch"  
indication.

>> Another issue is that simple shell shortcuts don't work, so something
>> as simple as "git-add -i ." will report:
>>
>>  error: pathspec '.' did not match any file(s) known to git.
>
> The sample code snippet I sent you probably is not doing
> get_pathspec() before using the "pattern" thing.  And I suspect
> that ...
>
>> Likewise it's not possible to validate pathspecs like "\*.sh" either,
>
> ... may be related to that.


Ok, thanks for the pointer. Didn't know what the get_pathspec()  
function was for, but will have a play with it.

Cheers,
Wincent