From: Junio C Hamano <gitster@pobox.com>
Date: Mon, 09 Jun 2008 23:41:06 GMT
Subject: Re: [PATCH] Introduce get_octopus_merge_bases() in commit.c
Message-ID: <7vlk1edvn1.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <alpine.DEB.1.00.0806100024310.1783@racer>

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> However, your remark about optimizing for the two-head case got me 
> thinking: should we not rather use the simple algorithm Miklos proposed 
> for octopus_merge_bases(), even if it is suboptimal?

Actually a quick glance at git-merge, a rather large case...esac after
that "show-branch --merge-base" tells me that we do not really use the
output from that operation and instead we check if we are fast-forward
from all the other heads by iterating over them.  merge-octupos would
accept it as the base but never looks at it.

So I think with a proper refactoring of the codepath you probably would
not even have to do the blanket "merge-base across all heads" at all.