Re: [PATCH] git-tag: Document that -m/-F create tag objects, too

From: Julian Phillips <julian@quantumfyre.co.uk>
Date: Apr 9, 2007, 16:15 UTC
Message-ID: <Pine.LNX.4.64.0704091641290.1420@beast.quantumfyre.co.uk>
In-Reply-To: <87y7l11s5m.fsf@mid.deneb.enyo.de>

On Mon, 9 Apr 2007, Florian Weimer wrote:

Show 13 quoted lines

> * Julian Phillips:
>
>> On Mon, 9 Apr 2007, Florian Weimer wrote:
>>
>>> And warn about the fact that lightweight tags are not normally
>>> transferred by clone/fetch.
>>
>> Um ... yes, they are (in v1.5.1 at least).
>
> Hmm.  This has to be a fairly recent change, see:
>
> <http://article.gmane.org/gmane.linux.kernel/513438>
>

It's not that recent a change. I have been cloning and fetching my tag-references for some time (being too lazy to use full tags for my one-man projects). Certainly since v1.4.4 days ...

Fetch doesn't care whether a tag is a full tag or merely a reference, it will automatically fetch them both in the same way (not sure about clone). It checks that the reference points at an object, but not that the object is a tag rather than a commit. This behaviour hasn't changed since the auto tag-following code was added (AFAIK). Any difference between the treatment of tags and tag-references is a subtle consequence of the former adding an object I expect.

In the case of the v2.6.21-rc6 tag-reference anyone who did a clone or fetch and may have got the tag-reference provided that they didn't already have the commit it referenced? I'm not 100% sure of the behaviour here, I haven't quite got to tag following in my C rewrite (I should be doing it later today).

-- 
Julian

  ---
Never call a man a fool.  Borrow from him.