WLS 5.1.0 SP10
USS/390 10.00 02 (on a 7060)
JDK: build 1.3.0, J2RE 1.3.0 IBM build hm130-20010207 (JIT disabled)
I'm currently porting an application to USS/390 and ran into a snag with JSP and RMI compilation. Both weblogic.rmic and weblogic.jspc fail with a null pointer exception:
java.lang.NullPointerException
at java.util.StringTokenizer.<init>(StringTokenizer.java:128)
at weblogic.utils.compiler.CodeGenerator.parse(CodeGenerator.java:508)
at weblogic.utils.compiler.CodeGenerator.generateCode(CodeGenerator.java
:312)
at weblogic.utils.compiler.CodeGenerator.generate(CodeGenerator.java:263
)
at weblogic.jspc.runBody(jspc.java:121)
at weblogic.utils.compiler.Tool.run(Tool.java:80)
at weblogic.jspc.main(jspc.java:153)
It looks like the string tokenizer is being constructed with NULL for the string argument, and is throwing an NPE. Working backwards with the help of a decompiler shows that CodeGenerator.parse is creating this string with the entry associated with "main" in a hashtable.
Now, this same build process works on other UNIX and Win32 systems, so it's likely some kind of subtle build issue. My question is, how would I go about debugging this? The various options to jspc and rmic do not give me additional info, and the classes I'm operating on are being built (we run the compilers with about 5-10 classes as arguments). I'm assuming some of classes are not being parsed completely and stubs being made, but I can't figure out which ones, short of decompiling the classes and comparing with my code.
Anybody share a clue? Is there a better way to get inside the jspc, rmic processes to find out what is going on?
-- jdv