Need help testing the EA builds

[Dmitry Samersoff]

Volker,

Currently tmtools (jstack, jmap, jinfo, ... ) has two modes of
operations - one is cooperative (attach API based) other one is
non-cooperative (SA based).

Long term plan is completely separate these two modes, with jhsdb as the
only entry point to SA.

This work is partially done - SA based mode is accessible via jhsdb.

But removal of SA based functions from existing tools has huge
compatibility impact (e.g. no more mixed stacktrace from jstack etc) and
can't be done at one shot.

Nothing besides tmtools depends to SA i.e. jcmd it self doesn't depend
to SA.

-Dmitry

On 2016-03-04 22:08, Volker Simonis wrote:
> Hi Alan, Dmitry,
> 
> I've tried today to build the jake repository on AIX. Unfortunately
> that doesn't work because we now have a hard dependency on SA which is
> required by jcmd:
> 
> $cat jdk/src/jdk.jcmd/share/classes/module-info.java
> module jdk.jcmd {
>     requires jdk.attach;
>     requires jdk.jvmstat;
>     requires jdk.hotspot.agent; // until JDK-8059035 is complete
> }
> 
> This will break every platform which does not implement the SA.
> 
> There already exists a bug for this issue:
> 8059035: Break the (implicit) dependency from jdk.jcmd to jdk.hotspot.agent
> https://bugs.openjdk.java.net/browse/JDK-8059035
> 
> Is there any chance that this will be resolved soon?
> Or is there an easy way to disable this dependency on AIX?
> 
> Regards,
> Volker
> 
> 
> On Wed, Mar 2, 2016 at 4:00 PM, Alan Bateman <Alan.Bateman at oracle.com> wrote:
>>
>> We are starting to think through how to bring the module system that is the
>> monster patch in jigsaw/jake forest into the JDK 9 main line.
>>
>> There are many reasons to do this. One reason is to reduce testing effort.
>> Another is to reduce the cost of sync ups as it takes a lot of effort each
>> week to merge in the changes from JDK 9. We also have confused some
>> developers by having two early access (EA) builds, one EA build with Project
>> Jigsaw, the other without. It should also be obvious that time is moving on
>> and we don't want to be rushing things right before JDK 9 Feature Complete.
>>
>> Merging into the JDK 9 main line will be a big step and there are many
>> things that need to be done to make this happen. The proposal will most
>> likely be to take snapshot of what we have in jake and bring it into JDK 9
>> after stabilization, testing and of course review. Once it is in JDK 9 then
>> we'll continue to iterate and work through the many issues that need
>> attention.
>>
>> One thing that would be useful is to get as many people as possible to try
>> out the EA builds [1]. It really helps confidence when there is confirmation
>> that existing libraries and applications work as before. If there are bugs
>> or issues that involve any of the supported interfaces then we want to know.
>> We know well that there are several compatibility issues and we've attempted
>> to capture those in the Risks and Assumption section of JEP 261 [2].
>>
>> We also interested in bug reports and issues from those trying out the EA
>> builds to create their own modules or migrating existing code or libraries
>> to modules. Bonus points for those brave enough to try out some of the many
>> new APIs too.
>>
>> -Alan
>>
>> [1] https://jdk9.java.net/jigsaw/
>> [2] http://openjdk.java.net/jeps/261


-- 
Dmitry Samersoff
Oracle Java development team, Saint Petersburg, Russia
* I would love to change the world, but they won't give me the sources.