Path: utzoo!attcan!uunet!ginosko!gem.mps.ohio-state.edu!tut.cis.ohio-state.edu!ucbvax!hplabs!hplabsz!mayer
From: mayer@hplabsz.HPL.HP.COM (Niels Mayer)
Newsgroups: comp.windows.x
Subject: Re: Parallel processes to X ?
Message-ID: <3839@hplabsz.HPL.HP.COM>
Date: 23 Aug 89 08:00:39 GMT
References: <140@elctr.UUCP>
Reply-To: mayer@hplabs.hp.com (Niels Mayer)
Organization: Hewlett-Packard Labs, Software Technology Lab, Palo Alto, CA.
Lines: 48
Summary:
Expires:
Sender:
Followup-To:

In article <140@elctr.UUCP> hbr@elctr.UUCP (Hans B| Randgaard) writes:
>Are there some way to invoke background processes still using X to
>control everything, but without getting a copy of the widgets ?

Xt programming is event-driven programming, so why not make your
background process send it's results back to your application by
running it in a subshell (see popen()) and having any output from the
pipe get processed as an "event" via XtAddInput().  If you don't like
it running in a subshell, make a server/client interface. Take a look
at xwebster 1.0 (from x11r3 contrib tape) for an example of this
approach. 

>However the analysis is invoked by pushing a command-button and it
>can take quite some time (several minutes) before the analysis is
>through.
>What I want to do is to make the analysis run as a parallel process
>so that the user can make other things(maybe kill the analysis
>clicking another button) during the analysis. 

You can certainly do that with the above approach.

Alternately, you can turn things upside down and do event processing
once per n iterations of your analysis loop:
  do {
    sprintf(temptext, "Trying webster server at address %s ...\n", host_addr);
    Display_Def_Text_Append(temptext);
    Display_Def_Refresh();
    XSync(display, FALSE);
    while (XtPending()) {	/* since we can't get back to XtMainLoop yet-*/
      XEvent event;		/*-we process the events here... (yech) */
      XtNextEvent(&event);
      XtDispatchEvent(&event);
    }
	/* iteration of time consuming operation */
	...
  } while (blablabla)...

The above comes from xwebster 1.1 (on expo.lcs.mit.edu's contrib ftp
dir) and I don't know what that XSync() is doing there... If I
remember, nested event loops require r3 Xt.

>I have made several
>fork experiments, but I always(of course) get a copy of the whole
>widget tree, which I certainly don't need or want. 

Scary!

-- Niels "don't ask me about webster servers" Mayer.