[M3devel] need help with cm3 problem before I deliver software this week

[Randy Coleburn]

Jay:
 
I have checked my Reactor v4.1 distribution and the problem also reproduces there as well, so this is not something new, its been around awhile.  I guess my memory must be failing that I don't remember this bug.  In any event, my customer is not happy, so I have to try and fix this and the pixmap problem.
 
I've returned back home from the customer site.  They have given me two weeks to fix these problems and to make some enhancements they want in the product before it is released to the end users.  So, that is my time interval--I've got 2 weeks max to fix everything.  The customer is not going to accept the product if I can't fix the cursor movement typein problem.  They don't like the pixmap problem, but since it is a non-functional issue, they will have to accept the defect if I can't fix it.  So, number one priority is to get the typein at cursor issue solved.
 
Note that this problem seems to reproduce in all FormsVBT places where you can type text, even in numerics.  Thus, it is a real problem.  My product uses several windows that have fields where the user has to type to change values.  In every case, when you click the mouse and also use the arrow keys, the first character typed goes to the wrong place.  I guess I never observed this because I always clicked the mouse to the exact place I wanted and didn't use the arrow keys.
 
Jay, you mention something about changing the TextPort model and ask "is this actually your bug, or just indicative of it"?  I'm not sure exactly what you are asking me, so if you will clarify your question, I'll try to give a better answer.  I don't think switching the TextPortModel will have a bearing on the problem I am having.
 
I have to leave now for a family reunion event, so I won't have email access again until late Sunday.  When I get back, I'll try to send a short program to demonstrate the pixmap problem, but then I'm not sure you can reproduce it unless you can set your screen resolution to 1920x1200.  At a minimum, I'll try and capture some screen images to show the difference in the same window when run at the various resolutions.
 
Thanks for your assistance in helping track down this problem.
 
Regards,
Randy

>>> Jay <jayk123 at hotmail.com> 8/1/2008 11:28 AM >>>
Clarification: there are other changes.
There was a change in font stuff.
There was added batching of painting.
And more.
 
I'm still stumped. I have to be away from this for a few hours or all day.
I think seeing if 4.1 repros will be good, and then maybe widen the net on either a) what changed (not just trestle?) and b) debugging it.
Really it shouldn't be hard..to figure how out it decides where to place the character.
 
Oh, and save the text file, see if it is in memory and on disk where it is drawn on the screen.
That would be useful to know.
 
ok..the text matches the display.
 
I can roll Wintrestle.i3 back to cm3.6, still repros.
 
Ok, the reason it is platform specific is because of:
 
Searching for 'TextPortModel: TEXT := "'...
C:\dev2\cm3.2\m3-ui\vbtkit\src\vbtkitutils\POSIX\VBTKitEnv.i3(30):  TextPortModel: TEXT := "emacs";
C:\dev2\cm3.2\m3-ui\vbtkit\src\vbtkitutils\WIN32\VBTKitEnv.i3(30):  TextPortModel: TEXT := "mac";

If you change Windows the the emacs model, no repro.
So try Posix to mac model, hopefully it repros.
 
Ok, Randy, is this actually your bug, or just indicative of it?
 
 - Jay



From: jayk123 at hotmail.com 
To: rcoleburn at scires.com; rodney.bates at wichita.edu; dabenavidesd at yahoo.es 
Date: Fri, 1 Aug 2008 14:45:02 +0000
CC: m3devel at elegosoft.com 
Subject: Re: [M3devel] need help with cm3 problem before I deliver software this week

 > It works ok with 3.6..

 
a *quick* and *incomplete* survey of 3.6 vs. current shows:
  - deal with change in Text type (platform independent) 
  - handle WM_CHAR (windows specific) 
  - handle newlines differently (platform independly, except maybe the data varies per platform) 

 I tried undoing the last two, still no luck.
 The first is harder to undo, maybe "impossible", just have to read/debug it and decide it is ok or not..?

 Next easy step is probably to see if repros with 4.1,
  and then proceed with comparing the "closer" repro vs. non-repro.
  If it does not repro with 4.1, and newer releases are available
  between it and current -- e.g. 5.2 -- would be good to narrow
  down when it started happening and compare those source trees.

The code in question is stuff like mtext and textport.

There are functions named key or Key or mouse or Mouse. Those are also important to look at, I think.

Not being particularly close to figuring this out, consider ignoring everything I have said. :)

It is hard to pass up such a simple repro. :)
And the knowledge that 3.6 works, and really not much has changed in the intervening ~12 years..

Oh, and there are pixmap changes.
Randy, I glossed over the mail about pixmaps.
You have a simple repro of that?
You can try it with 3.6?

3.6 should still be readily downloadable from the web, and it really isn't
difficult to get it up and running, at least on Windows.
(I can't speak for Linux. I ran 3.6, albeit hardly at all, on Linux 1.2 I think it was,
and NT 3.51, maybe later NT and maybe Win9x)
(and 3.6 is liberally licensed, so anyone can send it around.
The notion that 5.1 was the open source release always sounded wrong...)

 - Jay





From: jayk123 at hotmail.com 
To: rcoleburn at scires.com; rodney.bates at wichita.edu; dabenavidesd at yahoo.es 
CC: m3devel at elegosoft.com 
Subject: RE: [M3devel] need help with cm3 problem before I deliver software this week
Date: Fri, 1 Aug 2008 14:06:45 +0000

It works ok with 3.6..


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://m3lists.elegosoft.com/pipermail/m3devel/attachments/20080801/811dfc2d/attachment-0002.html>