Jordan Brown (Sun) wrote:
> Roland Mainz wrote:
>> Erm... who or what defines "fatal" ? Even SIGSEGV can be caught within
>> applications (either using alternative stacks or a seperate thread
>> (assuming I don't mix-up things again...)) and is not fatal...
>
> I don't know for sure, but I think you just answered your own question.
> If caught, it's not fatal. If not caught, it's fatal.
That pretty much sums it up.
Remember also that only a signal originating from a process in
another contract will trigger the fatal signal logic.
> I assume that the decision is at its lowest level "are the low-order 8
> bits of the exit code that would be returned by wait(2) non-zero?".
> That is, did the process die as the result of an uncaught signal?
The signal delivery code generates the event after determining that:
a) the signal came from outside the contract, and
b) the process needs to die as a result of receiving the signal.
See the bottom of psig() in common/os/sig.c.
Dave
