Re: [PATCH v4] Allow update hooks to update refs on their own.

From: Steven Grimm <koreth@midwinter.com>
Date: Dec 6, 2007, 07:50 UTC
Message-ID: <602F3FCF-C761-4E0A-9854-723F9F8F96F7@midwinter.com>
In-Reply-To: <20071206063626.GA18698@coredump.intra.peff.net>

On Dec 5, 2007, at 10:36 PM, Jeff King wrote:

> Ah, I thought his argument was "we have to send back a bit, so why not
> just send the hash we made for informational purposes? It doesn't  
> hurt,
> and maybe we can make use of it later."

Yeah, that was more or less my thinking. Keep it simple for now, but it seems like that information is bound to be useful at some point. In particular, if you don't send it down, it's really difficult to unambiguously get back after the fact (given that a fetch might contain subsequent revisions unrelated to yours.)

My v3 patch (which I will combine with a modified form of the documentation update now that it sounds like transmitting the SHA1 isn't objectionable) actually sent it down twice: once in the protocol message and once in the human-readable push status report.

-Steve