PC updates while interpreting bytecode

[Kyle Galloway <kgallowa at redhat dot com>]

Hi all,

I was wondering about updates to the pc while interpreting bytecodes.  
I  know that the actual pc is a local variable to the particular 
instance of _Jv_InterpMethod::run (or run_debug), but I was wondering 
how often frame_desc.pc is updated.  The reason I ask is that to get a 
stack trace at an arbitrary point in time, it is necessary for the pc 
inside the frame structure to be correct in order for the location 
reported to be correct.  this is particularly important when single 
stepping through code in between method calls, where currently, though 
the VM is stepping to the right place, the pc in the frame info is not 
updated which causes eclipse to incorrectly display the source line 
currently executing at some points.

I have thought of a couple of solutions to this, and I was wondering 
what opinions there are on this.  The first is to update the pc inside 
NEXT_INSN in the debug case, though I think this is still less that 
ideal since there is a race there if the thread is suspended after the 
local pc has changed but before the frame info pc is updated.  The 
second is to pin the PC with a pointer inside the frame structure in the 
debug case like this:

pc_t pc;
frame_desc.pc_ptr = &pc;

I would assume it would be preferable in this case to maintain the 
current pc and to access the "real" pc through this pointer, which would 
be only a small change to the current stack tracing code.  I tend to 
think this would be the preferable solution, but I thought I would get 
input from people far wiser than I am in this department.

Thanks,

Kyle