[asterisk-dev] policy change for correct handling of fatal program conditions

[SF Markus Elfring]

> Well, if not, then all calls fail in their own due time.  This is still the
> "right thing" to do, and you have not made an argument for why the process
> would do better to call exit().

The function "set_var" shows a simple example. It tries to add a 
variable to a global list in a thread-unsafe way. How important is this 
operation for correct program behaviour?
http://svn.digium.com/view/asterisk/trunk/utils/check_expr.c?rev=47232&view=markup

A memory allocation failure will be silently ignored there because the 
current function interface has got no means in the C programming 
language to report this error back to its caller. A C++ implementation 
would have got the opportunity to throw an exception. How many efforts 
would you like to invest to block or defer process termination?

Regards,
Markus