Oracle's Committers to staging repo
Steve Poole
spoole at linux.vnet.ibm.com
Mon Jun 17 04:20:59 PDT 2013
Volker - we're setting a poor example for how to run an open source project if we make people committers to the PPC project just because Oracles infrastructure does not allow separate access to the staging repo.
Having said that I do understand why you want to have this happen asap so I will vote yes on the understanding this is purely for mechanical reasons...
On 17 Jun 2013, at 10:08, Volker Simonis <volker.simonis at gmail.com> wrote:
> Hi Steve, (Neil, Jonathan,)
>
> that's all defined in the OpenJDK Bylaws [1]. We're an OpenJDK project and the only way to get committers into the project after the project has been started is by voting them in. I really don't see a problem with this approach - every OpenJDK project handles it this way.
>
> It's actually our (i.e. the project committer's) power that we can (and must) decide who will get commit rights to our repository.
>
> What problems do you see with this approach?
>
> If you don't see any insuperable problems I would kindly ask you to vote with yes. If all committers vote with yes, we can end the vote [2] and give Vladimir and Albert commit rights. They've already reviewed some of our changes and are waiting to push them via JPRT and I don't see why we shouldn't allow them to do so.
>
> I you will not vote we will have to wait 14 days until the vote is over before we can approve Vladimir and Albert as committers (by Lazy Consensus - see [2]). But I think that would be an unnecessary delay of the project.
>
> Regards,
> Volker Simonis
>
> [1]http://openjdk.java.net/bylaws
> [2]http://openjdk.java.net/projects/#committer-vote
>
>
>
> On Mon, Jun 17, 2013 at 10:28 AM, Steve Poole <spoole at linux.vnet.ibm.com> wrote:
> Let me try and ask the question again - why do we have to give Vladimir and others write access to the staging area via this means? We can't we update the access controls for the staging area to give existing main repos commiters immediate write access?
>
>
> On 14 Jun 2013, at 18:54, Iris Clark <iris.clark at oracle.com> wrote:
>
> > Hi, Volker.
> >
> > I think that Vladimir answered the question very well. Similar to what we've done for JDK releases, we'll have a closed forest which "overlays" on the open forest (creating nested forests). This is what the hotspot portion would look like:
> >
> > stage/hotspot [OPEN]
> > stage/hotspot/src/closed [CLOSED]
> > stage/hotspot/make/closed [CLOSED]
> > stage/hotspot/test/closed [CLOSED]
> >
> > You won't be able to see any of that [CLOSED] source so from your perspective nothing will change.
> >
> > The existence of the closed forest which corresponds to ppc-aix-port/stage will allow Vladimir and others to modify and test the closed source. Ideally there are "synchronized" pushes to all of the affected trees at the same time if one fix needs to change multiple trees. This minimizes the window for broken tests or build targets. I think we had one a couple of weeks ago between langtools and jdk for lambda bridge methods.
> >
> > Thanks,
> > Iris
> >
> > "synchronized" because we just execute the pushes in series, there is no magic.
> >
> > -----Original Message-----
> > From: Vladimir Kozlov
> > Sent: Friday, June 14, 2013 7:51 AM
> > To: Volker Simonis
> > Cc: Steve Poole; ppc-aix-port-dev at openjdk.java.net; Iris Clark
> > Subject: Re: Oracle's Committers to staging repo
> >
> > On 6/14/13 6:00 AM, Volker Simonis wrote:
> >> Per the OpenJDK Bylaws, it is only possible to appoint committers at project startup.
> >> Later on, the only way to become committer for a project is to cast a vote.
> >>
> >> Vladimir and Albert are mostly expected to work on the "closed" part
> >> of our repository which we (i.e. "we from outside
> >> Oracle") can't do anyway.
> >>
> >> As far as I a understand it, Oracle will push their "closed" PPC port
> >> into our staging repository. We won't be able to see it because the
> >> closed part is a separate sub-forest in the hotspot directory. But the Oracle folks will then be able to build and test their closed PPC port right from our staging repository.
> >> @Iris: is my understanding right here? Is it really possible to place
> >> the "closed" PPC port into our "open" staging repository?
> >
> > It is forest. Closed repos are created on side on different file server and not related to open part in any way (in Mercurial sense). And it is the same setup as in our Hotspot sources.
> > When we need to push related changes (for example, change method signature) into open and close repos we create separate changesets in those repos but with the same changeset header/descriptor, at least with the same bugid:synopsis and summary:description lines (in ppc-aix project we will get different authors and reviewers).
> > To prepare for testing we pull them together in one place. Closed repos are cloned as sub-directories of open sources:
> > hotspot/src/closed, hotspot/make/closed and hotspot/test/closed. So from build system view it is one source with top /hotspot/ directory.
> > We have to test and push such changes together since they are related. That is why we asked permission to push into open part of this project.
> >
> > Note: our testing system JPRT keeps builds from push job for long time (years) so later, when some more complex testing finds a problem, we can track which push/changes introduced the regression. It helps diagnose the problem.
> > I want to use JPRT to push all changes which modify shared sources and may affect our closed parts.
> >
> >> I admit that's a strange constellation for an open source project and
> >> I totally agree with you that it should be Oracles sole responsibility
> >> to fix problems with their closed ports but taking into account the
> >> real circumstances I think it is the best and most pragmatic solution
> >> would to go this way. After all - the faster Oracle fixes the problems
> >> with their closed port, the faster we get into the main branch:)
> >
> > Yes, it is my main motivation. It is faster to deal with one conflict at time than allocate resources later to fix 100s conflicts. This way we will be more confident about these changes when we will make decision to push it onto main branch.
> >
> >> And if we all do vote now (after all we're currently just 4
> >> committres) we wouldn't even have to wait until the full voting period
> >> of 14 days ends:)
> >
> > Thanks,
> > Vladimir
> >
> >>
> >> Regards,
> >> Volker
> >>
> >>
> >>
> >> On Fri, Jun 14, 2013 at 1:32 PM, Steve Poole <spoole at linux.vnet.ibm.com <mailto:spoole at linux.vnet.ibm.com>> wrote:
> >>
> >> Hmm - shouldn't the staging repo be setup so that both ppc-aix and jdk8 committers have write permissions? (Rather
> >> than having to artifically add people to the ppc project?)
> >>
> >>
> >>
> >>
> >> On 14 Jun 2013, at 09:47, Volker Simonis <volker.simonis at gmail.com <mailto:volker.simonis at gmail.com>> wrote:
> >>
> >>> Sorry, it seems I was overhasty!
> >>> I can only appoint Authors, so I'll start a vote for you to make you committers.
> >>>
> >>> Regards,
> >>> Volker
> >>>
> >>>
> >>>
> >>> On Fri, Jun 14, 2013 at 10:36 AM, Volker Simonis <volker.simonis at gmail.com <mailto:volker.simonis at gmail.com>> wrote:
> >>>
> >>> Sure, and the good thing is that we don't need any vote for this! I'll just appoint you as committers -
> >>> somtimes autarchy is great:)
> >>> I'll send a mail to the openjdk registrar.
> >>>
> >>> Regards,
> >>> Volkre
> >>>
> >>> On Thu, Jun 13, 2013 at 7:01 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com
> >>> <mailto:vladimir.kozlov at oracle.com>> wrote:
> >>>
> >>> Volker,
> >>>
> >>> Do we (me and Albert) need to have ppc-aix project committer status to be able push into Hotspot staging
> >>> repo http://hg.openjdk.java.net/__ppc-aix-port/stage/hotspot
> >>> <http://hg.openjdk.java.net/ppc-aix-port/stage/hotspot>?
> >>>
> >>> Albert is located in your time zone and will help with running tests on Oracle side (with closed repos)
> >>> and using JPRT. And he also may help with periodic sync staging repo with jdk8/tl. We needs this if we
> >>> have to adjust Oracle closed sources.
> >>>
> >>> Regards,
> >>> Vladimir
> >>>
> >>>
> >>>
> >>
> >>
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/ppc-aix-port-dev/attachments/20130617/3de533ac/attachment.html
More information about the ppc-aix-port-dev
mailing list