Re: Allowing Applications to continue after exception...

[Richard Frith-Macdonald]

On 5 Feb 2009, at 05:50, David Ayers wrote:
> http://lists.gnu.org/archive/html/discuss-gnustep/2004-10/ 
> msg00092.html
>
> I still believe that handling generic handling of exceptions in the
> runloop is a dangerously wrong and an implementation detail that we
> shouldn't try to be compatible with.  But others may disagree.

Well I agree that it's obviously dangerously wrong ... but I'm not  
sure that means we shouldn't do it anyway.
Pragmatically, it seems like the vast majority of exceptions do not  
indicate anything very harmful, so continuing is not a problem for  
those cases, so compatibility might generally make sense.

What we could do is make the behavior selectable (with a user default  
for instance), and use it as a selling points that we can switch  
applications over to a 'secure' mode where they won't screw up like  
Apple ones might.

One more thing to note ... it's not as simple as just looking at the  
case of catching exceptions in the application's event loop.  GNUstep- 
base is already broken in this way in that it's Apple compatible in  
catching/logging exceptions inside timers and notification handlers  
(possibly other places too I guess).  This has actually bitten me  
once ... where an error in a timer meant that we got a huge logfile  
generated, filling the disc and bringing an entire server down because  
the process kept running and our monitoring software therefore didn't  
alert us of the error.  In this case, not being Apple compatible would  
have saved us half an hour of downtime, which is important when you  
are running 24x7 servers.