Re: [PATCH v2] contrib: git-cpcover: copy cover letter

[Junio C Hamano <gitster@pobox.com>]

Eric Sunshine <sunshine@sunshineco.com> writes:

> On Tue, Dec 3, 2019 at 11:45 PM Jonathan Nieder <jrnieder@gmail.com> wrote:
>> Michael S. Tsirkin wrote:
>> > My flow looks like this:
>> > 2. vi <dir>/v<n-1>-0000-cover-letter.patch <dir>/v<n>-0000-cover-letter.patch
>> > copy subject and blurb, avoiding patchset stats
>> > 3. add changelog update blurb as appropriate
>> >
>> > The following perl script automates step 2 above.
>>
>> Neat.  I wonder, should "git format-patch" learn an option for this?
>>         git format-patch -v<n> --cover-letter \
>>                 --last-cover-letter=<dir>/v<n-1>-0000-cover-letter.patch \
>>                 -o <dir>
>
> That was my first thought, as well, although, as this has similar
> purpose to the new git-format-patch --cover-from-description= option,
> perhaps a more suitable name might be --copy-cover-from= or something?
>
> I could even imagine a new option -V<n> which has the combined effect
> of setting the re-roll count (like -v) and automagically copying the
> cover letter material from cover letter v<n-1> located in <dir>.

I actually looked into doing something similar but without any new
option (i.e. unconditionally --cover-letter with -v<n> would check
for v<n-1>-0000-cover.letter and does the right thing) some time
ago.  I do not recall why I gave up (not that I tried very hard),
but IIRC, the current reroll-count was not passed down in the
callchain to make_cover_letter() to do this.

But I think that was even before we integrated the range-diff stuff,
which does seem to use the "given we are doing <n>, let's compare
with <n-1>" thing, so perhaps it is not too difficult.

I am just saying that I think the change would not have to be opt-in,
but can be unconditionally made, simply because replacing the BLURB
HERE placeholder with *anything* written by human user previously is
a 100% improvement ;-)

Thanks.