From: Frans Klaver <fransklaver@gmail.com>
Date: Fri, 30 Mar 2012 20:22:39 GMT
Subject: Re: [PATCH] run-command: treat inaccessible directories as ENOENT
Message-ID: <op.wbz2v2k60aolir@keputer>
In-Reply-To: <7vmx6ym54m.fsf@alter.siamese.dyndns.org>

On Fri, 30 Mar 2012 18:21:29 +0200, Junio C Hamano <gitster@pobox.com>  
wrote:

> Frans Klaver <fransklaver@gmail.com> writes:
>
>> isn't it rather more sensible from a unit-test perspective to see
>> if start_command returns 127 instead of 128 in this specific case?
>
>
> You are welcome to add another test that checks lower level  
> implementation
> detail, but this specific test is to make sure the gripe "Why does git
> deny my aliases when I have inaccessible directory on my PATH?" will  
> never
> come back.

I think I didn't word carefully enough there. I didn't mean to dispute the  
use of the test. The test I proposed would make sense in t0061, but I  
would rather have expected the test in Jeff's patch in a tests that  
specifically targets aliases. It would be less surprising, wouldn't it?  
The fact that git goes through start_command before doing aliases is  
merely an implementation detail, from my point of view.