From: Nguyen Thai Ngoc Duy <pclouds@gmail.com>
Date: Thu, 24 May 2012 13:12:30 GMT
Subject: Re: [PATCH v2 6/6] sha1_loose_object_info: do not complain out loud on non-existent objects
Message-ID: <CACsJy8CpqR4WGoYqNpUESigqTgHr0cL4gK2t7sCcb8P5StX3-w@mail.gmail.com>
In-Reply-To: <7vvcjm28ff.fsf@alter.siamese.dyndns.org>

On Wed, May 23, 2012 at 11:01 PM, Junio C Hamano <gitster@pobox.com> wrote:
>> while it's still making sense for me, i think it's more logical  to
>> move the check to the caller, where "entry in pack?" check is also
>> done.
>
> I think most of the callers of sha1_object_info_extended() are using this
> function, saying "We expect this object to exist somewhere, perhaps in
> pack or perhaps in a loose form, and trying to see what it is", and they
> rely on the first error message "unable to find" to be issued.

hmm.. if you see it from that angle, yes it makes sense

> So in that sense, I do not see how this patch makes any sense at all.
> Care to point out a codepath where we throw a random 20 bytes at it in
> order to see if an object with the given object name exists?  That would
> be the only case where "unable to find" might be an unwanted error
> message.

packed_delta_info(), fast-import (I think) and cat-file do not check
for object existence before calling this.
-- 
Duy