contribution process

Wed Jul 2 19:33:39 UTC 2014

Hi Augie,

(Please don't read this email as advocating for not using the ML for review.)

On Wed, Jul 2, 2014 at 2:26 PM, Augie Fackler <raf at durin42.com> wrote:
>
> On Jul 2, 2014, at 3:15 PM, cowwoc <cowwoc at bbs.darktech.org> wrote:
>
>> Augie,
>>
>> I brought up this topic because someone brought it up off-list in response to my post. I have never attempted to contribute to hg (mostly because my lack of Python experience) but I can definitely see how (quoting Paul) "emailing patches is a crummy process".
>
> How? Please elaborate. I've been doing review-by-email for 6 or 7 _years_ as the sole way I've been involved with tools like Mercurial. I'd like specifics on how this is bad for potential contributors so I can have meaningful conversations on how to improve our contribution process for everyone.

I think that from a *descriptive* standpoint, for better or for worse,
there is a generation of developers right now whose relationship with
email is different than it would have been 6-7 years ago.  Email is
now de facto a website, not a set of interoperable protocols.

>
> I'd really like to chase down /what/ about patches-by-email is perceived to be awful. Right now I'm only hearing that y'all hate it, but not WHY.
>

The arguments I've heard are that:

 * Overhead to setting up SMTP server is high.
 * If you're using a non-IMAP/POP client, importing patches can be harder.
 * The "single place to look" for the currently under review patches
is clearer for pull requests.

>> I understand your reluctant to rely on non-OSS services, but I think people would benefit greatly from a better-organized, more visual contribution process as Github-style pull requests provide. I'm not asking you to stop accepting patches through the mailing list, but rather suggesting that accepting pull requests off Bitbucket would be a major step in the right direction.
>
> I've already stated how frustrating PRs are, both on BB and GH. That being a contribution mechanism for hg wouldn't be likely to get love from me or mpm, and I suspect the other reviewers would also not be fans of doing reviews there.

Is the review process on Phabricator or Rietveld, which to my
recollection both have preferences for squashing commits, any better?
Is the problem the email interoperability, or is it more tuned to the
issues related to the UI and the preference for sequential commits?

>
>
> [elided some "just try it", which I see as a nonstarter because it fragments the patch stream and the PR interface is lousy for reviewers in my experience]
>
>> I think you could pretty easily generate mailing list traffic from pull requests, but not the other way around (as you seem to be asking).
>
> That's an interesting-ish option. Would we also be able to get the reviews posted back to the ML so that the list remains the sole source of truth about what's going on?

This should be feasible with the webhook process for pull requests on
BitBucket.  We have a polling-based CI server for BitBucket that could
probably do this.  Again though, as you note BitBucket is not FLOSS.

-Matt

>
>
>> Gili
>>
>> On 02/07/2014 2:10 PM, Paul Nathan wrote:
>>> Development by emailing patches is a crummy process. That has been remarked upon more than a few times in the mailing list. I won't reiterate reasons previously mentioned. If contributions are down, use bitbucket as the central staging repo and use their pull requests.  There's also a foss clone of bb that I can't recall the name of atm.
>>>
>>> I don't see hg as lacking features, personally.  It beats git in nearly every way imo.
>>>
>>> ---
>>> Regards,
>>> Paul Nathan
>>>
>>> On July 2, 2014 10:28:24 AM PDT, Augie Fackler <raf at durin42.com> wrote:
>>>> On Jul 2, 2014, at 11:34 AM, cowwoc <cowwoc at bbs.darktech.org> wrote:
>>>>
>>>>> Augie,
>>>>>
>>>>> Perhaps it would help if we could streamline the contribution
>>>> process…?
>>>>
>>>> http://mercurial.selenic.com/wiki/ContributingChanges is the current
>>>> process, for reference. If you've got advice on how we could make that
>>>> more inviting, let's talk it out?
>>>>
>>>>> Right now the main hg repository is http://selenic.com/repo/hg. As
>>>> far as I can tell, there is no "pull request" process so contributing
>>>> could certainly be made easier. I don't think this is *the* answer for
>>>> fixing Mercurial's popularity but faster growth would certainly help us
>>>> catch up on any missing features (which, in my opinion, is half the
>>>> problem).
>>>>
>>>> I'm not sure how pull requests would help here? It'd add a place I have
>>>> to check for patches. Also, I'm completely unaware of
>>>> free-as-in-freedom tools that support pull requests - if you're aware
>>>> of something, please let me know. I'd at least spend a little while
>>>> investigating.
>>>>
>>>> If someone was interested in writing some sort of web service that
>>>> could turn our mailing list traffic into a somewhat pull-request-like
>>>> UI, that'd be something of great interest to me - that'd be the best of
>>>> both worlds.
>>>>
>>>>> Gili
>>>>>
>>>>> On 01/07/2014 1:36 PM, Augie Fackler wrote:
>>>>>> On Jun 30, 2014, at 1:35 PM, cowwoc <cowwoc at bbs.darktech.org> wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> I'd like to bring your attention to:
>>>>>>>
>>>> http://www.google.ca/trends/explore?hl=en-US&q=mercurial,+git,+github,+bitbucket&cmpt=q&content=1
>>>>>>> It seems to me that GitHub is directly responsible for Git's
>>>> popularity
>>>>>>> skyrocketing in 2009. By comparison, Mercurial's popularity seems
>>>> to be
>>>>>>> stagnant for a while. Bitbucket initially did a decent job, but
>>>> ever since
>>>>>>> they got bought out by Atlassian all public announcements emphasize
>>>> Git
>>>>>>> support and don't mention Mercurial at all.
>>>>>>>
>>>>>>> Case in point:
>>>>>>>
>>>> http://blog.bitbucket.org/2012/12/10/feature-branches-just-got-better/
>>>>>>> How do we go about reversing this trend?
>>>>>> I think one thing that would help would be users blogging more about
>>>> their workflows and how hg makes them productive. I keep meaning to do
>>>> this, but I have too many things pulling at my attention to ever manage
>>>> to get started.
>>>>>>> Gili
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> View this message in context:
>>>> http://mercurial.808500.n3.nabble.com/Mercurial-popularity-is-stagnant-tp4011549.html
>>>>>>> Sent from the General mailing list archive at Nabble.com.
>>>>>>> _______________________________________________
>>>>>>> Mercurial mailing list
>>>>>>> Mercurial at selenic.com
>>>>>>> http://selenic.com/mailman/listinfo/mercurial
>>>>
>>>>
>>>> ------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Mercurial mailing list
>>>> Mercurial at selenic.com
>>>> http://selenic.com/mailman/listinfo/mercurial
>>
>
>
> _______________________________________________
> Mercurial mailing list
> Mercurial at selenic.com
> http://selenic.com/mailman/listinfo/mercurial
>