Zsombor,
>> Instead of discouraging contributions
My suggestion was to discourage patches that have *only* very, very tiny and
non-essential updates – like changing case of a label word from “G” to “g”,
removing a blank line, etc; or patches that touch hundreds of files for
trivial/non-essential updates. I think these can’t be a way to ease into the
project. A better way would be to take time to understand the project
functionality and contribute towards enhancing it. I think an excellent example
in this regard would be Colm’s initial contributions, which significantly
improved the unit test coverage for Ranger plugins. It will help the Ranger
community immensely if contributors take such a path.
>> Is there any particular reason, why the current workflow used as is ?
Ranger community has been using review-board since its beginning, and it seems
to be working well. Many Apache project do use this approach.
As new contributors join the community, their experience/expectation might be
different. Use of ‘pull requests’ to deal with patch/review/merge was suggested
sometime back as well. I think we should give this a shot, to find if this
approach makes it easier for everyone. However, bear in mind that people might
have bias towards a specific approach, based on their prior experience and
comfort level. Forcing everyone to change might take effort/time.
Thanks,
Madhan
On 5/24/17, 4:31 AM, "Zsombor" <gzsom...@gmail.com> wrote:
Instead of discouraging contributions, the project should ease the
contribution - and also the review process.
The current JIRA + Reviewboard infrastructure needs a lot's of unnecessary
manual steps, which hurts everyone who wants to help.
Currently, the contributor needs to do the following, after they have a
working commit in their git repository
1, open a jira ticket
2, generate a patch from git
2, create a review boad request, uploading the patch to it
3, upload the patch to jira
4, if something is not correct, they have to repeat 2-3.
Similarly, for the reviewer, he has to manually download, apply, and run
the tests locally, even when he thinks the patch is ok.
My suggestion is to switch to a pull request based workflow, where the
manual patch creation, upload-download process could be omitted, and travis
or some other automatic build service should be introduced, to ensure, that
the base sanity tests are not omitted accidentally.
With this process, a commiter could review and approve a trivial commits
in less than a minute.
Is there any particular reason, why the current workflow used as is ?
Regards,
Zsombor
On Wed, May 24, 2017 at 12:28 PM, Gautam Borad <gbo...@gmail.com> wrote:
> +1 for Madhan's recommendations,
>
> Colm, I agree that we should not discourage new contributions. However, I
> think, we should also not encourage such single line/whitespace
> contributions. We want contributors who can do more functionals/feature
> changes and while doing that they can also fix the trivial issues
> (whitespace etc)
>
> Since each contribution to Ranger requires creating Jira/RR, if we start
> having lot of such trivial contributions, the community will be
overwhelmed
> with activities(mails etc) like this and that can lead to ignoring of a
> real functional change, when it comes.
>
> In fact, the Apache page on Contributors itself says :
>
> "Being a contributor simply means that you take an interest in the project
> and contribute in some way, ranging from asking sensible questions (which
> documents the project and provides feedback to developers) through to
> providing *new features* as patches."
>
>
> So yes, we should encourage contributors, but encourage them to try and
> understand Ranger and add more features/functionalities and eventually
> "earn" the title of a committer. Thanks.
>
>
>
> On Wed, May 24, 2017 at 2:19 PM, Colm O hEigeartaigh <cohei...@apache.org>
> wrote:
>
> > Hi Madhan,
> >
> > Trivial commits provide a path to get new contributors on board to the
> > project - something that the project needs IMO. Yes it may make
> backporting
> > fixes a little more difficult, but it's hardly an intractable problem to
> > figure out some whitespace changes between branches - it's not as if
> Ranger
> > is a particularly large project.
> >
> > Having said that I agree that some of the very trivial patches could
> maybe
> > be consolidated a bit more. I will encourage future review requests that
> > have a very trivial spelling fix to hold on to the fix for a while, so
> that
> > we can fix multiple spelling fixes etc. at the same time.
> >
> > Colm.
> >
> > On Wed, May 24, 2017 at 7:27 AM, Madhan Neethiraj <mad...@apache.org>
> > wrote:
> >
> > > All,
> > >
> > >
> > >
> > > I notice a number of recent patches address trivial issues like white
> > > space, spelling mistakes (one patch just changed a single letter in a
> > > label). And few other patches update a large number of files for
> > > trivial/non-functional changes – like whitespaces. I strongly suggest
> we
> > > refrain from authoring/encouraging such patches – for many reasons.
One
> > of
> > > the main reasons is the overhead such updates add in backporting
> > > real/critical fixes (that would come later) to other branches, as
these
> > > changes might force dealing with merge conflicts.
> > >
> > >
> > >
> > > Since the changes introduced in such patches are not essential, I
would
> > > suggest to take these up when these source files are updated for other
> > > functional fixes. I would greatly appreciate if the patches focus on
> > > fixing/enhancing Ranger functionality; this would be benefit the
> > community
> > > immensely.
> > >
> > >
> > >
> > > Thanks,
> > >
> > > Madhan
> > >
> > >
> > >
> >
> >
> >
> > --
> > Colm O hEigeartaigh
> >
> > Talend Community Coder
> > http://coders.talend.com
> >
>
>
>
> --
> Regards,
> Gautam.
>
