Re: Pull-Requests not worked on

[Colin McCabe <cm...@apache.org>]

So, there has been some instability in the junit tests recently due to
some big features landing.  This is starting to improve a lot, though,
so hopefully it won't be a problem for much longer.

If you do hit what you think is a bogus unit test failure, you can type
"retest this please" in the PR to trigger another junit run.

It would be good to close some of the stale PRs.  I have some myself
that I should probably close since they have been superseded by
different approaches.

With regard to OSGi specifically, it might be hard to find someone who
has the relevant expertise to review the patch.  Since our understanding
is limited, maybe it makes sense to have  someone else who understands
OSGi repackage the software for that system?  I'm just tossing out ideas
here, though-- I could be wrong.

best,
Colin


On Tue, May 23, 2017, at 09:48, Jeff Widman wrote:
> I agree with this. As a new contributor, it was a bit demoralizing to
> look
> at all the open PR's and wonder whether when I sent a patch it would just
> be left to sit in the ether.
> 
> In other projects I'm involved with, more typically the maintainers go
> through periodically and close old PR's that will never be merged. I know
> at this point it's an intimidating amount of work, but I still think it'd
> be useful to cut down this backlog.
> 
> Maybe at the SF Kafka summit sprint have a group that does this? It's a
> decent task for n00bs who want to help but don't know where to start to
> ask
> them to help identify PR's that are ancient and should be closed as they
> will never be merged.
> 
> On Tue, May 23, 2017 at 4:59 AM, <ma...@sdv-it.de> wrote:
> 
> > Hello everyone
> >
> > I am wondering how pull-requests are handled for Kafka? There is currently
> > a huge amount of PRs on Github and most of them are not getting any
> > attention.
> >
> > If the maintainers only have a look at PR which passed the CI (which makes
> > sense due to the amount), then there is a problem, because the CI-pipeline
> > is not stable. I've submitted a PR myself which adds OSGi-metadata to the
> > kafka-clients artifact (see 2882). The pipeline fails randomly even though
> > the change only adds some entries to the manifest.
> > The next issue I have is, that people submitting PRs cannot have a look at
> > the failing CI job. So with regards to my PR, I dont have a clue what went
> > wrong.
> > If I am missing something in the process please let me know.
> > Regarding PR 2882, please consider merging because it would safe the
> > osgi-community the effort of wrapping the kafka artifact and deploy it
> > with different coordinates on maven central (which can confuse users)
> > regards
> > Marc
> >

Re: Pull-Requests not worked on

[Ismael Juma <is...@juma.me.uk>]

No, only Apache Infra people have commit rights to the GitHub project. The
Kafka committers push to the Apache Git repo and that gets synced
automatically to GitHub. My understanding is that there is a desire to
avoid a situation where people can push directly to GitHub for provenance
reasons (I don't remember the details).

Ismael

On Tue, Jun 20, 2017 at 12:52 AM, Jeff Widman <je...@netskope.com> wrote:

> On Thu, Jun 8, 2017 at 5:00 PM, Ismael Juma <is...@juma.me.uk> wrote:
>
> > One of the reasons why some PRs remain open even if they should be
> > closed is that we have no way to close PRs ourselves without referencing
> > them in a commit or asking Apache Infra for help. So, abandoned PRs tend
> to
> > remain open. This is very frustrating, but there is no solution in sight.
> >
>
>
> Why is this?
>
> Don't those with commit rights to the project have collaborator permissions
> and thus the ability to close PRs (ala most Github-based projects?)
>

Re: Pull-Requests not worked on

[Jeff Widman <je...@netskope.com>]

On Thu, Jun 8, 2017 at 5:00 PM, Ismael Juma <is...@juma.me.uk> wrote:

> One of the reasons why some PRs remain open even if they should be
> closed is that we have no way to close PRs ourselves without referencing
> them in a commit or asking Apache Infra for help. So, abandoned PRs tend to
> remain open. This is very frustrating, but there is no solution in sight.
>


Why is this?

Don't those with commit rights to the project have collaborator permissions
and thus the ability to close PRs (ala most Github-based projects?)

Re: Pull-Requests not worked on

[Ismael Juma <is...@juma.me.uk>]

Hi all,

A few clarifications:

1. Anyone can see the console output for a failed job, a login is not
required for that.

2. One of the reasons why some PRs remain open even if they should be
closed is that we have no way to close PRs ourselves without referencing
them in a commit or asking Apache Infra for help. So, abandoned PRs tend to
remain open. This is very frustrating, but there is no solution in sight.

3. As Colin pointed out, the stability of the JUnit tests has improved a
lot in recent weeks. Thanks to everyone who has helped with this.

4. "retest this please" became available once I switched the Jenkins PR
builder to the community one (Apache previously only allowed the usage of
the Cloudbees Enterprise one). I updated the relevant wiki pages and I
think I sent an email to the mailing list at the time.

Ismael



On Fri, Jun 9, 2017 at 12:47 AM, James Cheng <wu...@gmail.com> wrote:

> I always thought that we had to repush a commit or close/open the PR in
> order to retrigger the build. This is the first time I've heard that you
> can do that.
>
> Does "retest this please" cause the tests to be re-run, simply because
> it's a comment on a PR? Or is it some magic with how we set things up? Or
> does someone simply end up reading the comment, and then pushing the button
> on behalf of the requester?
>
> -James
>
> > On Jun 8, 2017, at 1:48 PM, Colin McCabe <cm...@apache.org> wrote:
> >
> > So, there has been some instability in the junit tests recently due to
> > some big features landing.  This is starting to improve a lot, though,
> > so hopefully it won't be a problem for much longer.
> >
> > If you do hit what you think is a bogus unit test failure, you can type
> > "retest this please" in the PR to trigger another junit run.
> >
> > It would be good to close some of the stale PRs.  I have some myself
> > that I should probably close since they have been superseded by
> > different approaches.
> >
> > With regard to OSGi specifically, it might be hard to find someone who
> > has the relevant expertise to review the patch.  Since our understanding
> > is limited, maybe it makes sense to have  someone else who understands
> > OSGi repackage the software for that system?  I'm just tossing out ideas
> > here, though-- I could be wrong.
> >
> > best,
> > Colin
> >
> >
> > On Tue, May 23, 2017, at 09:48, Jeff Widman wrote:
> >> I agree with this. As a new contributor, it was a bit demoralizing to
> >> look
> >> at all the open PR's and wonder whether when I sent a patch it would
> just
> >> be left to sit in the ether.
> >>
> >> In other projects I'm involved with, more typically the maintainers go
> >> through periodically and close old PR's that will never be merged. I
> know
> >> at this point it's an intimidating amount of work, but I still think
> it'd
> >> be useful to cut down this backlog.
> >>
> >> Maybe at the SF Kafka summit sprint have a group that does this? It's a
> >> decent task for n00bs who want to help but don't know where to start to
> >> ask
> >> them to help identify PR's that are ancient and should be closed as they
> >> will never be merged.
> >>
> >> On Tue, May 23, 2017 at 4:59 AM, <ma...@sdv-it.de> wrote:
> >>
> >>> Hello everyone
> >>>
> >>> I am wondering how pull-requests are handled for Kafka? There is
> currently
> >>> a huge amount of PRs on Github and most of them are not getting any
> >>> attention.
> >>>
> >>> If the maintainers only have a look at PR which passed the CI (which
> makes
> >>> sense due to the amount), then there is a problem, because the
> CI-pipeline
> >>> is not stable. I've submitted a PR myself which adds OSGi-metadata to
> the
> >>> kafka-clients artifact (see 2882). The pipeline fails randomly even
> though
> >>> the change only adds some entries to the manifest.
> >>> The next issue I have is, that people submitting PRs cannot have a
> look at
> >>> the failing CI job. So with regards to my PR, I dont have a clue what
> went
> >>> wrong.
> >>> If I am missing something in the process please let me know.
> >>> Regarding PR 2882, please consider merging because it would safe the
> >>> osgi-community the effort of wrapping the kafka artifact and deploy it
> >>> with different coordinates on maven central (which can confuse users)
> >>> regards
> >>> Marc
> >>>
>
>

Re: Pull-Requests not worked on

[James Cheng <wu...@gmail.com>]

I always thought that we had to repush a commit or close/open the PR in order to retrigger the build. This is the first time I've heard that you can do that. 

Does "retest this please" cause the tests to be re-run, simply because it's a comment on a PR? Or is it some magic with how we set things up? Or does someone simply end up reading the comment, and then pushing the button on behalf of the requester?

-James

> On Jun 8, 2017, at 1:48 PM, Colin McCabe <cm...@apache.org> wrote:
> 
> So, there has been some instability in the junit tests recently due to
> some big features landing.  This is starting to improve a lot, though,
> so hopefully it won't be a problem for much longer.
> 
> If you do hit what you think is a bogus unit test failure, you can type
> "retest this please" in the PR to trigger another junit run.
> 
> It would be good to close some of the stale PRs.  I have some myself
> that I should probably close since they have been superseded by
> different approaches.
> 
> With regard to OSGi specifically, it might be hard to find someone who
> has the relevant expertise to review the patch.  Since our understanding
> is limited, maybe it makes sense to have  someone else who understands
> OSGi repackage the software for that system?  I'm just tossing out ideas
> here, though-- I could be wrong.
> 
> best,
> Colin
> 
> 
> On Tue, May 23, 2017, at 09:48, Jeff Widman wrote:
>> I agree with this. As a new contributor, it was a bit demoralizing to
>> look
>> at all the open PR's and wonder whether when I sent a patch it would just
>> be left to sit in the ether.
>> 
>> In other projects I'm involved with, more typically the maintainers go
>> through periodically and close old PR's that will never be merged. I know
>> at this point it's an intimidating amount of work, but I still think it'd
>> be useful to cut down this backlog.
>> 
>> Maybe at the SF Kafka summit sprint have a group that does this? It's a
>> decent task for n00bs who want to help but don't know where to start to
>> ask
>> them to help identify PR's that are ancient and should be closed as they
>> will never be merged.
>> 
>> On Tue, May 23, 2017 at 4:59 AM, <ma...@sdv-it.de> wrote:
>> 
>>> Hello everyone
>>> 
>>> I am wondering how pull-requests are handled for Kafka? There is currently
>>> a huge amount of PRs on Github and most of them are not getting any
>>> attention.
>>> 
>>> If the maintainers only have a look at PR which passed the CI (which makes
>>> sense due to the amount), then there is a problem, because the CI-pipeline
>>> is not stable. I've submitted a PR myself which adds OSGi-metadata to the
>>> kafka-clients artifact (see 2882). The pipeline fails randomly even though
>>> the change only adds some entries to the manifest.
>>> The next issue I have is, that people submitting PRs cannot have a look at
>>> the failing CI job. So with regards to my PR, I dont have a clue what went
>>> wrong.
>>> If I am missing something in the process please let me know.
>>> Regarding PR 2882, please consider merging because it would safe the
>>> osgi-community the effort of wrapping the kafka artifact and deploy it
>>> with different coordinates on maven central (which can confuse users)
>>> regards
>>> Marc
>>>