Re: [PATCH 4/3] git-fetch: test avoiding unnecessary copying from alternates

From: Shawn O. Pearce <spearce@spearce.org>
Date: Nov 8, 2007, 08:34 UTC
Message-ID: <20071108083440.GK14735@spearce.org>
In-Reply-To: <20071108082213.GA17054@spearce.org>

"Shawn O. Pearce" <spearce@spearce.org> wrote:

> This test verifies my prior "avoid local fetching from alternate"
> patch is functional and doesn't regress in the future during any
> additional improvements made to git.

...

Show 9 quoted lines

> +test_expect_success 'quickfetch should not copy from alternate' '
> +
> +	(
> +		mkdir quickclone &&
> +		cd quickclone &&
> +		git init-db &&
> +		(cd ../.git/objects && pwd) >.git/objects/info/alternates &&
> +		git remote add origin .. &&
> +		git fetch -k -k

Hmmph. On second thought I think this is a little sketchy for a test. Versions without my quickfetch patch fail this test and versions with it pass. But it depends on the implementation of `-k -k` to always call index-pack over unpack-objects.

I'm using -k -k here to ensure we keep the pack fetched as we're only fetching 6 objects and they are already reachable in the quickclone repository thanks to the alternate ODB. If we use unpack-objects during this fetch we will still pass this test because the objects won't be unpacked if they are already reachable locally.

Of course this test is for a performance optimization. For 6 tiny objects it really doesn't matter if we copy them or not, or if we copy them over a pipe only to discard them because they are already reachable. It does however matter when you are talking about nearly 300MB worth of objects. :-\

-- 
Shawn.