From: Alex Riesen <raa.lkml@gmail.com>
Date: Thu, 29 May 2008 17:51:57 GMT
Subject: Re: [PATCH] make commit --interactive lock index
Message-ID: <20080529175157.GB5596@steel.home>
In-Reply-To: <483EC079.1050008@gnu.org>

Paolo Bonzini, Thu, May 29, 2008 16:40:57 +0200:
>>>>> +	assert (!(interactive && pathspec && *pathspec));
>>>> As pathspec is specified indirectly by the user, I think an 
>>>> assert() here is actively wrong.
>>> But the program may still guarantee a condition by checking it  
>>> elsewhere.  I don't need to teach you about that, do I?  In 
>>> particular, the assert checks that this:
>>>
>>> if (interactive && argc > 0)
>>>         die("Paths with --interactive does not make sense.");
>>>
>>> ... is equivalent to !pathspec || !*pathspec.
>>
>> Okay, I have to spell it out:
>>
>> I think that the assert() here is not helpful at all, and that you 
>> should rather do the "if () die()" thingie.
>
> The "if() die ()" thingie is already in builtin-commit.c, so we won't  
> ever get a pathspec in the "add --interactive" case.  If we do,  
> something else has already been done incorrectly before -- not by the  
> user but by the programmer.

What could that be?