From: Michael J Gruber <git@drmicha.warpmail.net>
Date: Mon, 29 Oct 2012 10:08:50 GMT
Subject: Re: [PATCH] replace: parse revision argument for -d
Message-ID: <508E55B2.6060502@drmicha.warpmail.net>
In-Reply-To: <20121029090419.GA29464@sigill.intra.peff.net>

Jeff King venit, vidit, dixit 29.10.2012 10:04:
> On Mon, Oct 29, 2012 at 10:02:47AM +0100, Michael J Gruber wrote:
> 
>> Jeff King venit, vidit, dixit 29.10.2012 07:58:
>>> On Fri, Oct 26, 2012 at 03:33:27PM +0200, Michael J Gruber wrote:
>>>
>>>>  	for (p = argv; *p; p++) {
>>>> -		if (snprintf(ref, sizeof(ref), "refs/replace/%s", *p)
>>>> +		q = *p;
>>>> +		if (get_sha1(q, sha1))
>>>> +			warning("Failed to resolve '%s' as a valid ref; taking it literally.", q);
>>>> +		else
>>>> +			q = sha1_to_hex(sha1);
>>>
>>> Doesn't get_sha1 already handle this for 40-byte sha1s (and for anything
>>> else, it would not work anyway)?
>>
>> What is "this"???
>>
>> So far, "git replace -d <rev>" only accepts a full sha1, because it uses
>> it literally as a ref name "resf/replace/<rev>" without resolving anything.
>>
>> The patch makes it so that <rev> gets resolved to a sha1 even if it is
>> abbreviated, and then it gets used.
>>
>> Or do you mean the warning?
> 
> Sorry, yeah, I meant the warning and fallback.
> 
> If I understand correctly, the fallback will never work unless we are
> fed a 40-byte sha1. But get_sha1 should always return a 40-byte sha1
> without doing any further processing.

You do understand correctly, and I misunderstood get_sha1(). I does not
check for the presence of that sha1 in the object db, so that it
succeeds no matter what, that is: if it's valid hex. So I should
probably rewrite the error handling: no more need to check for lengths.

Michael