Re: Re: [Prothon-commits] rev 149 - in trunk:include/prothon src

Are you talking about this line?

pr_lock(&thread_registry_lock); pr_unlock(&thread_registry_lock);

I unlock it on the same line.  I guess this is a good argument for never
putting two statements on the same line.

----- Original Message ----- 
From: "Ben Collins" <[email protected]>
To: "Mark Hahn" <[email protected]>
Cc: <[email protected]>
Sent: Saturday, March 27, 2004 7:40 PM
Subject: Re: [Prothon-dev] Re: [Prothon-commits] rev 149 - in
trunk:include/prothon src


> > > + if (main_argv_obj && list_len(ist, main_argv_obj)) {
> > > + new_thread_obj((apr_thread_start_t)main_thread, main_argv_obj);
> > > + /* Grab this lock, so we know main1 thread has started */
> > > + pr_lock(&thread_registry_lock); pr_unlock(&thread_registry_lock);
> > >
> > > + }
> > > +
> > >  threads = get_attr(ist, sys_argv_obj, sym(ist, "threads"));
> > >  if (threads) {
> > >  int i;
> > >  for (i = 0; i < list_len(ist, threads); i++)
> > > - start_new_thread((apr_thread_start_t)main_thread, list_item(ist,
> > threads, i));
> > > + new_thread_obj((apr_thread_start_t)main_thread, list_item(ist,
> > threads, i));
> > >  }
>
> After you create the first new_thread_obk() for main args, you grab the
> lock to make sure it completes startup. However you never release it
> before calling new_thread_obj() again for the -t options. So the first
> time you call new_thread_obj() for one, it will deadlock waiting for
> thread_registry_lock to be released.
>
> You should only grab that lock once you have started all the threads you
> want to, and then release it.
>
> -- 
> Debian     - http://www.debian.org/
> Linux 1394 - http://www.linux1394.org/
> Subversion - http://subversion.tigris.org/
> WatchGuard - http://www.watchguard.com/
>
>
>