JDK submit repo

Andrew Dinn adinn at redhat.com
Fri Mar 9 10:21:58 UTC 2018

On 08/03/18 17:57, Volker Simonis wrote:
>  On Tue, Mar 6, 2018 at 7:05 PM, Vladimir Kozlov
> <vladimir.kozlov at oracle.com> wrote:
>> My understanding is currently it is not enough for *HotSpot* changes. You
>> still need to ask for sponsor to run additional Hotspot tests.
> At FOSDEM 2018, the OpenJDK Lead Mark Reinhold answered my comment
> that it is still not possible for external committers to push HotSpot
> changes with the words:
> "Actually it is possible!" (see 7:51min of [1])
> He further detailed:
> "If you are working on shared code and you run it trough the
> submission repo ... and all test pass you should be clear to push it."
> (see 8:16 min of [1])
> So if that's not true it is, to put it mildly, at least disappointing.
> It then seems like Oracle now creates a "fake" submission forest
> before every FOSDEM (you may remember that we already had a different
> one before FOSDEM 2017 which could also not be used for submitting,
> i.e. pushing, real change) just to shine at the "OpenJDK Governing
> Board Q&A" [2] session.

I think Mark qualified that with some other words that ought to be taken
into account.

He said that /just like Oracle engineers/ you would still need to talk
to Jesper -- or whoever was the gatekeeper for the repo -- to confirm
that a push could actually go ahead after it had passed the tests
auto-run by the submit repo.

To me that is not prima facie unreasonable, given that the repo
gatekeeper may well be aware of other things going on in the update
process that might require i) extra testing to be scheduled for your
change, ii) delay of your proposed commit to allow validation of another
pending/recent commit, or whatever ... For me, the problem is not having
to obtain such an ack but whether this ack most often turns out to be a
simple nod vs regularly requiring some tortuous and slow follow-on process.

> If the submission repo can neither be used for pushing changes
> directly nor for pushing them manually after the tests executed by the
> repo have passed, please remove it completely or rename to something
> like "run_some_random_tests_if_youd_like_to_burn_some_cpu_power_in_the_oracle_cloud-repo".
> There's really no reason why any developer should use it in that case!

Sorry, but I think this is a rather extreme position to adopt.

> 1. If all the tests pass, I have to ask an Oracle developer anyway to
> run "some other" tests and push my change. So why should I use the
> submission forest in the first place?

If it actually means that changes can frequently go in 'on the nod' then
it saves a lot of unnecessary leg-work by Oracle staff trying to
co-ordinate pushing changes on your behalf that you could do yourself a
lot easier and with a lot better understanding in case they fail. That
would be a win-win.

> 2. If the tests don't pass, I get a cryptic, completely meaningless
> error mail. Again, I have to ask an Oracle developer to tell me what
> went wrong. I could have much more easily come to the same result by
> just asking an Oracle developer to push my change right from the
> start. He would have run it through "some internal tests" and told me
> right away what's wrong.

Well, obviously, better errors would help. However, if most of the time
submissions don't fail and are nodded through then -- even with the
current cryptic messages - that doesn't invalidate this as a potential
mutual win (see previous answer).

> In both cases neither I nor the Oracle sponsor gets any benefit from
> the current submission forest.

I don't think it is fair or reasonable to take this position.

> So Mark, you in your role as OpenJDK Lead, could you please give an
> authoritative to the question what the submission forest is good for
> and remove it in the case it turns out to be useless?
Apologies are perhaps due for interceding here -- I have no desire to
pre-empt Mark. I just wanted to point out that I don't see things as
being as bad as Volker obviously does.


Andrew Dinn
Senior Principal Software Engineer
Red Hat UK Ltd
Registered in England and Wales under Company Registration No. 03798903
Directors: Michael Cunningham, Michael ("Mike") O'Neill, Eric Shander

More information about the jdk-dev mailing list