[Sugar-devel] Patches to mail list, or patches to trac?

[Simon Schampijer]

On 07/05/2010 02:47 PM, Tomeu Vizoso wrote:
> On 07/01/2010 08:32 PM, Gary Martin wrote:
>> Hi folks,
>>
>> Just wanted to ask the obvious question. Patches to mail list, or
>> patches to trac?
>
> About acceptance, the current process is in place until its wiki page is
> changed:
>
> http://wiki.sugarlabs.org/go/Development_Team/Code_Review
>
> Now, that process already mentioned reviews on mailing lists in some
> cases and of course, doesn't forbid people from sending patches to the
> ml for non-acceptance reviews.
>
>> Patches to mail-list seem great for quick random 'easy' feedback from
>> and for any one who cares to give it, and I really like seeing little
>> code snippets go past. However it seems vital that with the current
>> process we at least make a final closing stage (currently tickets in
>> trac) where the hard final call can be made and the loop closed.
>
> Yup, I guess this is the acceptance part of it which needs to be very
> clear to submitters and thus should be agreed by all maintainers of
> official modules.
>
>> Personally I read mail in a bunch of different places/devices and
>> there's no way I can currently (and sanely) keep track of all the
>> list activity and know who suggested what, what was actually agreed,
>> and what is still outstanding. I've had a few patches and reviews now
>> from kind folk posting to the mail-list, but for me, I've ended up
>> having to ask folks to create git clones so I can pull in patches in
>> a maintainable work flow.
>>
>> I dread to think what it must be like trying to look after several
>> large core modules!
>
> Last we discussed this, I understood that the conclusion is that the
> parties interested in a particular submission will keep pinging the
> maintainer in public (irc, ml) until it gets reviewed.
>
> I still think it would be much better to have a simpler queue such as
> http://bugs.sugarlabs.org/wiki/TomeuReviewQueue in which you know more
> clearly what is awaiting review and exactly which patches they are
> (well, when the submitter properly discards all patches and don't put
> several patches in the same ticket), but it was contended that the
> "bureaucracy" needed was too much  and discouraged contributions.
>
> Regards,
>
> Tomeu

The main problem I am seeing with having trac and the mailing list being 
used is that the two parts get out of sync. As a "well-behaving" 
maintainer, I have to follow both discussions in order to follow up.

For example take http://bugs.sugarlabs.org/ticket/1926  There is a 
ticket and a mailing list discussion with patches, icons etc. Now, there 
is no link in the ticket to this discussion, and one needs to follow 
both tracks. Of course adding policies such as "link the ml-discussion 
in the ticket" can help here.

Do people have a plan/policy for this?

Regards,
    Simon