Re: [PATCH] Additional merge-base tests

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Jul 5, 2006, 17:04 UTC
Message-ID: <Pine.LNX.4.63.0607051900200.29667@wbgn013.biozentrum.uni-wuerzburg.de>
In-Reply-To: <44ABE596.40103@gmail.com>

Hi,

On Wed, 5 Jul 2006, A Large Angry SCM wrote:

> If grafts in your repository create a cycle, the misbehavior of merge-base
> should be among the least of your concerns.

Right. BUT grafts can be very helpful to connect branches which were independently imported into git. And in these cases, the clockSkew really is no clockSkew. But in that case, the generation number would have to be recalculated also.

Anyway, I think that it should be a configurable, which defaults to off, i.e. in the normal case merge-base should behave as it does now. And if we have that configurable, we might as well take the safe but dumb approach to just traverse _everything_.

Ciao, Dscho