Re: [RFC 2/3] am: semi working --cover-at-tip

From: Junio C Hamano <gitster@pobox.com>
Date: Nov 17, 2017, 01:54 UTC
Message-ID: <xmqqlgj5u0u6.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <4bbaaf33-3796-4aa2-6434-ab79182274f5@suse.de>

Nicolas Morey-Chaisemartin <NMoreyChaisemartin@suse.de> writes:

>> I thought about that.

Show 7 quoted lines

>> Is there a use case for cover after the last patch works and
>> removes the need to touch am_next (can be done out of the loop in
>> am_run).
>
> Do you have an opinion on that ? It has quite a big impact on how things are done !
> Single series only would mean a simple flush at the end.
> Multiple series makes things a whole lot complex.

I am not sure what you even mean. Are you wondering what "am" should do to a mbox with, say, these messages?

	1: [PATCH 0/2] Cover for series A
	2: [PATCH 1/2] patch 1 of series A
	3: [PATCH 2/2] patch 2 of series A
	4: [PATCH 0/3] Cover for series B
	5: [PATCH 1/3] patch 1 of series B
	6: [PATCH 2/3] patch 2 of series B
	7: [PATCH 2/3] patch 3 of series B

Running "am" on the whole thing and expecting covers to become the capping empty commit at the tip is crazy, I would think, for such a mbox, as there is no way to tell the command (after it processes 1, 2 and 3, to create commits out of 1, 2 and then an empty one out of 0 to finish one topic off) that it must create a new branch to store the next series, and building the second series on top of the capping empty commit at the tip of first series would not make any sense---the "tip empty commit" for the first series will no longer be at the "tip".

What I was alluding to is a different case, in which additional patches are sent as a follow-up later, ending up in a mbox like this:

	1: [PATCH 0/2] Cover for series A
	2: [PATCH 1/2] patch 1 of series A
	3: [PATCH 2/2] patch 2 of series A
	4: [PATCH 3/2] patch 3 of series A

Naturally, the cover letter may not list 3/2 in its short-log section, but the description for the overall goal and approach of the series in it should still be valid even with patch 3/2. The total number of messages mailsplit gives us would be 4, your subject parser would read "2" as the number of patches, which would make the number of messages for the series to be expected "3" (i.e. "2" plus cover), but "am" would want to create commits for patches 1, 2, and 3, and then cap it with the cover material.