Re: building from cvs

[Michael Koch <konqueror at gmx dot de>]

On Sat, Jan 24, 2004 at 10:45:37PM +0100, Jakob Praher wrote:
> hi all,
> 
> first of all kudos to you all. You are doing a great job.
> 
> I wanted to test some of the newer APIs in libgcj and so I checked out
> the gcc source and did a fresh compile. (For instance the
> ByteBuffer.allocateDirect(I) )
> 
> After some problems, I managed to compile the whole thing and install it
> into a custome prefix $HOME/gcc.
> 
> I sourced my gcc prefix (PATH,LD_LIBRARY_PATH) and compiled a simple
> Test.java. 
> Now, no matter which Java code, I compile, after trying to run the
> binary I get a sigsegv. Gdb told me that the natClass.cc tries to access
> a apprantly Null pointer in :
> 
> _Jv_linkExceptionClassTable(java::lang::Class*) 
> 
> struct _Jv_CatchClass *catch_record = self->catch_classes;
>   if (!catch_record || catch_record->classname)
>     return;  
> 
> the upper to calls on the call stacks are:
> 
> #0  0x4044e484 in _Jv_linkExceptionClassTable(java::lang::Class*) (
>     self=0x407b40e0) at java/lang/natClass.cc:1769
> #1  0x4044caad in java::lang::Class::initializeClass() (this=0x407b40e0)
>     at java/lang/natClass.cc:800
> 
> so did I screw anything up, when compiling gcj?
> I did some research, but after all I think the catch_classes struct is
> filled in during code generatation by the gcj compiler and, apparantly
> it should never be null, or otherwise there would have been a check in 
> _Jv_linkExceptionClassTable, right?

What platform are you working on ? x86 ?
What configure options have you used ? 
Perhaps you got just a bad state from cvs. Have you tried updating ?

I compile and use gcc CVS trunk on a dialy basis on x86 since a long
time.


Michael