[Bug 740] Changed - divide by zero exception in x11drv.dll

http://bugs.winehq.com/show_bug.cgi?id=740

*** shadow/740	Fri May 31 13:26:39 2002
--- shadow/740.tmp.7952	Fri May 31 13:26:50 2002
***************
*** 251,253 ****
--- 251,300 ----

------- Additional Comments From [e-mail]  2002-05-31 13:26 -------
  Adding Andriy to CC's
+ 
+ ------- Additional Comments From [e-mail]  2002-05-31 13:26 -------
+ ok, I have some info on the problem...
+ 
+ here is basically what happens:
+ 
+ fgBits gets initialized ;)
+ fgBits gets set to zero
+ then things screw up...
+ 
+ but you knew that I know, just keep reading...
+ 
+ 2 lines below where fgBits gets set to zero, xmax gets set to whatever, 
+ possibly zero, same with ymax...
+ 
+ then we run a for on y=0 and compare it to ymax
+ inside there we run a for on x=0 and compare it to xmax...
+ then theres a switch that may be important to this bug
+ and then the crucial if (red+green+blue > threshold)
+ which is the only place that fgBits gets incremented before that divide by zero 
+ error, so obviously 1 of 3 things is happening:
+ 
+ 1) ymax is being set to zero and so the for fails
+ 2) xmax is being set to zero and so the for fails
+ 3) red+green+blue is not being set high enough to pass the threshold if
+ 
+ and so fgBits++ isn't taking place...
+ 
+ now that I have looked back up at the switch, I see that lo and behold:
+ 
+ theChar = theImage[byteIndex++];
+ red = theChar;
+ 
+ and like david said above, the whole theImage array is failing...
+ 
+ now, since I don't have access to a Linux box ATM (i have to go home to access 
+ it) I can't add my own trace messages to see what is causing theImage to fail, 
+ but it looks like either
+ 
+ 1) byteIndex++ isn't working for some really weird reason
+ 2) byteIndex++ is working, but theImage[1] doesn't exist
+ 3) &theMask[ptr->nWidth/8 * ptr->nHeight]; isn't working...
+ 
+ So... I will add my stuff tonight and try to post the results tomorrow....
+ 
+ all of this information is against the current mouse.c, not rev 1.2 which is 
+ the first version with this problem...
\ No newline at end of file

[David Ronis]

As I reported, I put in some trace's in the innermost x & y loops.
The problem is that the entire theImage is filled with zeros. Both
xmax and ymax are 32 and I can't belive the compiler is so broken that
the ++ operator fails; if so, we have more problems here, ground
control.

So only your last possibility seems reasonable.  On the other hand,
why can't a cursor mask be all zero's?

[e-mail] writes:
> http://bugs.winehq.com/show_bug.cgi?id=740
>
> *** shadow/740	Fri May 31 13:26:39 2002
> --- shadow/740.tmp.7952	Fri May 31 13:26:50 2002
>
> *** 251,253 ****
> --- 251,300 ----
>
> ------- Additional Comments From [e-mail]  2002-05-31 13:26 -------
>   Adding Andriy to CC's
> -
> - ------- Additional Comments From [e-mail]  2002-05-31 13:26 -------
> - ok, I have some info on the problem...
> -
> - here is basically what happens:
> -
> - fgBits gets initialized ;)
> - fgBits gets set to zero
> - then things screw up...
> -
> - but you knew that I know, just keep reading...
> -
> - 2 lines below where fgBits gets set to zero, xmax gets set to whatever,
> - possibly zero, same with ymax...
> -
> - then we run a for on y=0 and compare it to ymax
> - inside there we run a for on x=0 and compare it to xmax...
> - then theres a switch that may be important to this bug
> - and then the crucial if (red+green+blue > threshold)
> - which is the only place that fgBits gets incremented before that divide by zero
> - error, so obviously 1 of 3 things is happening:
> -
> -
> - ymax is being set to zero and so the for fails
> -
> - xmax is being set to zero and so the for fails
> -
> - red+green+blue is not being set high enough to pass the threshold if
> -
> - and so fgBits++ isn't taking place...
> -
> - now that I have looked back up at the switch, I see that lo and behold:
> -
> - theChar = theImage[byteIndex++];
> - red = theChar;
> -
> - and like david said above, the whole theImage array is failing...
> -
> - now, since I don't have access to a Linux box ATM (i have to go home to access
> - it) I can't add my own trace messages to see what is causing theImage to fail,
> - but it looks like either
> -
> -
> - byteIndex++ isn't working for some really weird reason
> -
> - byteIndex++ is working, but theImage[1] doesn't exist
> -
> - &theMask[ptr->nWidth/8 * ptr->nHeight]; isn't working...
> -
> - So... I will add my stuff tonight and try to post the results tomorrow....
> -
> - all of this information is against the current mouse.c, not rev 1.2 which is
> - the first version with this problem...
>
> \ No newline at end of file