Re: Exception Handling and _NS_COUNT_HACK

[Richard Frith-Macdonald]

On 28 Oct 2008, at 02:23, Samantha Rahn wrote:

> Very helpful! Thanks! I am learning about the handle <signal>
> [keywords] commands of GDB. Much appreciated. However, in Excel, the
> SIGSEGV *does* get sent after a handful of _NS_COUNT_HACK calls, but
> it does *not* result in recover() getting called. Any ideas there?

It occurs to me that it's just possible that you have the same problem  
as the bug report at https://savannah.gnu.org/bugs/index.php?24526

While the actual bug in this case is really unconfirmed (it hasn't  
been reproduced with current code afaik), it does sound like a case of  
the signal handler not getting called somehow.  My only guess is that  
some other thread could have changed the signal handler in the window  
between the code setting it and the signal being generated.  However,  
you might want to look at the code and see if you can spot some other  
possibility that I've missed.

Maybe the simplest thing to do right now would be to add a configure  
option to disable stack traces, so that anyone whose gcc has this bug  
in stackframe checking can stop it being used at all.