Read only archive ; use https://github.com/JacORB/JacORB/issues for new issues
Bug 994 - NullPointerException during ORB initialization
Summary: NullPointerException during ORB initialization
Status: RESOLVED FIXED
Alias: None
Product: JacORB
Classification: Unclassified
Component: ORB (show other bugs)
Version: 3.5
Hardware: PC Windows NT
: P5 major
Assignee: Mailinglist to track bugs
URL:
Depends on:
Blocks:
 
Reported: 2014-10-01 08:59 UTC by franky.bugzilla
Modified: 2014-11-06 18:10 UTC (History)
1 user (show)

See Also:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description franky.bugzilla 2014-10-01 08:59:53 UTC 
I'm running multiple ORB instances (each with different properties) within a single JVM. The differnt ORB instances are initialized in parallel on multiple threads. All goes well until today where I got a NullPointerException during orb initialization:
java.lang.NullPointerException
at org.omg.CONV_FRAME.CodeSetComponentHelper.write(CodeSetComponentHelper.java:68)
at org.omg.CONV_FRAME.CodeSetComponentInfoHelper.write(CodeSetComponentInfoHelper.java:68)
at org.jacorb.orb.standardInterceptors.CodeSetInfoInterceptor.<init>(CodeSetInfoInterceptor.java:57)
at org.jacorb.orb.standardInterceptors.IORInterceptorInitializer.post_init(IORInterceptorInitializer.java:61)
at org.jacorb.orb.ORB.interceptorPostInit(ORB.java:2147)
at org.jacorb.orb.ORB.internalInit(ORB.java:2005)
at org.jacorb.orb.ORB.set_parameters(ORB.java:1953)
at org.omg.CORBA.ORB.init(ORB.java:251)

After some digging in the code, I've found out that the problem lies in org.jacorb.orb.giop.CodeSet's method "getLocalCodeSetComponentInfo" which is not correctly implemented for concurrent access: In this method a first thread can have created "localCodeSetComponentInfo" but not filled in the field "ForCharData" yet and a second thread can then return this half initialized "localCodeSetComponentInfo" object as the first check on "localCodeSetComponentInfo" being null is not within the synchronized block.
Comment 1 Nick Cross 2014-10-01 10:18:09 UTC 
Yes this does look possible. Are you able to reproduce it easily as a test case?
Comment 2 Nick Cross 2014-11-06 18:10:02 UTC 
Hopefully fixed by SHA 1a0a0c8a3fa9cc27d5c2929ff86101d5f06187e2