On 11/24/2015 11:14 PM, Paul Eggert wrote:
> Daniel Colascione wrote:
>> C provides many error checking alternatives. Longjmping outside Emacs is
>> not one of them.
> 
> Surely there is a reasonably natural way to address the problem without
> requiring module authors to call longjmp, or to have a complex and
> error-prone check after every call.

How is it any worse than checking whether open(2) returns -1? I'd be
fine with just making future calls fail until a module clears the
pending condition or returns to Emacs.