[Tcl-bugs] [ tktoolkit-Bugs-720350 ] Memory leak or ? in canvas, WinXP Home

[SourceForge.net <no...@so...>]

Bugs item #720350, was opened at 2003-04-12 12:28
Message generated for change (Comment added) made by mrfitz
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112997&aid=720350&group_id=12997

Category: 04. Canvas Basics
Group: 8.4.2
Status: Open
Resolution: Works For Me
Priority: 5
Submitted By: Mike Fitzgibbon (mrfitz)
Assigned to: Jeffrey Hobbs (hobbs)
Summary: Memory leak or ? in canvas, WinXP Home

Initial Comment:
ActiveState ActiveTcl 8.4.2 (fresh install, also seen
with previous version, 8.4.0.1?)
WinXP Home, ver 2002, sp1
Mobile Athlon XP 1500+, 1.33 GHz
240 MB Ram

Taking the Colliding Balls script from the wiki, 

http://wiki.tcl.tk/8573

I appended the following code to add and delete balls
on button presses:

# Maximum ball number
set State(ballCount) [llength $State(ballList)]

# Create a new ball on right-click
bind $State(canvas) <Button-3> {
	incr State(ballCount)
	set ball ball$State(ballCount)
	createBall $ball
	lappend State(ballList) $ball
}

# Delete the closest ball on left-click
bind $State(canvas) <Button-1> {
	set id [$State(canvas) find closest %x %y]
	set ball [lindex [$State(canvas) gettags $id] 0]
	set idx [lsearch $State(ballList) $ball]
	if { $idx >= 0 } {
		set State(ballList) \
                      [lreplace $State(ballList) $idx $idx]
		$State(canvas) delete $id
	}
}

I also added the following to the end of proc moveBalls
{ } :

	# Remove slow balls
     foreach ball $State(ballList) {
         foreach {xvel yvel} $State(vel,$ball) {}
         set vel [expr {sqrt( $xvel*$xvel + $yvel*$yvel )}]
         if {$vel < 0.5} {
             set id [$State(canvas) find withtag $ball]
             set idx [lsearch $State(ballList) $ball]
             set State(ballList) [lreplace
$State(ballList) $idx $idx]
             $State(canvas) delete $id
         }
     }


After a while of creating balls and letting the slow
ones be removed, the canvas froze and the wish84
application shows 99% CPU utilization and progressively
increasing mem usage until ~150MB then drops to between
~20-40 MB (fluctuating) while the commit charge
continues to increase.  Eventually a "Fatal Error in
Wish" box pops up stating "unable to alloc n bytes"
(where in my case n was ~100660000 first, 2400 second).  
The windows error the first time showed:

AppName: wish.exe	 AppVer: 8.4.2.2	 ModName: ntdll.dll
ModVer: 5.1.2600.1106	 Offset: 000267cd

appcompat.txt attached


----------------------------------------------------------------------

>Comment By: Mike Fitzgibbon (mrfitz)
Date: 2003-04-12 23:45

Message:
Logged In: YES 
user_id=121466

The code did not appear in the "existing files" section, or
show as attached in the response email, so I am trying to
upload it again.
(I forgot the checkbox, oops)

----------------------------------------------------------------------

Comment By: Mike Fitzgibbon (mrfitz)
Date: 2003-04-12 23:38

Message:
Logged In: YES 
user_id=121466

The addition to moveBalls was made just prior to the after line.
It is a recent addition, the problem occurred with the prior
TCL version
without it.  I just added it to help automate the process.
It takes awhile for the problem to manifest itself, ~25
minutes just now.
Whenever the number of balls dropped to 8, I added 8 more.
I am attaching the code I am running that exhibits the problem
It is named CollidingBalls.tcl

----------------------------------------------------------------------

Comment By: Jeffrey Hobbs (hobbs)
Date: 2003-04-12 13:39

Message:
Logged In: YES 
user_id=72656

Attached is my interpretation of what the user was using.  I 
do not have problems on a P4/2.4ghz machine running 
Win2K and 8.4.2.  Memory does climb, CPU usage is usually 
under 10% unless I have more than 20 balls on the screen.

I could see a possible problem that if code were added after 
the after and took too long, you would end up with multiple 
calls in moveBalls that could create eventually constant 
moveBalls calls, but I couldn't simulate the problem.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112997&aid=720350&group_id=12997