----- On Apr 19, 2018, at 8:10 AM, Matthew Wilcox willy@xxxxxxxxxxxxx wrote:
> On Thu, Apr 19, 2018 at 06:21:42AM +0900, Junio C Hamano wrote:
>> Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> writes:
>>
>> > But IMO this patch is really lacking a few things before being ready:
>> >
>> > 1. You have no tests for this. See t/t9001-send-email.sh for examples,
>> > ...
>> > 2. Just a few lines down from your quoted hunk we have this:
>> > ... code about $supress_cc{<token>} ...
>> > Your change should at least describe why those aren't being updated,
>> > but probably we should add some other command-line option for
>> > ignoring these wildcards, e.g. --[no-]wildcard-by-cc=reviewed
>> > --[no-]wildcard-by-cc=seen etc, and we can make --[no-]signed-off-by
>> > a historical alias for --[no-]wildcard-by-cc=signed-off.
>> > 3. Ditto all the documentation in "man git-send-email" about
>> > ...
>>
>> Thanks, I agree that 2. (the lack of suppression) is a showstopper.
>
> I agree with that (and the lack of tests, obviously)
>
>> I'd further say that these new CC-sources should be disabled by
>> default and made opt-in to avoid surprising existing users.
>
> But I disagree with this. The current behaviour is surprising to
> existing users, to the point where people are writing their own scripts
> to replace git send-email (which seems crazy to me).
We could perhaps go with a whitelist approach. The four
main match I would be tempted to add are: Acked-by, Reported-by,
Reviewed-by, and Tested-by.
My workflow is to initially CC a bunch of relevant maintainers
when sending out a patch, and as the Acked, Reviewed and Tested
by tags come it, I replace those CC with the relevant tag.
I never expected them to stop being CC'd when switching between
those categories.
Thanks,
Mathieu
>
>> One thing we also need to be very careful about is that some of the
>> fields may not even have an e-mail address. We can expect that
>> S-o-b and Cc would be of form "human readable name <email@xxxxxxxx>"
>> by their nature, but it is perfectly fine to write only human
>> readable name without address on random lines like "suggeted-by" and
>> "helped-by". There needs a way for the end-user to avoid using data
>> found on such lines as if they are valid e-mail addresses.
>
> I also agree with this. I'll add some test-cases and make sure we only
> add these if they're valid email addresses.
--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
