errata.graphicsgemsiii

[Game.Programming].Academic - Graphics Gems (6 books source code)

Errata to _Graphics Gems III_, first edition, edited by David Kirk
([e-mail]), Academic Press 1992.  Code available online in
http://www.graphicsgems.org/

compiled by Eric Haines ([e-mail]) from author and reader contributions

version 1.16
date:  5/13/2002

-----

Errors in the text:

p. 67, equation 2.2 at bottom: should read "a := c - a;"

p. 201, just before "Timing Measurements" section: "corresponds to collinear
	line segments" should read "corresponds to parallel or collinear line
	segments".

p. 204, middle of page: the y-component of the center should be negative, so
	that it matches the equation two lines further down.

p. 234, final two equations need minus signs in the denominators:

    P[u] = M[v]*Jd - N[v]*Id
           ---------------------
           M[u]*N[v] - M[v]*N[u]

    P[v] = N[u]*Id - M[u]*Jd
           ---------------------
           M[u]*N[v] - M[v]*N[u]

p. 285, third paragraph, first sentence:  change "with the complement of the
	residency mask" to "with the residency mask".

-----

The following are errors in the code listings (corrected in the online code at
princeton.edu:pub/Graphics/GraphicsGems/GemsIII).  Note that many of the code
listings online are different in minor and major ways from the code in the
book (see Addenda, at the end, for new code not in the book).


Serious errors (ones your compiler cannot or may not catch):

p. 330, Equation (4):  put a minus sign just before the 1/2pi term.

p. 394: Delete FLOOR and CEILING macros (they're more like truncations).
	Change ROUND macro to (i.e. add parentheses around "a"):
	#define ROUND(a)	((a)>0 ? (int)((a)+0.5) : -(int)(0.5-(a)))
	Change SGN macro to (i.e. change positive condition result to "1"):
	#define SGN(a)		(((a)<0) ? -1 : 1)

p. 401, line 9:  in V3Combine change last "result->y" to "result->z".

p. 422, third from last line:  change "tmp->xsize" to "dst->ysize".

p. 425-427:  bitbuf is used before being set; initialized bitbuf = 0 before
	use in the three scale_bitmap_* routines.

p. 430, line 8:  change "uranx" to "urany" in the "jittery" macro.

p. 438, after line 18, "dir = neighbor(...", add the following test:

	/***	maybe done with contour ***/

	if ( (x == start_x) && (y == start_y) && (chain != NULL) )
		if (dir == chain->dir)
			return(1);

	and the code 20 lines later labelled "maybe done with contour"
	should be deleted (3 lines).

p. 474:  the variable "rho" is not used, delete it.
p. 477, next to last two lines:  call sq_ellipsoid_tensor with five "1."'s and
	sq_toroid_tensor with six "1."'s.

p. 498, lines 53-54:  put "&" in front of a1,b1,c1 and a2,b2,c2.
p. 499, lines 27-28:  put "&" in front of xa,ya and xb,yb.

p. 500-1: change COLLINEAR to PARALLEL (i.e. when f==0, all we can conclude is
	that the two lines are parallel).

p. 521: Numerical instability can result during point-triangle intersection.
        A better SIGN3 macro is:
		#define EPS 10e-5
		#define SIGN3( A ) \
		  (((A).x < EPS) ? 4 : 0 | ((A).x > -EPS) ? 32 : 0 | \
		   ((A).y < EPS) ? 2 : 0 | ((A).y > -EPS) ? 16 : 0 | \
		   ((A).z < EPS) ? 1 : 0 | ((A).z > -EPS) ? 8 : 0)
	and the point_triangle_intersection test (p. 526) changes to:
	   return ((sign12 & sign23 & sign31) == 0) ? OUTSIDE : INSIDE;
	   [addition of parentheses thanks to Martin Roth]

p. 525: change calls to check_point. The third argument to each call should
	switch p1 and p2. For example, the first correction should be:
		if (check_point(p1,p2,( .5-p1.x)/(p2.x-p1.x),0x3e) == INSIDE)
	  		return(INSIDE);

p. 547, beginning:  add line "#include <math.h>"

p. 570, line 49:  change to "major = V3New(0.0, 0.0, 1.0);"

p. 580, line 19:  change "f+= acos(VDOT(s, t)) * VDOT(sxt, Ns);" to
	"f-= acos(VDOT(s, t)) * VDOT(sxt, Nr);", i.e. "Ns" becomes "Nr".
	"Ns" can be removed from the list of calling parameters for
	"computeUnoccludedFormFactor()".

p. 606, beginning:  add "#include <string.h>"

-----

Syntax errors (ones that are not usually harmful, or are easily caught):

p. 396-404, throughout:  replace "};" with "}" to make lint happy
p. 402, gcd():  the variable "k" is set but not used - remove it

p. 411, top:  for ANSI compatibility, add procedure declarations:
	void RectStretch(long xs1,long ys1,long xs2,long ys2,long xd1,long yd1,
		long xd2,long yd2) ;
	void Stretch(long x1,long x2,long y1,long y2,long yr,long yw) ;
	void CircleStretch(long SBMINX,long SBMAXX,long xc,long yc,long r) ;
	void Stretch2Lines(long x1,long x2,long y1,long y2,long yr1,long yw1,
		long yr2,long yw2) ;

p. 411-413:  cast expressions inside "abs()" calls to type "(int)".

p. 414-424:  replace "ceiling" with "ceil" (on some machines)
p. 416, lines 18-19:  remove "char *p;" and "int width, height;" declarations;
	unused.
p. 420-421:  replace "(*filter)" with "(*filterf)" to avoid redeclaration of
	"filter".
p. 423, line 23:  remove "register char *p;" declaration; unused.

p. 425, line 33 and p. 426, line 51:  remove declaration of "i", as it is
	unused.

p. 427, line 29:  remove declarations for "*p" and "*q"; unused.

p. 428, line 35:  remove declarations for "i" and "j"; unused.

p. 441, line 37:  remove "nx" declaration; unused.
p. 443, line 2:  remove "xl, xr, yt, yb" declarations; unused.
p. 443, lines 3-4:  remove "k" and "ptb" declarations; unused.

p. 472, line 19:  remove "result" declaration; unused.
p. 474, lines 17-18:  remove "iworld" and "R" declarations; unused.
p. 477, line 32:  remove "x" declaration; unused.
p. 476-477:  note that sqellipsoidposn() and sqtoroidposn() are not
	particularly useful in their current forms; they do not return any
	values, e.g. x,y,z are computed but not returned.

p. 489, line 50:  cast "intsct" to "(void *)" to make lint happy.

p. 496, line 10:  remove global "r, p1, p2, p3, p4" declarations.
p. 496, line 10 and p. 499, throughout:  change "v1" to "gv1" and "v1" to
	"gv2".
p. 497, line 31:  remove "xt" and "yt" declarations; unused.
p. 498, line 27:  remove "t" and "u" declarations; unused.

p. 500, line 25:  remove "temp" declaration; unused.
p. 500, line 26:  remove "x3lo,x3hi" and "y3lo,y3hi" declarations; unused.

p. 513, line 17:  add the declarations for the static routines:
	static void computePlaneEq() ;
	static Node *insertPlaneEq() ;
	static int comparePlaneEqs() ;

p. 522, line 5:  remove "vect23, vect31" declarations; unused.

p. 543, line 45:  change "malloc" to "(StackPtr)malloc"
p. 544, line 54:  change "malloc" to "(BinNodePtr)malloc"
p. 546, line 17:  change "malloc" to "(BinNodePtr)malloc"

p. 551-554:  Note that this code uses definitions and subroutines that are
	found in Rayshade.  Most are easy enough to understand and code (or
	use Steve Hollasch's vector macros on pages 405-407).

p. 559, line 22:  remove "tmp" declaration; unused.

p. 570, line 9:  remove "j" declaration; unused.
p. 572, line 18:  remove "a, b, c" and "x1, y1, z1" declarations; unused (i.e.
	replace line with "double d;".
p. 573, lines 15 and 28:  remove "i" declarations; unused.

p. 577, line 12:  add the declarations for the static routines:
	static float computeFormFactor() ;
	static float computeUnoccludedFormFactor() ;
	static void splitQuad() ;
	static float quadArea() ;

p. 579, last line:  remove "j" declaration, as "j" is not used.

p. 607, line 40:  change to "memset((void *)acc, ..." to make lint happy.

-----

The following are typographical errors in the comments:

p. 429, line 24:  change "varables" to "variables"

p. 431, line 7:  change "Cyshosz" to "Cychosz"

p. 443, line 54:  change "seqment" to "segment"

p. 454, line 3:  change "architechture" to "architecture"

p. 458, line 21:  change "consistancy" to "consistency"

p. 484, line 46:  change "Inplementation" to "Implementation"

p. 493, lines 20 and 23:  change "subsegements" to "subsegments"

p. 539, line 18:  change "upto" to "up to"

p. 550, line 6:  change "parrallel" to "parallel"

p. 556, line 48:  change "miminum" to "minimum"

p. 572, line 7:  change "homogenous" to "homogeneous"

p. 578, line 15:  change "visiblity" to "visibility"

p. 606, line 32:  change "initialzed" to "initialized"

-----

Addenda:

There is test software (not in the text) for the "Fast n-Dimensional Extent
Overlap Testing" (exthit), "Accurate Polygon Scan Conversion Using Half-Open
Intervals" (accurate_scan), and "Partitioning a 3-D Convex Polygon with an
Arbitrary Plane" (partition3d).

There is C code for Badouel and Wurthrich's Gem II.9 (ndline.c) in the code
distribution (but not in the text).

There is more explanatory C code for Woo's Gem VII.1 in the code distribution.

There is an improved rgbvary.c included in the code distribution for Microsoft
Windows named rgbvaryW.c.  On PC's this code takes up less space and is 8
times faster.  It looks straightforward to regroove VaryDIB24() to other
bitmap architectures.

Here is a little additional explanation to the "Use of Residency Masks..."
gem on p. 284-287 from Joe Cychosz (note errata correction above):

	The process goes something like this:

	ray_mask = 0

	foreach (cell the ray passes through) {
	    foreach (object in the cell) {
		if  (ray_mask & object_mask == 0) {
		    compute the ray-object intersection
		}
	    }
	    update ray_mask marking for cell
	}

	The important thing here is to mark the cell after it is processed.

The filter.c file is updated as the filter_rcg.c by Ray Gardener. This new
version of filter.c is faster, and also includes Targa file I/O for testing
purposes.