On Oct 15, 2010, at 12:48 PM, Ted Kirby wrote:
> Thanks Kevan. I did try the build with JRE 1.6.0 IBM Windows 32 build
> pwi3260sr8fp1-20100624_01 (SR8 FP1), and think I got past this
> problem. I got out of memory:
>
> [INFO]
> ------------------------------------------------------------------------
> [INFO] Building Geronimo Plugins, Connector 1.6 :: Builder
> [INFO] task-segment: [clean, install]
> [INFO]
> ------------------------------------------------------------------------
> [INFO] [clean:clean {execution: default-clean}]
> [INFO] Deleting file set:
> C:\svn\g\trunk\plugins\connector-1_6\geronimo-connector-builder-1_6\target
> (included: [**], excluded: [])
> [INFO] [genesis:validate-configuration {execution: default}]
> [INFO] [geronimo-property:set-property {execution: set-property}]
> [INFO] [xmlbeans:xmlbeans {execution: default}]
> Time to build schema type system: 0.921 seconds
> Time to generate code: 9.844 seconds
> JVMDUMP006I Processing dump event "systhrow", detail
> "java/lang/OutOfMemoryError" - please wait.
> JVMDUMP032I JVM requested Snap dump using
> 'C:\svn\g\trunk\Snap.20101015.111238.8152.0001.trc' in res
> ponse to an event
> JVMDUMP010I Snap dump written to
> C:\svn\g\trunk\Snap.20101015.111238.8152.0001.trc
> JVMDUMP032I JVM requested Heap dump using
> 'C:\svn\g\trunk\heapdump.20101015.111238.8152.0002.phd' in
> response to an event
> JVMDUMP010I Heap dump written to
> C:\svn\g\trunk\heapdump.20101015.111238.8152.0002.phd
> JVMDUMP032I JVM requested Java dump using
> 'C:\svn\g\trunk\javacore.20101015.111238.8152.0003.txt' in
> response to an event
> JVMDUMP010I Java dump written to
> C:\svn\g\trunk\javacore.20101015.111238.8152.0003.txt
> JVMDUMP013I Processed dump event "systhrow", detail
> "java/lang/OutOfMemoryError".
> java.lang.OutOfMemoryError: Failed to create a thread: retVal
> -1073741830, errno 12
> null
> java.lang.OutOfMemoryError: Failed to create a thread: retVal
> -1073741830, errno 12
>
> I was using MAVEN_OPTS=-Xmx1500m -XX:MaxPermSize=512m
-XX:MaxPermSize=512m is not going to have any effect on an IBM runtime.
Search for ("Failed to create a thread" outofmemoryerror ) and you'll find
information like
http://publib.boulder.ibm.com/infocenter/javasdk/v6r0/index.jsp?topic=/com.ibm.java.doc.diagnostics.60/diag/problem_determination/win_oom_thread.html
Counter-intuitively, you may find that reducing your -Xmx setting (e.g.
-Xmx1024) may fix this problem. The .txt file may have information on the
number of threads being created. There are also settings to reduce the amount
of real memory that is allocated by default for each thread (IIRC). I can help
look at those files, if you'd like.
I'm pretty confident that if there's sufficient motivation to get your building
running on Windows IBM JDK, we can get it running...
--kevan
>
> I then went back to the Sun JDK and the build ran successfully with
> MAVEN_OPTS=-Xmx1024m -Xms256m -XX:PermSize=64m -XX:MaxPermSize=256m
> -XX:ReservedCodeCacheSize=64m
>
> On Thu, Oct 14, 2010 at 11:14 PM, Kevan Miller <[email protected]> wrote:
>>
>> On Oct 14, 2010, at 7:48 PM, Ted Kirby wrote:
>>
>>> Thanks Shawn and Janet. I switched to Sun JDK, and not only got by
>>> this problem, but, after a couple others, I got a successful build!
>>> WooHoo!
>>>
>>> I was getting the error with JRE 1.6.0 IBM Windows 32 build
>>> pwi3260sr7-20091217_01 (SR7). Do we support building only with Sun
>>> JDK?
>>
>> Umm. No. I can't think why we'd say that. Would seem there are two
>> possibilities:
>>
>> 1) IBM JDK has more strict validation and we should fix our source (I'm not
>> sure if Shawn and Janet are using Sun or IBM JDK), or
>> 2) There's a compiler bug
>>
>> Google 'inconvertible types' and you'll get a number of hits for compiler
>> bugs in this area. Your IBM JDK is nearly a year old. I'd guess there's a
>> decent chance that upgrading would fix your problem.
>>
>> --kevan
>>
>>
>>
