Well, I was going to try and have a look at this over the weekend.  
Unfortunately, NetBSD/arm32 just isn't up to debugging an application as 
large as this (groan).

I started by trying to build a debug version of Mozilla; this failed 
because something in the link needed more than 8M of string space, and 
since the strings were being put on the stack using alloca it tripped the 
hard kernel limit for my shark (I've got a patch for ld that I will PR 
later today).

Then linking crashed my shark (wedged solid, keyboard not responding, so 
couldn't get into the debugger to find out where).  Rebooted and tried 
again, this time linking successful.

Finally, it is impossible to debug this on a 32M shark; gdb is 
continuously being fobbed off by the kernel ptrace call with "Operation 
not permitted" whenever the executable hits a breakpoint.  This means that 
it cannot even get enough information to report the registers or stack 
frame.

Sorry, I can't do any more on this until ptrace is fixed.   *Please* could 
someone who understands the VM system (I think that is the guilty party) 
have a look to find out why it keeps throwing out ptrace calls to legal 
addresses?

R.