Re: [Swarm-Support] possible memory problem on mac g5

[Bill Northcott]

On 06/04/2004, at 2:24 AM, Paul Box wrote:
>  I have just set up a G5 mac, and installed swarm (Thanks again 
> Bill!).  I
> have tried running some of my favorite swarm programs, and found that 
> I am
> getting segmentation faults when too many agents are added to the 
> system.  The
> problem originally seemed to be memory: the program runs and eventually
> crashes with the following message:
>
> Program received signal EXC_BAD_ACCESS, Could not access memory.
> objc_msg_lookup (receiver=0x29007265, op=0xf860) at
> ../../../libobjc/objc/sarray.h:232
> 232     ../../../libobjc/objc/sarray.h: No such file or directory.
>
> A backtrace gives the following:
>
> (gdb) bt
> #0  objc_msg_lookup (receiver=0x29007265, op=0xf860) at
> ../../../libobjc/objc/sarray.h:232
> #1  0x0095691c in objc_msg_lookup (receiver=0x29007265, op=0xf860) at
> ../../../libobjc/sendmsg.c:150
> #2  0x0000c024 in -[Plant colonize].48 (self=0x11800a48, _cmd=0xf860) 
> at
> Plant.m:107
> #3  0x0000bf68 in -[Plant grow].47 (self=0xf860, _cmd=0xf860) at 
> Plant.m:77
> #4  0x00004978 in -[Cell step].45 (self=0x29007265, _cmd=0xf860) at 
> Cell.m:73
> #5  0x0099586c in -[Collection_any forEach:].57 (self=0xb36fec, 
> _cmd=0xf860,
> aSelector=0xf464) at ../../../../src/collections/Collection.m:216

Looking at that backtrace, in frame 3 'self' should obviously be a 32 
bit pointer.  However, it is probably the debugger that is confused 
rather than the program.  To get meaningful arguments in the backtrace, 
you need to compile with -g -O0.  This will probably mean altering the 
values of 'CC' and 'OCBJC' in $SWARMHOME/etc/swarm/makefile.common and 
possibly swarm.config in the same directory.

Bill Northcott