Fwd: kio: slave process dies when using https

[Dawit A.]

On Sunday 09 June 2002 12:24, Stephan Kulow wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On Freitag, 7. Juni 2002 21:21, Waldo Bastian wrote:
> > The problem is that we shouldn't die/exit on a SIGPIPE. However, we do
> > exactly that since the following change to slavebase.cpp
> >
> > revision 1.114
> > date: 2002/02/28 13:35:53;  author: coolo;  state: Exp;  lines: +3 -2
> > Don't endless loop in sigpipe_handler if an application is killed having
> > a download in progress
> > -    signal(SIGPIPE,&sigpipe_handler);
> > +    exit(0);
> > +    // signal(SIGPIPE,&sigpipe_handler);
> >
> > Maybe coolo can comment why that change was necassery and why the endless
> > looping couldn't be fixed with some error checking in the slave itself.
> > Otherwise I will revert that change.
>
> Well, just reverting doesn't help. If the slave exists or idles endlessly
> in sigpipe_handler doesn't really mater. Just that the later causes a lot
> more CPU usages and doesn't give the user any feedback.

But that would only happen if the io-slave simply tries to retransmit the data 
w/o checking the result of doing an io-operation, doesn't it ?  Otherwise it 
would simply know that the current socket is invalid and create a new 
connection to the given site like we do in kio_http.

Dawit A.