bugs in stackTrace.c

[Godmar Back]

> 
> Sorry to say, but you were wrong :)
> 

 You're right.  I just tried starting kaffe with a non-existing
class and it said ClassNotFoundException, just like it should, so I wondered.
I think that's because the seg violation is caught, handled, and
ignored somehow.  One can see the failure by running the interpreter with
-vmdebug EXCEPTION or under gdb.

I assume you mean STACKTRACEMETH(trace)!=1 in the code below?

	- Godmar

> if the top stackframe contains a meth = (Method*)1 then Tnext is never
> called. A possible patch would be to have STACKTRACEINIT() loop until it
> has found a valid frame. Or another one would be to have buildStackTrace
> look like this instead:
> 
> 	STACKTRACEINIT(trace, base, base);
> 	cnt = 0;
> 	while(!STACKTRACEEND(trace)) {  /* <<<--- new code */
> 		if(STACKTRACEMETH(trace)!=-1)
> 			break;
> 		STACKTRACESTEP(trace);
> 	}
> 
> 	while(!STACKTRACEEND(trace)) {
> 		STACKTRACESTEP(trace);
> 		cnt++;
> 	}				  /* <<<--- end new code */
> 
> 	/* Build an array of stackTraceInfo */
> 	info = gc_malloc(sizeof(stackTraceInfo) * (cnt+1), GC_ALLOC_NOWALK);
> 
> 	cnt = 0;
> 
> 	STACKTRACEINIT(trace,base,base);
> 	while(!STACKTRACEEND(trace)) {	/* <<<--- new code */
> 		if(STACKTRACEMETH(trace)!=-1)
> 			break;
> 		STACKTRACESTEP(trace);
> 	}  				/* <<<--- new code */
> 
> 	for(; !STACKTRACEEND(trace); STACKTRACESTEP(trace)) {
> 		info[cnt].pc = STACKTRACEPC(trace);
> #if defined(INTERPRETER)
> 		info[cnt].meth = STACKTRACEMETH(trace);
> #endif
> #if defined(TRANSLATOR)
> 		info[cnt].meth = 0;	/* We do this lazily */
> #endif
> 		cnt++;
> 	}
> 
> 
> 
> regards, Stefan Burstroem
> 
> -----------------------------------------------------------------------
> >> Irl: Stefan Burstroem <<  >> Omnipresence Intl. <<  >> Irc: Yabba <<
>  >> Phone: +46 (0)46-211 40 84 << >> EMail: stefan at omnipresence.com <<
> -----------------------------------------------------------------------
> 
>