Re: debugging kse

On Sat, 27 Dec 2003, Alfred Perlstein wrote:

> * Daniel Eischen <eischen_at_vigrid.com> [031227 22:41] wrote:
> > On Fri, 26 Dec 2003, Alfred Perlstein wrote:
> > 
> > > I have a copy of webstone that will use kse available here:
> > > http://people.freebsd.org/~alfred/webstone/
> > 
> > Here's a diff to eliminate TLS usage in webstone.  It can
> > be cleaned up further, and I can't see any reason for the
> > use of TLS.  Operations on FILE should be thread-safe and
> > the remaining uses of TLS were mostly for temporary string
> > manipulations.
> > 
> >   http://people.freebsd.org/~deischen/webstone.diffs
> > 
> > I haven't enabled threads, but it should be simple enough.
> > Note that you want the parent thread to call alarm(), not
> > each child thread (client).  I'd have the parent thread use
> > sigwait() to catch the signal, set the flag for the other
> > clients to see, then join to the client threads.  I don't
> > see any need for synchronization objects/semaphores.
> 
> heh!
> 
> A bit ahead of you over here...
> http://people.freebsd.org/~alfred/webstone/

I don't see why you want to emulate the windows API when it is
much simpler to add in the necessary pthread bits when threads
are enabled.  I updated my patch to add in pthread support
for webclient (see patch-ad):

  http://people.freebsd.org/~deischen/webstone.diffs

I didn't add any of the necessary configure stuff that is
in your patch.  I also didn't add a synchronization point
to start the client threads off and running.  The fork()
method didn't use any so I didn't see why using threads
would be any different.  I did join to the threads at
the end, but even that isn't necessary since they will
continue to run after the main thread ends (until the
alarm goes off).

I'm still not sure how to test this.  Do I need a
web server?

-- 
Dan Eischen