harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexei Fedotov <alexei.fedo...@gmail.com>
Subject Fwd: Quick Apache Harmony JNI question
Date Wed, 18 Feb 2009 18:22:44 GMT
Hello Pavel,
The following mail from Bert gives better perception and some answers
about his use cases. If this could help you fixing HARMONY-5505! :-)

---------- Forwarded message ----------
From: Alexei Fedotov <alexei.fedotov@gmail.com>
Date: 2008/12/12
Subject: Re: Quick Apache Harmony JNI question
To: webmaster@quadmore.com

Hello Bert,
I would appreciate if you redirected the following questions to the
mail list dev@harmony.apache.org. This is the proper way how Apache
provides non-commercial support. If confidential matters were
involved, you'd better ask for commercial support on the same list.

Here goes a short answer.
* Yes, we tried to implement DestroyJavaVM correctly. It does not
segfault on simple scenarios. Note, Apache Harmony VM does not allow
several VMs to coexist in the same address space.
* I believe Apache Harmony VM was better tested for Java 5.
* Check the date of the posts about Sun's DestroyJavaVM. Now it works
much better. Let me make a strong guess that unless you've spend more
than a week debugging your scenario, it is more likely that a bug is
in your invocation code then in Sun's DestroyJavaVM.
* I tried DestroyJavaVM in J9 and it worked as well. You may ask Tim
(on the list) more questions about IBM's VM.


On Fri, Dec 12, 2008 at 9:22 PM,  <webmaster@quadmore.com> wrote:
> Hello Alexei,
> I am writing you in regard to the Apache Harmony post:
> http://markmail.org/message/i2dzuxvhp2jwinsg#query:Harmony%20DestroyJAVAVM+page:1+mid:osffotk7clugto6n+state:results
> I am a developer in Quebec City, Canada. We're working on a 10 year old
> maritime traffic control system built in C++ and have to integrate a new
> radar interface which is in Java.
> I've done some prototypes which use JNI Invocation which we like but they
> leak some memory and over time we would need to stop and restart the JVM
> from the same C++ process.
> The DestroyJavaVM JNI method was not correctly implemented by Sun, reference:
> http://forums.sun.com/thread.jspa?forumID=52&threadID=305978
> but your exchange, as well as the new Intel JNI degugger, seem to suggest
> that Apache Harmony has implemented correctly the DestroyJavaVM method.
> Could you confirm it is now possible with Harmony to destroy the JVM and
> start a new one from the same C++ executable, and which version of Harmony
> (5 or 6) does this ? Any info would be appreciated.
> Thank you,
> Bert Szoghy

С уважением,
Алексей Федотов,
ЗАО <<Телеком Экспресс>>

С уважением,
Алексей Федотов,
ЗАО <<Телеком Экспресс>>

View raw message