[uClibc] kern.debug kernel: NWFPE: iwconfig[736] takes exception 00000002 at c014d5cc from 40054594

Ralph Siemsen ralphs at netwinder.org
Tue Jun 21 17:26:33 UTC 2005


John Clark wrote:
> 
> The philosophical question is: why this would not cause the same sort of 
> trap with glibc? 'fixing' the driver to return a non-zero
> value always, suggests that this bug will continue to exist because 
> someone didn't condition iwconfig to check for zero on
> return.

The correct behaviour AFAIK is that the userspace process should be 
terminated with SIGFPE.  That is what happens on x86 with glibc.  In 
this case iwconfig would exit immediately, unless it has installed a 
handler for this signal.  Might be worth checking if calls signal().

-R



More information about the uClibc mailing list