From: Jon Seymour <jon.seymour@gmail.com>
Date: Thu, 30 Jun 2005 01:33:25 GMT
Subject: Re: [PATCH 2/2] Fix for git-rev-list --merge-order B ^A (A,B share common base)
Message-ID: <2cfc403205062918336a55e8da@mail.gmail.com>
In-Reply-To: <7v1x6k1z6c.fsf@assigned-by-dhcp.cox.net>

On 6/30/05, Junio C Hamano <junkio@cox.net> wrote:
> >>>>> "JS" == Jon Seymour <jon.seymour@gmail.com> writes:
> 
> I am puzzled about this part.
> 
> JS> The unit test changes in this patch remove use of the --show-breaks
> JS> flags from certain unit tests. The changed --merge-order behaviour
> JS> changed the annotation that --show-breaks prints for certain test cases.
> JS> The new behaviour is reasonable and irrelevant to the intent of the tests
> JS> so that tests have been changed to eliminate the spurious behaviour.
> 
> If the behaviour of --show-breaks subtly changes, and if that
> changed behaviour is something still acceptable, why not update
> the test to show the new expected results since you are updating
> the test anyway?

I can do it this way, if you prefer. The issue was that the expected output was:

= l2
| l1 
| l0

but became:

^ l2
| l1
| l0

The annotation changes because there is no longer a single head in the
start list. There are now multiple heads, it just happens that one of
the heads is also a prune point.

> 
> Showing that "subtle" change in the diff may draw people's
> attention and would help you to verify that the behaviour change
> is not something that would be unacceptable to them.

Fair enough, I'll resubmit with a less drastic change to the test case.

> 
> Also if you are changing t6001, could you also merge Mark
> Allen's BSD portability fix while you are at it?
> 
>     Message-ID: <20050628014337.18986.qmail@web41205.mail.yahoo.com>
> 
> 

Ok.

jon.