[udf-dlm-develop] core dump

[Harald F. <hf...@ap...>]

I though I should send this experience to the group, maybe it helps somebody who 
still gets segmentation fauls or core dumps.

Our experience here is that this is ALWAYS caused by mistakes by the programmer. 
What we did here is certainly very old-fashioned, but helped us to trace so far 
all problems. In the pidf there is one line like this

    int num_sensors = 171;                         $$ no. of sensors

We changed this number first to 5, working only with the first 5 sensors, where 
it is unlikely (though not impossible) that there is an error in there. Then we 
increased this number to 50, 100 and checked if we got our segmentation fault. 
After it happened at 100 we got back to 80 etc, until we came close to the bad 
sensor by 5. Then it was easy to find the wrong statement and it always turned 
out to be our mistake.

So, if somebody still fights with segmentation faults, try this method (I still 
learned it in the university as a method to determine the square root of any 
number. Do you remember these times?)

Harald


=========================================================
Harald U. Frey 
Space Sciences Lab	   phone: 510-643-3323
University of California   fax:   510-643-2624      
Berkeley, CA 94720-7450    email: hf...@ss...