Re: Forcing Exception Handling

["Randy Brukardt" <randy@rrsoftware.com>]

"iloAda" <egzgheib@gmail.com> wrote in message 
news:91988389-b17a-42cb-8f5b-9b1205604b55@y3g2000vbh.googlegroups.com...
...
>exception
> When others =>
>     --  Do something that will allow the system to keep on running
>
>have u guys done something like that before?

Sure, for the spam filter and web server, each task has a handler like this 
at the outer level to log failures; then return a failure code to the 
caller. (For the web server, this returns a 503 code to the requester; for 
the spam filter, this causes the message to be quarantined for later 
hand-analysis -- and fixing of the bug that caused the crash.) This is done 
in a block inside of a loop so that the (worker) task still continues to be 
usable for processing other requests/messages.

Both of these are "real-time" systems in that not answering a web request in 
a reasonable causes the requestor to get nothing (which is bad) or causes 
e-mail to get lost (which is bad). But the timing constraints are fairly 
loose; we have seconds (not milliseconds) to handle the problem.

For the Janus/Ada compiler and tools, we used to have such handlers, but 
they covered up errors so well that we got rid of them. It's better for the 
compiler to just crash outright because that makes customers call us with a 
report right away rather than trying to figure out what went wrong. (And 
often we can figure out the bug just from the default error walkback and a 
bit of poking in the source code -- saves lots of debugging time.)

                               Randy.