--- Peace Jerome <peace_the_dreamer at yahoo.com> wrote:

> --- ned at squeakland.org wrote:
> > 
> > "Change Set:		SecondClickTimeoutFix-nk
> > Date:			26 July 2004
> > Author:			Ned Konz
> > 
> > Fixes two problems in MouseClickState:
> > - down/up/down/wait would generate an incorrect
> > doubleClick.
> > - down/up/down/move would not generate startDrag
> > 
> > Also corrects and clarifies the
> > MouseClickState-related comments in
> > HandMorph.
> > 
> > Also corrects the balloonText instructions in
> > DoubleClickExample.
> > "
> > 
> Thanks for taking the time to respond. I tried out
> your fixes and things seem to be better. I'm still
> looking over the code trying to puzzle out all that
> it
> does and all its ramifications. I just realized
> MouseClickState>>handleEvent:from: is recursive in
> that it sends HandMorph>>handleEvent which has a
> sent
> to  ....  I haven't gone thru the logic but I bet it
> would be interesting to see a history of what 
> events
> pass thru ClickState. 
> 
> Cheers and joy, --Jer 
> 

Well, I finished diagraming the logic yesterday. And 
it sure looked like clickstate was calling itself
recursively with a second mouse up. Checking
eventstats showed that mouseups were exceding mouse
downs when DoubleClickExample was doodled and not very
much otherwise.

So I wrote a investigator subclass to show the event
history and got these results:

2: a WizClickState[#secondClickDown]
		clickClient: a DoubleClickExample<DoubleClick>(2653)
		clickState: #secondClickDown
		firstClickDown: [869 at 245 mouseDown red 2333079]
		firstClickUp: [869 at 245 mouseUp 2333169]
		firstClickTime: 2333079
		clickSelector: #click:
		dblClickSelector: #doubleClick:
		dblClickTime: 350
		dblClickTimeoutSelector: #doubleClickTimeout:
		dragSelector: #startDrag:
		dragThreshold: 10
		history: an OrderedCollection()
			1: [869 at 245 869 at 245 mouseMove red 2333097]
			2: [869 at 245 869 at 245 mouseMove red 2333117]
			3: [869 at 245 869 at 245 mouseMove red 2333137]
			4: [869 at 245 869 at 245 mouseMove red 2333157]

			5: [869 at 245 mouseUp 2333169]
			6: [869 at 245 mouseUp 2333169]

			7: [869 at 245 mouseDown red 2333191]
			8: [869 at 245 869 at 245 mouseMove red 2333237]
			9: [869 at 245 869 at 245 mouseMove red 2333257]
			10: [869 at 245 869 at 245 mouseMove red 2333277]
			11: [869 at 245 869 at 245 mouseMove red 2333297]
			12: [869 at 245 mouseUp 2333301]

history is all the events going thru one clickstate
from inseption til reset.  I've isolated the
duplicated mouse down caused by the recursive call to
handleEvent:.

I've won my first bet. My second is that this is a
significant bug explaining many other symptoms.

Attached is a changeset with the subclass to collect
the history and clickstates.  The above example is
taken from a fresh image with just your changes added
(and the look tools of course.)

The tools don't depend on specific methods in
MouseClickState. So the can be added to before or
after your most recent fixes. 

I didn't have time to house train them completely. I
was more interested in getting the debugging info.
To turn them off completely revert the method that
subsitutes the subclass for MouseClickState.


'From Squeak3.8alpha of ''17 July 2004'' [latest
update: #5976] on 29 July 2004 at 10:30:54 am'!
"Change Set:		ClickTestTools-wiz
Date:			29 July 2004
Author:			Jerome Peace (wiz)

Third attempt. wiz 7/29/2004 10:29
Added a size limit to PastHistory. No sense causing
out of space crashes.

Second attempt at ClickTestTools. 

This is a first pass at tools to examine 
MouseClickState.

I made a nosy subclass of MouseClickState which copys
the events as they are sent to a history collection.

The state's objects themselves are preserved in a
classVar ordered collection with the idea of them
being explored and reset.

Other than that I modified Double Cllick Example to
implement the timeout selector. It cycles trough
several border sizes.

And created sub instances to test varioius
combinations of selectors since the behavior of
clickstate differs depending on their presence or not.

HandMorph class was modified to give access to the
eventStats that handleEvent stores.

HandMorph was modified to use and preserve my
clickstate subclass.
"!

Gzip cs (to which the above is a preamble) attached.


Yours in Service,  Jerome Peace



	
		
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - 100MB free storage!
http://promotions.yahoo.com/new_mail 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ClickTestTools-wiz.2.cs.gz
Type: application/x-gzip
Size: 5082 bytes
Desc: ClickTestTools-wiz.2.cs.gz
Url : http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20040729/27b1c132/ClickTestTools-wiz.2.cs.bin