Re: [Swarm-Support] GDB backtrace

[Marcus G. Daniels]

Crile Doscher wrote:

>  Could anyone help me decipher this backtrace? 

The backtrace isn't helpful, obviously.  Usually an incremental approach 
of breakpoints in GDB or direct instrumentation in your model is the way 
to go.  For example, you can put in diagnostics like printf("Made it to 
checkpoint #1");  in your code (checkpoint #2,3,4,[etc]).  Then you can 
interpolate between the checkpoints by scrutinizing your code to posit 
places that could fail due to things like memory management mistakes 
(running over the end of arrays, using uninizalized object handles, 
etc).     One thing to note with GDB is that if your model is being 
compiled with -O2 (the default), you can get counterintuitive results on 
GDB's perception of where line numbers are.   To really use GDB for high 
level debugging, you'll want to compile with -O0.  Frankly, since most 
Swarm models are not large, it is easy to recompile them quickly, and 
thus using manual diagnostics is reliable and pretty fast.