[oe] [meta-java][PATCH] OpenJDK-8: For x86 and ARM

[Otavio Salvador]

On Fri, Sep 11, 2015 at 4:55 AM, Richard Purdie
<richard.purdie at linuxfoundation.org> wrote:
> On Fri, 2015-08-28 at 14:29 -0300, Otavio Salvador wrote:
>> On Fri, Aug 28, 2015 at 4:12 AM, Richard Leitner
>> <richard.leitner at skidata.com> wrote:
>> > On 08/28/2015 03:14 AM, Vinay K. Awasthi wrote:
>> >> Rebased it again with latest change-set and tested it on oe-core master with latest mate-openembedded.
>> >>
>> >> There is no change in recipe.
>> >>
>> >> This enables OpenJDK-8 jre/jdk for x86 and ARM.
>> >>
>> >> This recipe is tested for arm (zero VM) and x86 (hotspot vm).
>> >>
>> >> To build just use bitbake openjdk-8.
>> >
>> > I successfully compiled and tested it for/on an ARMv7 Cortex A9 CPU:
>> > # /usr/lib/jvm/java-8-openjdk/jre/bin/java -version
>> > openjdk version "1.8.0-internal"
>> > OpenJDK Runtime Environment (build 1.8.0-internal-leri_2015_08_28_08_47-b00)
>> > OpenJDK Zero VM (build 25.60-b23, interpreted mode)
>> >
>> > Also X11/awt/swing works fine! Only problem for me is that ZeroVM is
>> > running in "interpreted mode". Therefore my programs are running really
>> > slow.
>> >
>> > Was the "mixed mode" available/working in your tests?
>> >
>> > IMHO a working JIT (=> mixed-mode) is a must-have for merging this patch
>> > upstream. What do you think Otavio?
>>
>> Yes, I agree. There are other people working in OpenJDK 8 here[1] (at
>> ml) it seems so maybe we could share some ideas on how to fix it.
>
> I'm hearing there is a little frustration about moving forward with
> this. I'd like to suggest that a staged approach might be better in this
> case, for example if OpenJDK-8 is working on IA, how about adding it
> with:
>
> # mixed-mode isn't working on ARM yet
> COMPATIBLE_HOST = '(i.86|x86_64).*-linux'
>
> With a large change like this, I think its perhaps ok to take a staged
> approach, as long as its clear what is/isn't expected to work?
> If/as/when the ARM people have time to make it work there, patches can
> be sent to change that? This way it unblocks people and doesn't hold
> keen contributors back.

I am without much time to look at this and do a proper testing. If any
other maintainer wants to handle this I would be thankful.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750