On Tue, 13 Jul 1999, Waldo Bastian wrote:

> Steffen Hansen wrote:
> 
> > It seems that SIGABRT isn't catchable (I thought is was). Often
> > kded crashes are caused by a failed assert(bla); the result is 
> > that the X atom stays after kded is died.
> > 
> > How do we solve this problem? Not using assert() isn't a fix.
> > The problem remains with other uncatchable signals (SIGKILL etc.)
> 
> What about catching the exception mico appearantly throws?
> 
> The problem is not kded. It is konqy which doesn't handle a dying
> kded.

Hmmmmmmm :-)

Well, actually this would mean we'd have to try{} *every* invokation.

But fortunately we hide all CORBA stuff behind our C++ interfaces, so we
*know* where the invokations to kded are done, in libkded. We wouldn't
even need to hack any client code (Konqy, etc) .

So we need a kind of restart mechanism in KdedInstance and we need to put
catch{}'es in all the KRemoteFoo classes where calls to kded are done.

I think this could work. Well done Waldo! :-)

Steffen - can you provide such a restart-thingy ?

I'll then look into placing the try{}catch{}'es to the appropriate places.

Greetings,
 Simon