POV-Ray

The following scene shows how photons are “boosted” by pass_through objects; removing one of the boxes will reduce the effect; the effect can be seen with 3.6 as well as current betas:

global_settings {
  max_trace_level 10 // makes a difference!
  photons { spacing 0.02 }
}

camera {
  right x*image_width/image_height
  location  <0,2.6,-10>
  look_at   <0,0.75,0>
}

light_source {
  <500,500,150>
  color rgb 1.3
  photons {
    refraction on
    reflection on
  }
}

sky_sphere {
  pigment {
    gradient y
    color_map {
      [0.0 rgb <0.6,0.7,1.0>]
      [0.7 rgb <0.0,0.1,0.8>]
    }
  }
}

plane {
  y, 0
  texture { pigment { color rgb <1.0, 0.8, 0.6> } }
}

#declare M_Glass=
material {
  texture {
    pigment {rgbt 1}
    finish {
      ambient 0.0
      diffuse 0.05
      specular 0.6
      roughness 0.005
      reflection { 0.1, 1.0 fresnel on }
      conserve_energy
    }
  }
  interior {
    ior 1.5
    fade_power 1001
    fade_distance 0.9
    fade_color <0.5,0.8,0.6>
  }
}


sphere {
  <1.1,1,-1.3>, 1
  material { M_Glass }
  photons {
    target 1.0
    refraction on
    reflection on
  }
}

cylinder {
  <-1.2,0.01,0.8>, <-1.2,2.5,0.8>, 1
  material { M_Glass }
  photons {  // photon block for an object
    target 1.0
    refraction on
    reflection on
  }
}

box {
  <2.4,0,-2.3>, <2.6,4,-0.3>
  material { M_Glass }
  photons { pass_through }
}

box {
  <2.9,0,-2.3>, <3.1,4,-0.3>
  material { M_Glass }
  photons { pass_through }
}

Some sample scenes are no longer up-to-date, causing warnings, and should be fixed. For instance, the advanced/benchmark scene still includes “Buffer_Output=Off” and “Buffer_Size=0” in its .ini file. This should be checked systematically, and fixed as appropriate.

Currently, output image is buffered using POV-Ray’s RGBFT color model and floating point values, leading to a memory consumption of 20 bit per pixel. This is an unproportionally large memory footprint, given that the only further processing performed on the buffered data is conversion of the data to the desired output file format, which will typically use only 3 bytes per pixel (at most 12 bit per pixel).

The situation can be improved by choosing the output image buffer container based on the desired output file format and parameters. To this end, the following code changes should be made:

• Bundle image file format handler code into classes (typically one per file format)
• Include a method to determine the best data container type for the image buffer
• Instantiate the desired image file format handler object prior to rendering, querying the handler object for an image buffer

In addition to picking a suitable image data container from the already existing palette, careful design might also allow to use custom containers to directly pass the data to a library (in case the library provides its own buffering anyway), or write it directly to the output file (e.g. when writing image data to stdout). To make this compatible with multithreaded rendering, the following changes would have to be made:

• Add code to detect when a row of SMP blocks has been finished, and call a certain method on the image handler object
• Design the custom containers in such a way that they can buffer any number of unfinished SMP block rows as needed

When running make on Linux, the backend source files (and possibly others?) are apparently compiled twice: first from the .../source/backend/ directory, and another time from the .../source/ directory. As an example, here are the corresponding lines for sphsweep.cpp:

g++ -DHAVE_CONFIG_H -I. -I../.. -I../.. -I../../source -I../../source -I../../source/base -I../../unix -I../../vfe 
-I../../vfe/unix -pthread -I/usr/include/OpenEXR    -pthread -I/usr/include  -pipe -Wno-multichar -Wno-write-strin
gs -fno-enforce-eh-specs -s -O3 -ffast-math -pthread -MT sphsweep.o -MD -MP -MF .deps/sphsweep.Tpo -c -o sphsweep.
o `test -f 'shape/sphsweep.cpp' || echo './'`shape/sphsweep.cpp

g++ -DHAVE_CONFIG_H -I. -I..  -I.. -I../source/backend -I../source/base -I../source/frontend -I../unix -I../vfe -I.
./vfe/unix -pthread -I/usr/include/OpenEXR    -pthread -I/usr/include  -pipe -Wno-multichar -Wno-write-strings -fno
-enforce-eh-specs -s -O3 -ffast-math -pthread -MT sphsweep.o -MD -MP -MF .deps/sphsweep.Tpo -c -o sphsweep.o `test 
-f 'backend/shape/sphsweep.cpp' || echo './'`backend/shape/sphsweep.cpp

This is especially annoying on platforms that are rather slow at compiling.

The following scene has four spheres with different pigment color & filter settings:

- Left: filter 1
- Right: filter 0

- Top: red 0.0 green 0.5 blue 1.0
- Bottom: red 0.00 green 0.05 blue 0.10 (10% of the above)

Background is set to black, so that we only see the diffuse component of the object’s effective color.

Theoretically, both left spheres should be invisible, as they are fully transmissive (with a filtering effect), but apparently with a high filter setting, reducing an object’s pigment color actually increases the object’s effective diffuse color.

//+w600 +h600

global_settings{ assumed_gamma 1.0 }

camera {
  orthographic
  location  <0,0,-10>
  right 4*x
  up 4*y
  look_at   <0,0,0>
}

light_source{<10,10,-10> color rgb 1 parallel }

background { color rgb 0 }

default {
  finish {
    ambient 0
    diffuse 1
    specular 0
    phong 0
    reflection { 0.0 }
  }
}


sphere { <-1, 1, 0>, 0.8 texture { pigment { color rgb <0,0.5,1.0> filter 1.0 } } }
sphere { < 1, 1, 0>, 0.8 texture { pigment { color rgb <0,0.5,1.0> filter 0.0 } } }

sphere { <-1,-1, 0>, 0.8 texture { pigment { color rgb <0,0.5,1.0>*0.1 filter 1.0 } } }
sphere { < 1,-1, 0>, 0.8 texture { pigment { color rgb <0,0.5,1.0>*0.1 filter 0.0 } } }

This bug has been around in 3.6 already.

When using fade_distance and fade_power in combination with area_illumination, the light source fading is not applied to materials with subsurface scattering; see the following code for an example:

#version 3.7;
global_settings {
  assumed_gamma 1.0
  mm_per_unit 10
  subsurface { samples 200,20 }
}

camera {
  right x*image_width/image_height
  angle 30
  location  <0,1.5,-4>
  look_at   <0,0,0>
}

sky_sphere {
  pigment {
    gradient y
    color_map {
      [0.0 rgb <0.6,0.7,1.0>]
      [0.7 rgb <0.0,0.1,0.8>]
    }
  }
}

plane {
  y, 0
  texture {
    pigment {
      checker
      color rgb <1.0, 0.8, 0.6>
      color rgb <1.0, 0.0, 0.0>
      scale 0.5
    }
  }
}

light_source {
  <50,50,50>
  color rgb 30
  area_light 5*x,5*y,17,17 adaptive 1 jitter circular orient
  area_illumination on
  fade_distance 10
  fade_power 2
}

cylinder {
  <0,0,0>, <0,0.2,0> 1
  texture {
    pigment { color rgb 1 }
    finish {
      ambient 0
      diffuse 0.7
      specular albedo 0.3
      reflection { 0.3 fresnel }
      conserve_energy
      subsurface { translucency 0.1 }
    }
  }
  interior { ior 1.5 }
}

sphere {
  <0,0.4,0>, 0.2
  texture {
    pigment { color rgb <1,0.6,0.0> }
    finish {
      ambient 0
      diffuse 0.0
      specular albedo 0.8 metallic
      reflection { 1.0 metallic }
      conserve_energy
    }
  }
}

Using an end value of 1048576 or larger in a #for loop will cause the last iteration to be skipped, as can be demonstrated by the following code:

#declare N = 2000000;
#debug concat(”N = “,str(N, 0,50),”\n”)
#debug concat(”N-5 = “,str(N-5,0,50),”\n\n”)
#for (I, N-5, N, 1)

#debug concat("I   = ",str(I,0,50),"\n")

#end

(The limit was observed with a Win64 build; other builds may exhibit other limits or might even work fine, depending on the floating point engine used.)

As this limit is still far below the numeric precision limit, and a corresponding #while loop works fine with much higher values, this must be considered a bug rather than an inevitable limitation.

The bug can be tracked down to a faulty condition in tokenize.cpp, Parser::Parse_Directive(), CASE(END_TOKEN), case FOR_COND:

    if ( ((Step > 0) && (*CurrentPtr >= End + EPSILON)) ||
         ((Step < 0) && (*CurrentPtr <= End - EPSILON)) )

which should instead be:

    if ( ((Step > 0) && (*CurrentPtr > End + EPSILON)) ||
         ((Step < 0) && (*CurrentPtr < End - EPSILON)) )

Bounding box computation for CSG intersection appears to be broken when one member is an arbitrarily transformed plane.

POV-Ray 3.6.2 has the same problem (can’t test for 3.6.1).

// +W640 +H480 +MB1

#include "transforms.inc"

camera {
  location  <-0.2, 0.5, -4.0>
  direction 1.5*z
  right     x*image_width/image_height
  look_at   <0.0, 0.0,  0.0>
}

sky_sphere {
  pigment {
    gradient y
    color_map {
      [0.0 rgb <0.6,0.7,1.0>]
      [0.7 rgb <0.0,0.1,0.8>]
    }
  }
}

light_source {
  <0, 0, 0>            // light's position (translated below)
  color rgb <1, 1, 1>  // light's color
  translate <-30, 30, -30>
}

plane {
  y, -1
  pigment { color rgb <0.7,0.5,0.3> }
}

intersection {
  sphere {
    0.0, 1 }
  plane { -x, 0 transform { Shear_Trans(x,y+x*0.3,z) } }
  texture {
    pigment {
      radial
      frequency 8
      color_map {
        [0.00 color rgb <1.0,0.4,0.2> ]
        [0.33 color rgb <0.2,0.4,1.0> ]
        [0.66 color rgb <0.4,1.0,0.2> ]
        [1.00 color rgb <1.0,0.4,0.2> ]
      }
    }
    finish{
      specular 0.6
    }
  }
  rotate -y*5
}

When aborting a render when there are unfinished blocks among finished ones, under certain conditions some of those blocks are skipped when continuing the render later.

The attached TIFF file was created with IC using compression. When used in an image_map, POV-Ray 3.7.0.beta.34 on Windows XP x64 renders the image all transparent, while POV-Ray 3.6.2 renders the file fine. The same effect can be seen with LZW-compressed TIFF files created with Adobe Photoshop 6.0.

Uncompressed TIFF files created with either IC or Photoshop render fine in both versions of POV-Ray.

Stepping through the code of POV-Ray 3.7.0 shows that the same code path is taken regardless of compression, but the libtiff library returns different alpha channel values, indicating a problem in that library. POV-Ray 3.7 still uses libtiff 3.6.1, whereas the POV-Ray 3.6 branch has been updated to libtiff 3.8.2, which according to the change history includes a few alpha-related changes.

Source code inspection during examination of issues with the scene published at http://povray.sitewww.ch/?p=177 show the following issues with current (beta.34) implementation of dispersion in POV-Ray 3.7:

• The same adjustment to the IOR that is applied at the very first dispersion interface is erroneously applied to all subsequent interfaces.

• As an exception, dispersion adjustment is erroneously not applied to any interface defined by the surface of a non-dispersing object embedded into a dispersing object.

While this still allows to use dispersion for artistic effect, it is neither physically realistic, nor does it match 3.6 behavior.

When rendering a series of scenes (e.g. animation, or render queue in POV-Ray for Windows), number_of_waves is not properly reset to its default value between scenes, causing the parameter to default to the value set by the previous scene.

For instance, rendering the following scenes from a queue will cause “arches.pov” to be rendered differently the second time:

1. scenes\textures\finishes\arches.pov
2. scenes\textures\normals\normavg.pov
3. scenes\textures\finishes\arches.pov (again!)

The parser does not accept mesh objects (or CSG objects including a mesh object) as a parameter to the inside() built-in function, reporting error “Solid object identifier expected”, even if the mesh is “solidified” by specifying an inside_vector.

(see news://news.povray.org:119/4a983716@news.povray.org)

Trace level is not handled properly with blobs using per-component reflective or refractive textures, leading to lockups. Stepping through the code, it seems that POV-Ray fails to properly mark the blob object as increasing the trace level.

(Until this bug is fixed, the issue can be worked around in most cases by assigning a reflective texture to the blob as a whole.)

In POV-Ray 3.6 (including 3.62), checker and cells patterns are off by 0.001 (1e-3) units, as can be demonstrated with this scene:

camera {
  location  <0.0, 0.0, -5.0>
  direction 1.5*z
  right     x*image_width/image_height
  look_at   <0.0, 0.0,  0.0>
}

box { <-1,-1,0>, <0,0,1> pigment { checker color rgb 1 color rgb 0 scale   0.2 translate <-0.5,-0.5,0> } finish { ambient 1 diffuse 0 } }
box { < 1, 1,0>, <0,0,1> pigment { checker color rgb 1 color rgb 0 scale   0.2 translate < 0.5, 0.5,0> } finish { ambient 1 diffuse 0 } }
box { < 1,-1,0>, <0,0,1> pigment { checker color rgb 1 color rgb 0 scale 200.0 translate < 0.5,-0.5,0> } finish { ambient 1 diffuse 0 } }
box { <-1, 1,0>, <0,0,1> pigment { checker color rgb 1 color rgb 0 scale 200.0 translate <-0.5, 0.5,0> } finish { ambient 1 diffuse 0 } }

The same can be demonstrated for the cells pattern.

POV-Ray 3.7 beta 34 is “clean”.

The configure script for unix uses a dumb string compare to test whether libpng version is 1.2.5 or higher, leading it to reject (for instance) libpng 1.2.27 and unnecessarily compile and statically link the older libpng version it comes with.

POV-Ray 3.7 does not show output options, such as:

Output Options
  Image resolution 640 by 480 (rows 1 to 480, columns 1 to 640).
  Output file: D:\foo\test.png, 24 bpp PNG
  Graphic display......On  (gamma: 2.2)
  Mosaic preview.......Off
  CPU usage histogram..Off
  Continued trace......Off

(the above is what 3.6 used to show)

Bounding box computation for CSG intersection appears to be broken when one member is an arbitrarily transformed plane.

// +W640 +H480 +MB1

#include "transforms.inc"

camera {
  location  <-0.2, 0.5, -4.0>
  direction 1.5*z
  right     x*image_width/image_height
  look_at   <0.0, 0.0,  0.0>
}

sky_sphere {
  pigment {
    gradient y
    color_map {
      [0.0 rgb <0.6,0.7,1.0>]
      [0.7 rgb <0.0,0.1,0.8>]
    }
  }
}

light_source {
  <0, 0, 0>            // light's position (translated below)
  color rgb <1, 1, 1>  // light's color
  translate <-30, 30, -30>
}

plane {
  y, -1
  pigment { color rgb <0.7,0.5,0.3> }
}

intersection {
  sphere {
    0.0, 1 }
  plane { -x, 0 transform { Shear_Trans(x,y+x*0.3,z) } }
  texture {
    pigment {
      radial
      frequency 8
      color_map {
        [0.00 color rgb <1.0,0.4,0.2> ]
        [0.33 color rgb <0.2,0.4,1.0> ]
        [0.66 color rgb <0.4,1.0,0.2> ]
        [1.00 color rgb <1.0,0.4,0.2> ]
      }
    }
    finish{
      specular 0.6
    }
  }
  rotate -y*5
}

In some cases it may be desirable to run POV-Ray from a batch file, without causing it to “steal the focus”.

I suggest making this dependant on whether POV-Ray is run with the /EXIT parameter.

Thin tori exhibit artifacts in 3.7.0 RC3 when the camera is placed inside the torus close to its “center plane”, as can be demonstrated with the following scene:

camera {
  location  <0.0, 0.0, -0.5>
  direction 1.5*z
  right     x*image_width/image_height
  look_at   <0.0, 0.0,  0.0>
  angle 1
}

light_source { <-30, 30, -30> color rgb 1 }

torus {
  1, 0.001
  texture { pigment { color red 1 } }
}

A pigment statement not wrapped in a texture statement does not properly override a pigment previously defined for the object. In the following SDL code:

  #declare PLANE = plane { y,0
    texture {
      pigment { checker color rgb 1 color rgb 0 scale 0.1 }
  } }
  object { PLANE
    pigment { checker color red 1 color blue 1 scale 1.0 }
  }

the scaling of the pigment previously specified for the PLANE object is retained for the new pigment. Compare:

  #declare PLANE = plane { y,0
    texture {
      pigment { checker color rgb 1 color rgb 0 scale 0.1 }
  } }
  object { PLANE
    texture {
      pigment { checker color red 1 color blue 1 scale 1.0 }
  } }

which behaves as expected.

The issue has been around at least since POV-Ray 3.6.2.

Blobs are currently limited to 1,000,000 components (with each cylindrical component counting as three: one cylinder + two end hemispheres); this limit may have served a historic purpose, but is now entirely arbitrary: The remaining code is limited only by the available RAM and the numeric limits of the int data type. The arbitrary maximum components limit per blob should therefore be removed.

Aside from unnecessarily limiting the power of the blob component, another drawback of the current test is that it is only performed after parsing of all the blob’s components, potentially hours after the limit had actually been reached.

For some particularly “heavy” SDL scripts, it might be desirable to override (or complement) the standard “Parsing 47110815K tokens” progress information with some more helpful custom info, e.g. “Planting trees... (37%)”, or “Generating terrain mesh row 47 of 500”.