Path: utzoo!attcan!uunet!seismo!dimacs.rutgers.edu!mips!sdd.hp.com!wuarchive!mit-eddie!bloom-beacon!eru!hagbard!sunic!news.funet.fi!hydra!serifos!havia
From: havia@serifos.Helsinki.FI (Jyrki Havia)
Newsgroups: comp.windows.x
Subject: Re: X11R4 xlock (SS1 SunOS 4.03 cc fix 1-14) doesn't work
Summary: A patch included
Message-ID: <8215@hydra.Helsinki.FI>
Date: 28 Oct 90 19:02:00 GMT
References: <TANA.90Oct24154056@lauda.ee.sophia.ac.jp>
Sender: news@cs.Helsinki.FI
Distribution: comp
Organization: University of Helsinki
Lines: 112


In article <TANA.90Oct24154056@lauda.ee.sophia.ac.jp> you write:
>
>I have installed X11R4 on the sparcstation 1 (Sun OS 4.03) patched
>with fix-1 to fix-14. Everything seems to work except for the client
>"xlock". It dies with the following message:
[removed... A VERY familiar message :-)]
>I noticed that both the xlock on X11R4 and R3 dies the same way (On
>the R4 server). Is this problem due to  some backward compatibility
>flag I set when compiling the whole system ? 

	No. This is due to a bug in R3 server, and R3 xlock. As the 
	bug get fixed, somebody forget to fix xlock, which 'used' 
	that bug. I have made a very simple patch for xlock.

	You can also set the R4 server to 'bug-to-bug' compatibility 
	mode with R3 server by executing command 'xset bc'.
 
>Also, is there an easy way to pin point the problem from an  error
>message such as the one I included above? I don't know much about
>programming in X.

	Well, as you have source code to look, the line 

  Major opcode of failed request:  26 (X_GrabPointer)

	in error message gives a hint which X protocol reguest failed.
	Here it is XGrabPointer(...) request. I first tracked the problem
	down to the server, and there is sequence (something like, we do 
	not have sources on-line) of code

	if (!bug_mode) {
		if (events BUTTON_PRESS or BUTTON_RELEASE are
		   requested explisitly) fail
	}

	So, I looked the xlock code, and there the 4th argument to 
	XGrabPointer is -1, ie. all bits on. THAT is a BUG! There 
	is (In Xlib or Xprotocol ref.) a statement like: 'When 
	grabbing the cursor, you should newer ask for Buttonxxxxxx 
	-events, as they are implicitly requested'. Someone has read 
	that (or it should have read from the beginning) 
	'.. must not ..' instead of '.. should not ..'.

	The patch I made includes the feature, that any movement of 
	mouse will bring the password-asking screen to life. I like 
	that, and some of my fellows don't like it. It is easy to 
	disbale by removign events PointerMotionMask and MotionNotify 
	from this patch.

----- Patch follows --------

 
*** xlock.c.orig	Mon Aug 20 16:02:22 1990
--- xlock.c	Mon Aug 20 16:34:08 1990
***************
*** 253,259 ****
  			   GrabModeAsync, GrabModeAsync, CurrentTime);
      if (status != GrabSuccess)
  	error("%s: couldn't grab keyboard! (%d)\n", status);
!     status = XGrabPointer(dsp, w, True, -1,
  			  GrabModeAsync, GrabModeAsync, None, mycursor,
  			  CurrentTime);
      if (status != GrabSuccess)
--- 253,259 ----
  			   GrabModeAsync, GrabModeAsync, CurrentTime);
      if (status != GrabSuccess)
  	error("%s: couldn't grab keyboard! (%d)\n", status);
!     status = XGrabPointer(dsp, w, True,  PointerMotionMask,
  			  GrabModeAsync, GrabModeAsync, None, mycursor,
  			  CurrentTime);
      if (status != GrabSuccess)
***************
*** 265,271 ****
  XChangeGrabbedCursor(cursor)
      Cursor	cursor;
  {
!     XGrabPointer(dsp, w, True, -1,
  		 GrabModeAsync, GrabModeAsync, None, cursor, CurrentTime);
  }
  
--- 265,271 ----
  XChangeGrabbedCursor(cursor)
      Cursor	cursor;
  {
!     XGrabPointer(dsp, w, True, PointerMotionMask,
  		 GrabModeAsync, GrabModeAsync, None, cursor, CurrentTime);
  }
  
***************
*** 483,489 ****
  		init(dsp, w, gc, color, inittime, count);
  	}
  	XNextEvent(dsp, &event);
!     } while (event.type != ButtonPress && event.type != KeyPress);
  }
  
  void
--- 483,491 ----
  		init(dsp, w, gc, color, inittime, count);
  	}
  	XNextEvent(dsp, &event);
!     } while (event.type != ButtonPress && 
!              event.type != KeyPress && 
!              event.type != MotionNotify);
  }
  
  void
----
Jyrki Havia                     : havia@cc.helsinki.fi
University of Helsinki,         : havia@finuha
Department of Computer Science  : HYLKA::HAVIA