Hello!

I have a reduced scene containing a height_field that would cause povwin
to crash (so I suppose, it's a bug).

I strongly suspect a division by zero error in the height_field code,
since the error can be avoided if you slightly change the perspective or
use Jitter when antialiasing output.

My setup:
Official Version 3.1g.watcom.win32
Win98 de 4.10.1998
Pentium II 300 MHz
64 MB RAM

Here is the scene:
------------start-----------
#declare Set_to_zero = 0.01;

camera
{
  location  -z*2
  direction z
  right     x
  up        y
  look_at   Set_to_zero
}

light_source
{
  <10, 4, -20>
  color rgb 1
}


height_field
{
  tga "hf.tga"
  texture {
    pigment {
      color rgb 1
    }
  }
  rotate x * 270
  translate <-0.5, -0.5, 0>
}
-------------end------------

INI and Command Line Options:
Original quickres.ini, 160x120, AA 0.3
-j


povray will crash if 1) the variable Set_to_zero is zero and 2) you
disable antialiasing jitter.

I posted the height_field tga (12 kB) to the binaries.images.group
(Xref: news.povray.org povray.binaries.images:16962, this message is a
reply to it)

/PETER/

-- 
Peter Santo (PUMP development)
Visit: http://www.mp3pump.de
linux quote of the week: 
It _is_ user-friendly.  It is not ignorant-friendly and idiot-friendly.

Post a reply to this message

Peter Santo wrote:
> 
> Hello!
> 
> I have a reduced scene containing a height_field that would cause povwin
> to crash (so I suppose, it's a bug).
> 
> I strongly suspect a division by zero error in the height_field code,
> since the error can be avoided if you slightly change the perspective or
> use Jitter when antialiasing output.
> 
> My setup:
> Official Version 3.1g.watcom.win32
> Win98 de 4.10.1998
> Pentium II 300 MHz
> 64 MB RAM

> povray will crash if 1) the variable Set_to_zero is zero and 2) you
> disable antialiasing jitter.
> 
> I posted the height_field tga (12 kB) to the binaries.images.group
> (Xref: news.povray.org povray.binaries.images:16962, this message is a
> reply to it)
> 
> /PETER/
> 
> --
> Peter Santo (PUMP development)
> Visit: http://www.mp3pump.de
> linux quote of the week:
> It _is_ user-friendly.  It is not ignorant-friendly and idiot-friendly.

I can verify that this does crash the render session at 24% completion in
a predictable manner. I am using the same build of Pov and windows 98 that
Peter is using on a p200 mmx system with 128 megs of ram.

Note: The image Peter used for the height field is an 8 bit compressed targa
file and I was also able to reproduce the fault with an 8 bit uncompressed
targa file. Also note that rendering at a resolution of 100x100 does not
cause it to crash but any higher resolution does.

-- 
Ken Tyler

Older   Links Page: http://home.pacbell.net/tylereng/links.htm
Updated Links Page: http://home.pacbell.net/tylereng/bkmrk999.htm

Post a reply to this message

On Sat, 07 Aug 1999 17:37:53 +0200, Peter Santo
<san### [at] poolinformatikrwth-aachende> wrote:

>Hello!
>
>I have a reduced scene containing a height_field that would cause povwin
>to crash (so I suppose, it's a bug).
>
>I strongly suspect a division by zero error in the height_field code,
>since the error can be avoided if you slightly change the perspective or
>use Jitter when antialiasing output.
>
>My setup:
>Official Version 3.1g.watcom.win32
>Win98 de 4.10.1998
>Pentium II 300 MHz
>64 MB RAM

I get this too.  As near as I can tell, it's crashing when the
intersection exactly hits the very edge of a triangle on the 
edge of the heightfield.  The line it crashes on says

      y3 = Get_Height(px,   pz+1, HField);

pz is 159, and HField->Data->max_z is 158.  Not a pretty picture.
I'm not sure what can be done about this, unfortunately, unless it's 
to subtract epsilon before rounding down to get pz.  Perhaps someone
else has an idea.

Post a reply to this message

Ron Parker wrote <37acc879.22494197@news.povray.org>...
>On Sat, 07 Aug 1999 17:37:53 +0200, Peter Santo
><san### [at] poolinformatikrwth-aachende> wrote:
>
>>Hello!
>>
>>I have a reduced scene containing a height_field that would cause povwin
>>to crash (so I suppose, it's a bug).
>>
>>I strongly suspect a division by zero error in the height_field code,
>>since the error can be avoided if you slightly change the perspective or
>>use Jitter when antialiasing output.
>>
>>My setup:
>>Official Version 3.1g.watcom.win32
>>Win98 de 4.10.1998
>>Pentium II 300 MHz
>>64 MB RAM
>
>I get this too.  As near as I can tell, it's crashing when the
>intersection exactly hits the very edge of a triangle on the
>edge of the heightfield.  The line it crashes on says
>
>      y3 = Get_Height(px,   pz+1, HField);
>
>pz is 159, and HField->Data->max_z is 158.  Not a pretty picture.
>I'm not sure what can be done about this, unfortunately, unless it's
>to subtract epsilon before rounding down to get pz.  Perhaps someone
>else has an idea.
>

Me too.
I am using POVRay 3.1a.watcom and I also got that error several times, with
different height_field TGA's and different scenes. (It once crashed in an
animation
on frame 20 of 500 and I didn't notice until hours later...arrgh)
I prepared a simple scene .ZIPped to 5K. If anyone needs it
for debugging, mail to bia### [at] inetmailde.
After 16% of this is rendered, POVRay causes a Windows error dialog box to
be
displayed ("Allgemeine Schutzverletzung" - what's that in the English
Windows version ?)
which said:

in Modul PVENGINE.EXE bei 015f:0046fb77.
[PVENGINE caused an error by illegal page in module PVENGINE.EXE at ...]
Register:
EAX=00000000 CS=015f EIP=0046fb77 EFLGS=00010246
EBX=018a210c SS=0167 ESP=0186fc50 EBP=0186fd34
ECX=b0b0b0b1 DS=0167 ESI=01880048 FS=13c7
EDX=018b0b28 ES=0167 EDI=00000001 GS=0000
Bytes bei CS:EIP:
66 8b 01 db 84 24 dc 00 00 00 89 84 24 dc 00 00
Stapelwerte:     [stack values]
00000000 3d100000 47bd4ee6 3fedbcc2
00000001 00000063 00000000 17971c19
00000004 0186f72c 0186f72c 0186f908
00000008 00000004 00000001 0186fd80

This must be the same problem as Peter Santo has.
My setup is:
iP 133 MHz, 32MB RAM, Win98
A friend of mine with an
iP2  333 MHz, 64MB RAM, Win95C
got the error at 16%, too.
Something has to be done about this, but I don't think I'd understand
the source code... (just about to start with C)

Post a reply to this message