Web lists-archives.com

Re: [RFC 0/3] Add support for --cover-at-tip

Le 13/11/2017 à 20:40, Jonathan Tan a écrit :
> On Mon, 13 Nov 2017 18:13:27 +0100
> Nicolas Morey-Chaisemartin <NMoreyChaisemartin@xxxxxxx> wrote:
>> v2:
>> - Enhance mailinfo to parse patch series id from subject
>> - Detect cover using mailinfo parsed ids in git am
> I noticed that this was done in the patch set by searching for "PATCH" -
> that is probably quite error-prone as not all patches will have a
> subject line of that form. It may be better to search for "0/" and
> ensure that it is immediately followed by an integer.

Yes this could be moved. I wasn't sure about the risk of colliding with something else.

> Also, it might be worth checking the message IDs to ensure that the
> PATCH M/Ns all indeed are replies to PATCH 0/N.

Doesn't that only work if mails [1..N] are reply to the cover ? Depending on the mailer and your option, this might not always be the case (I think).

>> - Support multiple patch series in a single run
> Is this done? I would have expected that some buffering of messages
> would be necessary, since you're writing a series of messages of the
> form <cover><patch 1>...<patch N> to the commits <patch 1>...<patch
> N><cover>.

This is for git am. It handles the fact that an input may contain multiple series (expect them to be in order).
When reaching patch N/N, it flushed the cover.

>> TODO:
>> - Add doc/comments
>> - Add tests
>> - Add a new "seperator" at the end of a cover letter.
>>   Right now I added a triple dash to all cover letter (manual or cover-at-tip) before shortlog/diff stat
>>   This allows manually written cover letters to be handle by git am --cover-at-tip without including the shortlog/diffstat but
>>   breaks compat with older git am as it is seen has a malformed patch. A new separator would solve that.
> I think the triple dash works. I tried "git am" with a cover letter with
> no triple dash, and it complains that the commit is empty anyway, so
> compatibility with older git am might not be such a big issue. (With the
> triple dash, it indeed complains about a malformed patch, as you
> describe.)

It kinda work but it makes the message difficult to understand for the user.
And change the behaviour from the previous releases.

Thanks for the feedback.