Merge branch 'master' of github.com:cnlohr/libsurvive

Conflicts:
	src/survive_data.c

10 files changed, 1821 insertions, 4 deletions

diff --git a/redist/glutil.c b/redist/glutil.c
new file mode 100644
index 0000000..6b0d0f3
--- /dev/null
+++ b/redist/glutil.c
@@ -0,0 +1,9 @@
+//
+//  glutil.c
+//  
+//
+//  Created by user on 2/4/17.
+//
+//
+
+#include "glutil.h"
diff --git a/redist/glutil.h b/redist/glutil.h
new file mode 100644
index 0000000..bb40174
--- /dev/null
+++ b/redist/glutil.h
@@ -0,0 +1,14 @@
+//
+//  glutil.h
+//  
+//
+//  Created by user on 2/4/17.
+//
+//
+
+#ifndef glutil_h
+#define glutil_h
+
+#include <stdio.h>
+
+#endif /* glutil_h */
diff --git a/redist/linmath.c b/redist/linmath.c
index 41fa18f..dd6bd15 100644
--- a/redist/linmath.c
+++ b/redist/linmath.c
@@ -1,7 +1,7 @@
 //Copyright 2013,2016 <>< C. N. Lohr.  This file licensed under the terms of the MIT license.
 
-#include "linmath.h"
 #include <math.h>
+#include "linmath.h"
 #include <float.h>
 #include <string.h>
 
diff --git a/redist/linmath.h b/redist/linmath.h
index 66a38ed..4e0cb77 100644
--- a/redist/linmath.h
+++ b/redist/linmath.h
@@ -25,7 +25,7 @@
 #define FLT_ACOS  acos
 #define FLT_ASIN  asin
 #define FLT_ATAN2  atan2
-#define FLT_FABS fabs
+#define FLT_FABS__ fabs
 
 #else
 
@@ -37,10 +37,15 @@
 #define FLT_ACOS  acosf
 #define FLT_ASIN  asinf
 #define FLT_ATAN2  atan2f
-#define FLT_FABS fabsf
+#define FLT_FABS__ fabsf
 
 #endif
 
+#ifdef TCC
+#define FLT_FABS(x) (((x)<0)?(-(x)):(x))
+#else
+#define FLT_FABS FLT_FABS__
+#endif
 
 
 //NOTE: Inputs may never be output with cross product.
diff --git a/redist/mymath.c b/redist/mymath.c
new file mode 100644
index 0000000..17dcc96
--- /dev/null
+++ b/redist/mymath.c
@@ -0,0 +1,550 @@
+//Copyright 2013,2016 <>< C. N. Lohr.  This file licensed under the terms of the MIT license.
+
+#include <math.h>
+#include <float.h>
+#include <string.h>
+#include "linmath.h"
+
+#define FLT float
+#define LINMATHPI ((FLT)3.141592653589)
+#define DEFAULT_EPSILON 0.001
+void quatnormalize( FLT * qout, const FLT * qin );
+void quatscale( FLT * qout, const FLT * qin, FLT s );
+void quatfrom2vectors(FLT *q, const FLT *src, const FLT *dest);
+
+
+void cross3d( FLT * out, const FLT * a, const FLT * b )
+{
+	out[0] = a[1]*b[2] - a[2]*b[1];
+	out[1] = a[2]*b[0] - a[0]*b[2];
+	out[2] = a[0]*b[1] - a[1]*b[0];
+}
+
+void sub3d( FLT * out, const FLT * a, const FLT * b )
+{
+	out[0] = a[0] - b[0];
+	out[1] = a[1] - b[1];
+	out[2] = a[2] - b[2];
+}
+
+void add3d( FLT * out, const FLT * a, const FLT * b )
+{
+	out[0] = a[0] + b[0];
+	out[1] = a[1] + b[1];
+	out[2] = a[2] + b[2];
+}
+
+void scale3d( FLT * out, const FLT * a, FLT scalar )
+{
+	out[0] = a[0] * scalar;
+	out[1] = a[1] * scalar;
+	out[2] = a[2] * scalar;
+}
+
+void normalize3d( FLT * out, const FLT * in )
+{
+	FLT r = ((FLT)1.) / FLT_SQRT(in[0] * in[0] + in[1] * in[1] + in[2] * in[2]);
+	out[0] = in[0] * r;
+	out[1] = in[1] * r;
+	out[2] = in[2] * r;
+}
+
+FLT dot3d( const FLT * a, const FLT * b )
+{
+	return a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
+}
+
+int compare3d( const FLT * a, const FLT * b, FLT epsilon )
+{
+	if( !a || !b ) return 0;
+	if( a[2] - b[2] > epsilon ) return 1;
+	if( b[2] - a[2] > epsilon ) return -1;
+	if( a[1] - b[1] > epsilon ) return 1;
+	if( b[1] - a[1] > epsilon ) return -1;
+	if( a[0] - b[0] > epsilon ) return 1;
+	if( b[0] - a[0] > epsilon ) return -1;
+	return 0;
+}
+
+void copy3d( FLT * out, const FLT * in )
+{
+	out[0] = in[0];
+	out[1] = in[1];
+	out[2] = in[2];
+}
+
+FLT magnitude3d(const FLT * a )
+{
+	return FLT_SQRT(a[0] * a[0] + a[1] * a[1] + a[2] * a[2]);
+}
+
+FLT anglebetween3d( FLT * a, FLT * b )
+{
+	FLT an[3];
+	FLT bn[3];
+	normalize3d( an, a );
+	normalize3d( bn, b );
+	FLT dot = dot3d(an, bn);
+	if( dot < -0.9999999 ) return LINMATHPI;
+	if( dot >  0.9999999 ) return 0;
+	return FLT_ACOS(dot);
+}
+
+/////////////////////////////////////QUATERNIONS//////////////////////////////////////////
+//Originally from Mercury (Copyright (C) 2009 by Joshua Allen, Charles Lohr, Adam Lowman)
+//Under the mit/X11 license.
+
+
+
+
+void quatsetnone(FLT * q)
+{
+	q[0] = 1; q[1] = 0; q[2] = 0; q[3] = 0;
+}
+
+void quatcopy(FLT * qout, const FLT * qin)
+{
+	qout[0] = qin[0];
+	qout[1] = qin[1];
+	qout[2] = qin[2];
+	qout[3] = qin[3];
+}
+
+void quatfromeuler( FLT * q, const FLT * euler )
+{
+	FLT X = euler[0]/2.0f; //roll
+	FLT Y = euler[1]/2.0f; //pitch
+	FLT Z = euler[2]/2.0f; //yaw
+
+	FLT cx = FLT_COS(X);
+	FLT sx = FLT_SIN(X);
+	FLT cy = FLT_COS(Y);
+	FLT sy = FLT_SIN(Y);
+	FLT cz = FLT_COS(Z);
+	FLT sz = FLT_SIN(Z);
+
+	//Correct according to
+	//http://en.wikipedia.org/wiki/Conversion_between_MQuaternions_and_Euler_angles
+	q[0] = cx*cy*cz+sx*sy*sz;//q1
+	q[1] = sx*cy*cz-cx*sy*sz;//q2
+	q[2] = cx*sy*cz+sx*cy*sz;//q3
+	q[3] = cx*cy*sz-sx*sy*cz;//q4
+	quatnormalize( q, q );
+}
+
+void quattoeuler( FLT * euler, const FLT * q )
+{
+	//According to http://en.wikipedia.org/wiki/Conversion_between_quaternions_and_Euler_angles (Oct 26, 2009)
+	euler[0] = FLT_ATAN2(2 * (q[0] * q[1] + q[2] * q[3]), 1 - 2 * (q[1] * q[1] + q[2] * q[2]));
+	euler[1] = FLT_ASIN(2 * (q[0] * q[2] - q[3] * q[1]));
+	euler[2] = FLT_ATAN2(2 * (q[0] * q[3] + q[1] * q[2]), 1 - 2 * (q[2] * q[2] + q[3] * q[3]));
+}
+
+void quatfromaxisangle( FLT * q, const FLT * axis, FLT radians )
+{
+	FLT v[3];
+	normalize3d( v, axis );
+	
+	FLT sn = FLT_SIN(radians / 2.0f);
+	q[0] = FLT_COS(radians / 2.0f);
+	q[1] = sn * v[0];
+	q[2] = sn * v[1];
+	q[3] = sn * v[2];
+
+	quatnormalize( q, q );
+}
+
+FLT quatmagnitude( const FLT * q )
+{
+	return FLT_SQRT((q[0] * q[0]) + (q[1] * q[1]) + (q[2] * q[2]) + (q[3] * q[3]));
+}
+
+FLT quatinvsqmagnitude( const FLT * q )
+{
+	return ((FLT)1.)/FLT_SQRT((q[0]*q[0])+(q[1]*q[1])+(q[2]*q[2])+(q[3]*q[3]));
+}
+
+
+void quatnormalize( FLT * qout, const FLT * qin )
+{
+	FLT imag = quatinvsqmagnitude( qin );
+	quatscale( qout, qin, imag );
+}
+
+void quattomatrix(FLT * matrix44, const FLT * qin)
+{
+	FLT q[4];
+	quatnormalize(q, qin);
+
+	//Reduced calulation for speed
+	FLT xx = 2 * q[1] * q[1];
+	FLT xy = 2 * q[1] * q[2];
+	FLT xz = 2 * q[1] * q[3];
+	FLT xw = 2 * q[1] * q[0];
+
+	FLT yy = 2 * q[2] * q[2];
+	FLT yz = 2 * q[2] * q[3];
+	FLT yw = 2 * q[2] * q[0];
+
+	FLT zz = 2 * q[3] * q[3];
+	FLT zw = 2 * q[3] * q[0];
+
+	//opengl major
+	matrix44[0] = 1 - yy - zz;
+	matrix44[1] = xy - zw;
+	matrix44[2] = xz + yw;
+	matrix44[3] = 0;
+
+	matrix44[4] = xy + zw;
+	matrix44[5] = 1 - xx - zz;
+	matrix44[6] = yz - xw;
+	matrix44[7] = 0;
+
+	matrix44[8] = xz - yw;
+	matrix44[9] = yz + xw;
+	matrix44[10] = 1 - xx - yy;
+	matrix44[11] = 0;
+
+	matrix44[12] = 0;
+	matrix44[13] = 0;
+	matrix44[14] = 0;
+	matrix44[15] = 1;
+}
+
+
+void quatfrommatrix( FLT * q, const FLT * matrix44 )
+{
+	//Algorithm from http://www.euclideanspace.com/maths/geometry/rotations/conversions/matrixToQuaternion/
+	float tr = matrix44[0] + matrix44[5] + matrix44[10];
+
+	if (tr > 0) {
+		float S = sqrt(tr+1.0) * 2; // S=4*qw
+		q[0] = 0.25 * S;
+		q[1] = (matrix44[9] - matrix44[6]) / S;
+		q[2] = (matrix44[2] - matrix44[8]) / S;
+		q[3] = (matrix44[4] - matrix44[1]) / S;
+	} else if ((matrix44[0] > matrix44[5])&(matrix44[0] > matrix44[10])) {
+		float S = sqrt(1.0 + matrix44[0] - matrix44[5] - matrix44[10]) * 2; // S=4*qx
+		q[0] = (matrix44[9] - matrix44[6]) / S;
+		q[1] = 0.25 * S;
+		q[2] = (matrix44[1] + matrix44[4]) / S;
+		q[3] = (matrix44[2] + matrix44[8]) / S;
+	} else if (matrix44[5] > matrix44[10]) {
+		float S = sqrt(1.0 + matrix44[5] - matrix44[0] - matrix44[10]) * 2; // S=4*qy
+		q[0] = (matrix44[2] - matrix44[8]) / S;
+		q[1] = (matrix44[1] + matrix44[4]) / S;
+		q[2] = 0.25 * S;
+		q[3] = (matrix44[6] + matrix44[9]) / S;
+	} else {
+		float S = sqrt(1.0 + matrix44[10] - matrix44[0] - matrix44[5]) * 2; // S=4*qz
+		q[0] = (matrix44[4] - matrix44[1]) / S;
+		q[1] = (matrix44[2] + matrix44[8]) / S;
+		q[2] = (matrix44[6] + matrix44[9]) / S;
+		q[3] = 0.25 * S;
+	}
+}
+
+
+void quattomatrix33(FLT * matrix33, const FLT * qin)
+{
+	FLT q[4];
+	quatnormalize(q, qin);
+
+	//Reduced calulation for speed
+	FLT xx = 2 * q[0] * q[0];
+	FLT xy = 2 * q[0] * q[1];
+	FLT xz = 2 * q[0] * q[2];
+	FLT xw = 2 * q[0] * q[3];
+
+	FLT yy = 2 * q[1] * q[1];
+	FLT yz = 2 * q[1] * q[2];
+	FLT yw = 2 * q[1] * q[3];
+
+	FLT zz = 2 * q[2] * q[2];
+	FLT zw = 2 * q[2] * q[3];
+
+	//opengl major
+	matrix33[0] = 1 - yy - zz;
+	matrix33[1] = xy - zw;
+	matrix33[2] = xz + yw;
+
+	matrix33[3] = xy + zw;
+	matrix33[4] = 1 - xx - zz;
+	matrix33[5] = yz - xw;
+
+	matrix33[6] = xz - yw;
+	matrix33[7] = yz + xw;
+	matrix33[8] = 1 - xx - yy;
+}
+
+void quatgetconjugate(FLT * qout, const FLT * qin)
+{
+	qout[0] = qin[0];
+	qout[1] = -qin[1];
+	qout[2] = -qin[2];
+	qout[3] = -qin[3];
+}
+
+void quatgetreciprocal( FLT * qout, const FLT * qin )
+{
+	FLT m = quatinvsqmagnitude(qin);
+	quatgetconjugate( qout, qin );
+	quatscale( qout, qout, m );
+}
+
+void quatsub( FLT * qout, const FLT * a, const FLT * b )
+{
+	qout[0] = a[0] - b[0];
+	qout[1] = a[1] - b[1];
+	qout[2] = a[2] - b[2];
+	qout[3] = a[3] - b[3];
+}
+
+void quatadd( FLT * qout, const FLT * a, const FLT * b )
+{
+	qout[0] = a[0] + b[0];
+	qout[1] = a[1] + b[1];
+	qout[2] = a[2] + b[2];
+	qout[3] = a[3] + b[3];
+}
+
+void quatrotateabout( FLT * qout, const FLT * q1, const FLT * q2 )
+{
+	//NOTE: Does not normalize
+	qout[0] = (q1[0]*q2[0])-(q1[1]*q2[1])-(q1[2]*q2[2])-(q1[3]*q2[3]);
+	qout[1] = (q1[0]*q2[1])+(q1[1]*q2[0])+(q1[2]*q2[3])-(q1[3]*q2[2]);
+	qout[2] = (q1[0]*q2[2])-(q1[1]*q2[3])+(q1[2]*q2[0])+(q1[3]*q2[1]);
+	qout[3] = (q1[0]*q2[3])+(q1[1]*q2[2])-(q1[2]*q2[1])+(q1[3]*q2[0]);
+}
+
+void quatscale( FLT * qout, const FLT * qin, FLT s )
+{
+	qout[0] = qin[0] * s;
+	qout[1] = qin[1] * s;
+	qout[2] = qin[2] * s;
+	qout[3] = qin[3] * s;
+}
+
+
+FLT quatinnerproduct( const FLT * qa, const FLT * qb )
+{
+	return (qa[0]*qb[0])+(qa[1]*qb[1])+(qa[2]*qb[2])+(qa[3]*qb[3]);
+}
+
+void quatouterproduct( FLT * outvec3, FLT * qa, FLT * qb )
+{
+	outvec3[0] = (qa[0]*qb[1])-(qa[1]*qb[0])-(qa[2]*qb[3])+(qa[3]*qb[2]);
+	outvec3[1] = (qa[0]*qb[2])+(qa[1]*qb[3])-(qa[2]*qb[0])-(qa[3]*qb[1]);
+	outvec3[2] = (qa[0]*qb[3])-(qa[1]*qb[2])+(qa[2]*qb[1])-(qa[3]*qb[0]);
+}
+
+void quatevenproduct( FLT * q, FLT * qa, FLT * qb )
+{
+	q[0] = (qa[0]*qb[0])-(qa[1]*qb[1])-(qa[2]*qb[2])-(qa[3]*qb[3]);
+	q[1] = (qa[0]*qb[1])+(qa[1]*qb[0]);
+	q[2] = (qa[0]*qb[2])+(qa[2]*qb[0]);
+	q[3] = (qa[0]*qb[3])+(qa[3]*qb[0]);
+}
+
+void quatoddproduct( FLT * outvec3, FLT * qa, FLT * qb )
+{
+	outvec3[0] = (qa[2]*qb[3])-(qa[3]*qb[2]);
+	outvec3[1] = (qa[3]*qb[1])-(qa[1]*qb[3]);
+	outvec3[2] = (qa[1]*qb[2])-(qa[2]*qb[1]);
+}
+
+void quatslerp( FLT * q, const FLT * qa, const FLT * qb, FLT t )
+{
+	FLT an[4];
+	FLT bn[4];
+	quatnormalize( an, qa );
+	quatnormalize( bn, qb );
+	FLT cosTheta = quatinnerproduct(an,bn);
+	FLT sinTheta;
+
+	//Careful: If cosTheta is exactly one, or even if it's infinitesimally over, it'll
+	// cause SQRT to produce not a number, and screw everything up.
+	if ( 1 - (cosTheta*cosTheta) <= 0 )
+		sinTheta = 0;
+	else
+		sinTheta = FLT_SQRT(1 - (cosTheta*cosTheta));
+
+	FLT Theta = FLT_ACOS(cosTheta); //Theta is half the angle between the 2 MQuaternions
+
+	if (FLT_FABS(Theta) < DEFAULT_EPSILON)
+		quatcopy( q, qa );
+	else if (FLT_FABS(sinTheta) < DEFAULT_EPSILON)
+	{
+		quatadd( q, qa, qb );
+		quatscale( q, q, 0.5 );
+	}
+	else
+	{
+		FLT aside[4];
+		FLT bside[4];
+		quatscale( bside, qb, FLT_SIN(t * Theta));
+		quatscale( aside, qa, FLT_SIN((1 - t)*Theta));
+		quatadd( q, aside, bside );
+		quatscale( q, q, ((FLT)1.)/sinTheta );
+	}
+}
+
+void quatrotatevector( FLT * vec3out, const FLT * quat, const FLT * vec3in )
+{
+	FLT tquat[4];
+	FLT vquat[4];
+	FLT qrecp[4];
+	vquat[0] = 0;
+	vquat[1] = vec3in[0];
+	vquat[2] = vec3in[1];
+	vquat[3] = vec3in[2];
+
+	quatrotateabout( tquat, quat, vquat );
+	quatgetconjugate( qrecp, quat );
+	quatrotateabout( vquat, tquat, qrecp );
+
+	vec3out[0] = vquat[1];
+	vec3out[1] = vquat[2];
+	vec3out[2] = vquat[3];
+}
+
+
+// Matrix Stuff
+
+Matrix3x3 inverseM33(const Matrix3x3 mat)
+{
+	Matrix3x3 newMat;
+	for (int a = 0; a < 3; a++)
+	{
+		for (int b = 0; b < 3; b++)
+		{
+			newMat.val[a][b] = mat.val[a][b];
+		}
+	}
+
+	for (int i = 0; i < 3; i++)
+	{
+		for (int j = i + 1; j < 3; j++)
+		{
+			FLT tmp = newMat.val[i][j];
+			newMat.val[i][j] = newMat.val[j][i];
+			newMat.val[j][i] = tmp;
+		}
+	}
+
+	return newMat;
+}
+
+void rotation_between_vecs_to_m3(Matrix3x3 *m, const FLT v1[3], const FLT v2[3])
+{
+	FLT q[4];
+
+	quatfrom2vectors(q, v1, v2);
+
+	quattomatrix33(&(m->val[0][0]), q);
+}
+
+void rotate_vec(FLT *out, const FLT *in, Matrix3x3 rot)
+{
+	out[0] = rot.val[0][0] * in[0] + rot.val[1][0] * in[1] + rot.val[2][0] * in[2];
+	out[1] = rot.val[0][1] * in[0] + rot.val[1][1] * in[1] + rot.val[2][1] * in[2];
+	out[2] = rot.val[0][2] * in[0] + rot.val[1][2] * in[1] + rot.val[2][2] * in[2];
+
+	return;
+}
+
+
+// This function based on code from Object-oriented Graphics Rendering Engine
+// Copyright(c) 2000 - 2012 Torus Knot Software Ltd
+// under MIT license
+// http://www.ogre3d.org/docs/api/1.9/_ogre_vector3_8h_source.html
+
+/** Gets the shortest arc quaternion to rotate this vector to the destination
+vector.
+@remarks
+If you call this with a dest vector that is close to the inverse
+of this vector, we will rotate 180 degrees around a generated axis if
+since in this case ANY axis of rotation is valid.
+*/
+void quatfrom2vectors(FLT *q, const FLT *src, const FLT *dest)
+{
+	// Based on Stan Melax's article in Game Programming Gems
+
+	// Copy, since cannot modify local
+	FLT v0[3];
+	FLT v1[3];
+	normalize3d(v0, src);
+	normalize3d(v1, dest);
+
+	FLT d = dot3d(v0, v1);// v0.dotProduct(v1);
+	// If dot == 1, vectors are the same
+	if (d >= 1.0f)
+	{
+		quatsetnone(q);
+		return;
+	}
+	if (d < (1e-6f - 1.0f))
+	{
+		// Generate an axis
+		FLT unitX[3] = { 1, 0, 0 };
+		FLT unitY[3] = { 0, 1, 0 };
+		
+		FLT axis[3];
+		cross3d(axis, unitX, src); // pick an angle
+		if ((axis[0] < 1.0e-35f) &&
+			(axis[1] < 1.0e-35f) &&
+			(axis[2] < 1.0e-35f)) // pick another if colinear
+		{
+			cross3d(axis, unitY, src);
+		}
+		normalize3d(axis, axis);
+		quatfromaxisangle(q, axis, LINMATHPI);
+	}
+	else
+	{
+		FLT s = FLT_SQRT((1 + d) * 2);
+		FLT invs = 1 / s;
+
+		FLT c[3];
+		//cross3d(c, v0, v1);
+		cross3d(c, v1, v0);
+
+		q[0] = c[0] * invs;
+		q[1] = c[1] * invs;
+		q[2] = c[2] * invs;
+		q[3] = s * 0.5f;
+
+		quatnormalize(q, q);
+	}
+
+}
+
+void matrix44copy(FLT * mout, const FLT * minm )
+{
+	memcpy( mout, minm, sizeof( FLT ) * 16 );
+}
+
+void matrix44transpose(FLT * mout, const FLT * minm )
+{
+	mout[0] = minm[0];
+	mout[1] = minm[4];
+	mout[2] = minm[8];
+	mout[3] = minm[12];
+
+	mout[4] = minm[1];
+	mout[5] = minm[5];
+	mout[6] = minm[9];
+	mout[7] = minm[13];
+
+	mout[8] = minm[2];
+	mout[9] = minm[6];
+	mout[10] = minm[10];
+	mout[11] = minm[14];
+
+	mout[12] = minm[3];
+	mout[13] = minm[7];
+	mout[14] = minm[11];
+	mout[15] = minm[15];
+
+}
+
diff --git a/redist/os_generic.h b/redist/os_generic.h
index 7ce22f2..aac425b 100644
--- a/redist/os_generic.h
+++ b/redist/os_generic.h
@@ -1,7 +1,7 @@
 #ifndef _OS_GENERIC_H

 #define _OS_GENERIC_H

 

-#ifdef WIN32

+#if defined( WIN32 ) || defined (WINDOWS)

 #define USE_WINDOWS

 #endif

 

diff --git a/redist/puff.c b/redist/puff.c
new file mode 100644
index 0000000..fdbf6de
--- /dev/null
+++ b/redist/puff.c
@@ -0,0 +1,955 @@
+/*
+ * puff.c
+ * Copyright (C) 2002-2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in puff.h
+ * version 2.1, 4 Apr 2010
+ *
+ * puff.c is a simple inflate written to be an unambiguous way to specify the
+ * deflate format.  It is not written for speed but rather simplicity.  As a
+ * side benefit, this code might actually be useful when small code is more
+ * important than speed, such as bootstrap applications.  For typical deflate
+ * data, zlib's inflate() is about four times as fast as puff().  zlib's
+ * inflate compiles to around 20K on my machine, whereas puff.c compiles to
+ * around 4K on my machine (a PowerPC using GNU cc).  If the faster decode()
+ * function here is used, then puff() is only twice as slow as zlib's
+ * inflate().
+ *
+ * All dynamically allocated memory comes from the stack.  The stack required
+ * is less than 2K bytes.  This code is compatible with 16-bit int's and
+ * assumes that long's are at least 32 bits.  puff.c uses the short data type,
+ * assumed to be 16 bits, for arrays in order to to conserve memory.  The code
+ * works whether integers are stored big endian or little endian.
+ *
+ * In the comments below are "Format notes" that describe the inflate process
+ * and document some of the less obvious aspects of the format.  This source
+ * code is meant to supplement RFC 1951, which formally describes the deflate
+ * format:
+ *
+ *    http://www.zlib.org/rfc-deflate.html
+ */
+
+/*
+ * Change history:
+ *
+ * 1.0  10 Feb 2002     - First version
+ * 1.1  17 Feb 2002     - Clarifications of some comments and notes
+ *                      - Update puff() dest and source pointers on negative
+ *                        errors to facilitate debugging deflators
+ *                      - Remove longest from struct huffman -- not needed
+ *                      - Simplify offs[] index in construct()
+ *                      - Add input size and checking, using longjmp() to
+ *                        maintain easy readability
+ *                      - Use short data type for large arrays
+ *                      - Use pointers instead of long to specify source and
+ *                        destination sizes to avoid arbitrary 4 GB limits
+ * 1.2  17 Mar 2002     - Add faster version of decode(), doubles speed (!),
+ *                        but leave simple version for readabilty
+ *                      - Make sure invalid distances detected if pointers
+ *                        are 16 bits
+ *                      - Fix fixed codes table error
+ *                      - Provide a scanning mode for determining size of
+ *                        uncompressed data
+ * 1.3  20 Mar 2002     - Go back to lengths for puff() parameters [Jean-loup]
+ *                      - Add a puff.h file for the interface
+ *                      - Add braces in puff() for else do [Jean-loup]
+ *                      - Use indexes instead of pointers for readability
+ * 1.4  31 Mar 2002     - Simplify construct() code set check
+ *                      - Fix some comments
+ *                      - Add FIXLCODES #define
+ * 1.5   6 Apr 2002     - Minor comment fixes
+ * 1.6   7 Aug 2002     - Minor format changes
+ * 1.7   3 Mar 2003     - Added test code for distribution
+ *                      - Added zlib-like license
+ * 1.8   9 Jan 2004     - Added some comments on no distance codes case
+ * 1.9  21 Feb 2008     - Fix bug on 16-bit integer architectures [Pohland]
+ *                      - Catch missing end-of-block symbol error
+ * 2.0  25 Jul 2008     - Add #define to permit distance too far back
+ *                      - Add option in TEST code for puff to write the data
+ *                      - Add option in TEST code to skip input bytes
+ *                      - Allow TEST code to read from piped stdin
+ * 2.1   4 Apr 2010     - Avoid variable initialization for happier compilers
+ *                      - Avoid unsigned comparisons for even happier compilers
+ */
+
+#include <setjmp.h>             /* for setjmp(), longjmp(), and jmp_buf */
+#include "puff.h"               /* prototype for puff() */
+
+#define local static            /* for local function definitions */
+#define NIL ((unsigned char *)0)        /* for no output option */
+
+/*
+ * Maximums for allocations and loops.  It is not useful to change these --
+ * they are fixed by the deflate format.
+ */
+#define MAXBITS 15              /* maximum bits in a code */
+#define MAXLCODES 286           /* maximum number of literal/length codes */
+#define MAXDCODES 30            /* maximum number of distance codes */
+#define MAXCODES (MAXLCODES+MAXDCODES)  /* maximum codes lengths to read */
+#define FIXLCODES 288           /* number of fixed literal/length codes */
+
+/* input and output state */
+struct state {
+    /* output state */
+    unsigned char *out;         /* output buffer */
+    unsigned long outlen;       /* available space at out */
+    unsigned long outcnt;       /* bytes written to out so far */
+
+    /* input state */
+    const unsigned char *in;          /* input buffer */
+    unsigned long inlen;        /* available input at in */
+    unsigned long incnt;        /* bytes read so far */
+    int bitbuf;                 /* bit buffer */
+    int bitcnt;                 /* number of bits in bit buffer */
+
+    /* input limit error return state for bits() and decode() */
+    jmp_buf env;
+};
+
+/*
+ * Return need bits from the input stream.  This always leaves less than
+ * eight bits in the buffer.  bits() works properly for need == 0.
+ *
+ * Format notes:
+ *
+ * - Bits are stored in bytes from the least significant bit to the most
+ *   significant bit.  Therefore bits are dropped from the bottom of the bit
+ *   buffer, using shift right, and new bytes are appended to the top of the
+ *   bit buffer, using shift left.
+ */
+local int bits(struct state *s, int need)
+{
+    long val;           /* bit accumulator (can use up to 20 bits) */
+
+    /* load at least need bits into val */
+    val = s->bitbuf;
+    while (s->bitcnt < need) {
+        if (s->incnt == s->inlen) longjmp(s->env, 1);   /* out of input */
+        val |= (long)(s->in[s->incnt++]) << s->bitcnt;  /* load eight bits */
+        s->bitcnt += 8;
+    }
+
+    /* drop need bits and update buffer, always zero to seven bits left */
+    s->bitbuf = (int)(val >> need);
+    s->bitcnt -= need;
+
+    /* return need bits, zeroing the bits above that */
+    return (int)(val & ((1L << need) - 1));
+}
+
+/*
+ * Process a stored block.
+ *
+ * Format notes:
+ *
+ * - After the two-bit stored block type (00), the stored block length and
+ *   stored bytes are byte-aligned for fast copying.  Therefore any leftover
+ *   bits in the byte that has the last bit of the type, as many as seven, are
+ *   discarded.  The value of the discarded bits are not defined and should not
+ *   be checked against any expectation.
+ *
+ * - The second inverted copy of the stored block length does not have to be
+ *   checked, but it's probably a good idea to do so anyway.
+ *
+ * - A stored block can have zero length.  This is sometimes used to byte-align
+ *   subsets of the compressed data for random access or partial recovery.
+ */
+local int stored(struct state *s)
+{
+    unsigned len;       /* length of stored block */
+
+    /* discard leftover bits from current byte (assumes s->bitcnt < 8) */
+    s->bitbuf = 0;
+    s->bitcnt = 0;
+
+    /* get length and check against its one's complement */
+    if (s->incnt + 4 > s->inlen) return 2;      /* not enough input */
+    len = s->in[s->incnt++];
+    len |= s->in[s->incnt++] << 8;
+    if (s->in[s->incnt++] != (~len & 0xff) ||
+        s->in[s->incnt++] != ((~len >> 8) & 0xff))
+        return -2;                              /* didn't match complement! */
+
+    /* copy len bytes from in to out */
+    if (s->incnt + len > s->inlen) return 2;    /* not enough input */
+    if (s->out != NIL) {
+        if (s->outcnt + len > s->outlen)
+            return 1;                           /* not enough output space */
+        while (len--)
+            s->out[s->outcnt++] = s->in[s->incnt++];
+    }
+    else {                                      /* just scanning */
+        s->outcnt += len;
+        s->incnt += len;
+    }
+
+    /* done with a valid stored block */
+    return 0;
+}
+
+/*
+ * Huffman code decoding tables.  count[1..MAXBITS] is the number of symbols of
+ * each length, which for a canonical code are stepped through in order.
+ * symbol[] are the symbol values in canonical order, where the number of
+ * entries is the sum of the counts in count[].  The decoding process can be
+ * seen in the function decode() below.
+ */
+struct huffman {
+    short *count;       /* number of symbols of each length */
+    short *symbol;      /* canonically ordered symbols */
+};
+
+/*
+ * Decode a code from the stream s using huffman table h.  Return the symbol or
+ * a negative value if there is an error.  If all of the lengths are zero, i.e.
+ * an empty code, or if the code is incomplete and an invalid code is received,
+ * then -10 is returned after reading MAXBITS bits.
+ *
+ * Format notes:
+ *
+ * - The codes as stored in the compressed data are bit-reversed relative to
+ *   a simple integer ordering of codes of the same lengths.  Hence below the
+ *   bits are pulled from the compressed data one at a time and used to
+ *   build the code value reversed from what is in the stream in order to
+ *   permit simple integer comparisons for decoding.  A table-based decoding
+ *   scheme (as used in zlib) does not need to do this reversal.
+ *
+ * - The first code for the shortest length is all zeros.  Subsequent codes of
+ *   the same length are simply integer increments of the previous code.  When
+ *   moving up a length, a zero bit is appended to the code.  For a complete
+ *   code, the last code of the longest length will be all ones.
+ *
+ * - Incomplete codes are handled by this decoder, since they are permitted
+ *   in the deflate format.  See the format notes for fixed() and dynamic().
+ */
+#ifdef SLOW
+local int decode(struct state *s, struct huffman *h)
+{
+    int len;            /* current number of bits in code */
+    int code;           /* len bits being decoded */
+    int first;          /* first code of length len */
+    int count;          /* number of codes of length len */
+    int index;          /* index of first code of length len in symbol table */
+
+    code = first = index = 0;
+    for (len = 1; len <= MAXBITS; len++) {
+        code |= bits(s, 1);             /* get next bit */
+        count = h->count[len];
+        if (code - count < first)       /* if length len, return symbol */
+            return h->symbol[index + (code - first)];
+        index += count;                 /* else update for next length */
+        first += count;
+        first <<= 1;
+        code <<= 1;
+    }
+    return -10;                         /* ran out of codes */
+}
+
+/*
+ * A faster version of decode() for real applications of this code.   It's not
+ * as readable, but it makes puff() twice as fast.  And it only makes the code
+ * a few percent larger.
+ */
+#else /* !SLOW */
+local int decode(struct state *s, struct huffman *h)
+{
+    int len;            /* current number of bits in code */
+    int code;           /* len bits being decoded */
+    int first;          /* first code of length len */
+    int count;          /* number of codes of length len */
+    int index;          /* index of first code of length len in symbol table */
+    int bitbuf;         /* bits from stream */
+    int left;           /* bits left in next or left to process */
+    short *next;        /* next number of codes */
+
+    bitbuf = s->bitbuf;
+    left = s->bitcnt;
+    code = first = index = 0;
+    len = 1;
+    next = h->count + 1;
+    while (1) {
+        while (left--) {
+            code |= bitbuf & 1;
+            bitbuf >>= 1;
+            count = *next++;
+            if (code - count < first) { /* if length len, return symbol */
+                s->bitbuf = bitbuf;
+                s->bitcnt = (s->bitcnt - len) & 7;
+                return h->symbol[index + (code - first)];
+            }
+            index += count;             /* else update for next length */
+            first += count;
+            first <<= 1;
+            code <<= 1;
+            len++;
+        }
+        left = (MAXBITS+1) - len;
+        if (left == 0) break;
+        if (s->incnt == s->inlen) longjmp(s->env, 1);   /* out of input */
+        bitbuf = s->in[s->incnt++];
+        if (left > 8) left = 8;
+    }
+    return -10;                         /* ran out of codes */
+}
+#endif /* SLOW */
+
+/*
+ * Given the list of code lengths length[0..n-1] representing a canonical
+ * Huffman code for n symbols, construct the tables required to decode those
+ * codes.  Those tables are the number of codes of each length, and the symbols
+ * sorted by length, retaining their original order within each length.  The
+ * return value is zero for a complete code set, negative for an over-
+ * subscribed code set, and positive for an incomplete code set.  The tables
+ * can be used if the return value is zero or positive, but they cannot be used
+ * if the return value is negative.  If the return value is zero, it is not
+ * possible for decode() using that table to return an error--any stream of
+ * enough bits will resolve to a symbol.  If the return value is positive, then
+ * it is possible for decode() using that table to return an error for received
+ * codes past the end of the incomplete lengths.
+ *
+ * Not used by decode(), but used for error checking, h->count[0] is the number
+ * of the n symbols not in the code.  So n - h->count[0] is the number of
+ * codes.  This is useful for checking for incomplete codes that have more than
+ * one symbol, which is an error in a dynamic block.
+ *
+ * Assumption: for all i in 0..n-1, 0 <= length[i] <= MAXBITS
+ * This is assured by the construction of the length arrays in dynamic() and
+ * fixed() and is not verified by construct().
+ *
+ * Format notes:
+ *
+ * - Permitted and expected examples of incomplete codes are one of the fixed
+ *   codes and any code with a single symbol which in deflate is coded as one
+ *   bit instead of zero bits.  See the format notes for fixed() and dynamic().
+ *
+ * - Within a given code length, the symbols are kept in ascending order for
+ *   the code bits definition.
+ */
+local int construct(struct huffman *h, short *length, int n)
+{
+    int symbol;         /* current symbol when stepping through length[] */
+    int len;            /* current length when stepping through h->count[] */
+    int left;           /* number of possible codes left of current length */
+    short offs[MAXBITS+1];      /* offsets in symbol table for each length */
+
+    /* count number of codes of each length */
+    for (len = 0; len <= MAXBITS; len++)
+        h->count[len] = 0;
+    for (symbol = 0; symbol < n; symbol++)
+        (h->count[length[symbol]])++;   /* assumes lengths are within bounds */
+    if (h->count[0] == n)               /* no codes! */
+        return 0;                       /* complete, but decode() will fail */
+
+    /* check for an over-subscribed or incomplete set of lengths */
+    left = 1;                           /* one possible code of zero length */
+    for (len = 1; len <= MAXBITS; len++) {
+        left <<= 1;                     /* one more bit, double codes left */
+        left -= h->count[len];          /* deduct count from possible codes */
+        if (left < 0) return left;      /* over-subscribed--return negative */
+    }                                   /* left > 0 means incomplete */
+
+    /* generate offsets into symbol table for each length for sorting */
+    offs[1] = 0;
+    for (len = 1; len < MAXBITS; len++)
+        offs[len + 1] = offs[len] + h->count[len];
+
+    /*
+     * put symbols in table sorted by length, by symbol order within each
+     * length
+     */
+    for (symbol = 0; symbol < n; symbol++)
+        if (length[symbol] != 0)
+            h->symbol[offs[length[symbol]]++] = symbol;
+
+    /* return zero for complete set, positive for incomplete set */
+    return left;
+}
+
+/*
+ * Decode literal/length and distance codes until an end-of-block code.
+ *
+ * Format notes:
+ *
+ * - Compressed data that is after the block type if fixed or after the code
+ *   description if dynamic is a combination of literals and length/distance
+ *   pairs terminated by and end-of-block code.  Literals are simply Huffman
+ *   coded bytes.  A length/distance pair is a coded length followed by a
+ *   coded distance to represent a string that occurs earlier in the
+ *   uncompressed data that occurs again at the current location.
+ *
+ * - Literals, lengths, and the end-of-block code are combined into a single
+ *   code of up to 286 symbols.  They are 256 literals (0..255), 29 length
+ *   symbols (257..285), and the end-of-block symbol (256).
+ *
+ * - There are 256 possible lengths (3..258), and so 29 symbols are not enough
+ *   to represent all of those.  Lengths 3..10 and 258 are in fact represented
+ *   by just a length symbol.  Lengths 11..257 are represented as a symbol and
+ *   some number of extra bits that are added as an integer to the base length
+ *   of the length symbol.  The number of extra bits is determined by the base
+ *   length symbol.  These are in the static arrays below, lens[] for the base
+ *   lengths and lext[] for the corresponding number of extra bits.
+ *
+ * - The reason that 258 gets its own symbol is that the longest length is used
+ *   often in highly redundant files.  Note that 258 can also be coded as the
+ *   base value 227 plus the maximum extra value of 31.  While a good deflate
+ *   should never do this, it is not an error, and should be decoded properly.
+ *
+ * - If a length is decoded, including its extra bits if any, then it is
+ *   followed a distance code.  There are up to 30 distance symbols.  Again
+ *   there are many more possible distances (1..32768), so extra bits are added
+ *   to a base value represented by the symbol.  The distances 1..4 get their
+ *   own symbol, but the rest require extra bits.  The base distances and
+ *   corresponding number of extra bits are below in the static arrays dist[]
+ *   and dext[].
+ *
+ * - Literal bytes are simply written to the output.  A length/distance pair is
+ *   an instruction to copy previously uncompressed bytes to the output.  The
+ *   copy is from distance bytes back in the output stream, copying for length
+ *   bytes.
+ *
+ * - Distances pointing before the beginning of the output data are not
+ *   permitted.
+ *
+ * - Overlapped copies, where the length is greater than the distance, are
+ *   allowed and common.  For example, a distance of one and a length of 258
+ *   simply copies the last byte 258 times.  A distance of four and a length of
+ *   twelve copies the last four bytes three times.  A simple forward copy
+ *   ignoring whether the length is greater than the distance or not implements
+ *   this correctly.  You should not use memcpy() since its behavior is not
+ *   defined for overlapped arrays.  You should not use memmove() or bcopy()
+ *   since though their behavior -is- defined for overlapping arrays, it is
+ *   defined to do the wrong thing in this case.
+ */
+local int codes(struct state *s,
+                struct huffman *lencode,
+                struct huffman *distcode)
+{
+    int symbol;         /* decoded symbol */
+    int len;            /* length for copy */
+    unsigned dist;      /* distance for copy */
+    static const short lens[29] = { /* Size base for length codes 257..285 */
+        3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
+        35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258};
+    static const short lext[29] = { /* Extra bits for length codes 257..285 */
+        0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
+        3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0};
+    static const short dists[30] = { /* Offset base for distance codes 0..29 */
+        1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
+        257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
+        8193, 12289, 16385, 24577};
+    static const short dext[30] = { /* Extra bits for distance codes 0..29 */
+        0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
+        7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
+        12, 12, 13, 13};
+
+    /* decode literals and length/distance pairs */
+    do {
+        symbol = decode(s, lencode);
+        if (symbol < 0) return symbol;  /* invalid symbol */
+        if (symbol < 256) {             /* literal: symbol is the byte */
+            /* write out the literal */
+            if (s->out != NIL) {
+                if (s->outcnt == s->outlen) return 1;
+                s->out[s->outcnt] = symbol;
+            }
+            s->outcnt++;
+        }
+        else if (symbol > 256) {        /* length */
+            /* get and compute length */
+            symbol -= 257;
+            if (symbol >= 29) return -10;       /* invalid fixed code */
+            len = lens[symbol] + bits(s, lext[symbol]);
+
+            /* get and check distance */
+            symbol = decode(s, distcode);
+            if (symbol < 0) return symbol;      /* invalid symbol */
+            dist = dists[symbol] + bits(s, dext[symbol]);
+#ifndef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+            if (dist > s->outcnt)
+                return -11;     /* distance too far back */
+#endif
+
+            /* copy length bytes from distance bytes back */
+            if (s->out != NIL) {
+                if (s->outcnt + len > s->outlen) return 1;
+                while (len--) {
+                    s->out[s->outcnt] =
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+                        dist > s->outcnt ? 0 :
+#endif
+                        s->out[s->outcnt - dist];
+                    s->outcnt++;
+                }
+            }
+            else
+                s->outcnt += len;
+        }
+    } while (symbol != 256);            /* end of block symbol */
+
+    /* done with a valid fixed or dynamic block */
+    return 0;
+}
+
+/*
+ * Process a fixed codes block.
+ *
+ * Format notes:
+ *
+ * - This block type can be useful for compressing small amounts of data for
+ *   which the size of the code descriptions in a dynamic block exceeds the
+ *   benefit of custom codes for that block.  For fixed codes, no bits are
+ *   spent on code descriptions.  Instead the code lengths for literal/length
+ *   codes and distance codes are fixed.  The specific lengths for each symbol
+ *   can be seen in the "for" loops below.
+ *
+ * - The literal/length code is complete, but has two symbols that are invalid
+ *   and should result in an error if received.  This cannot be implemented
+ *   simply as an incomplete code since those two symbols are in the "middle"
+ *   of the code.  They are eight bits long and the longest literal/length\
+ *   code is nine bits.  Therefore the code must be constructed with those
+ *   symbols, and the invalid symbols must be detected after decoding.
+ *
+ * - The fixed distance codes also have two invalid symbols that should result
+ *   in an error if received.  Since all of the distance codes are the same
+ *   length, this can be implemented as an incomplete code.  Then the invalid
+ *   codes are detected while decoding.
+ */
+local int fixed(struct state *s)
+{
+    static int virgin = 1;
+    static short lencnt[MAXBITS+1], lensym[FIXLCODES];
+    static short distcnt[MAXBITS+1], distsym[MAXDCODES];
+    static struct huffman lencode, distcode;
+
+    /* build fixed huffman tables if first call (may not be thread safe) */
+    if (virgin) {
+        int symbol;
+        short lengths[FIXLCODES];
+
+        /* literal/length table */
+        for (symbol = 0; symbol < 144; symbol++)
+            lengths[symbol] = 8;
+        for (; symbol < 256; symbol++)
+            lengths[symbol] = 9;
+        for (; symbol < 280; symbol++)
+            lengths[symbol] = 7;
+        for (; symbol < FIXLCODES; symbol++)
+            lengths[symbol] = 8;
+        construct(&lencode, lengths, FIXLCODES);
+
+        /* distance table */
+        for (symbol = 0; symbol < MAXDCODES; symbol++)
+            lengths[symbol] = 5;
+        construct(&distcode, lengths, MAXDCODES);
+
+        /* construct lencode and distcode */
+        lencode.count = lencnt;
+        lencode.symbol = lensym;
+        distcode.count = distcnt;
+        distcode.symbol = distsym;
+
+        /* do this just once */
+        virgin = 0;
+    }
+
+    /* decode data until end-of-block code */
+    return codes(s, &lencode, &distcode);
+}
+
+/*
+ * Process a dynamic codes block.
+ *
+ * Format notes:
+ *
+ * - A dynamic block starts with a description of the literal/length and
+ *   distance codes for that block.  New dynamic blocks allow the compressor to
+ *   rapidly adapt to changing data with new codes optimized for that data.
+ *
+ * - The codes used by the deflate format are "canonical", which means that
+ *   the actual bits of the codes are generated in an unambiguous way simply
+ *   from the number of bits in each code.  Therefore the code descriptions
+ *   are simply a list of code lengths for each symbol.
+ *
+ * - The code lengths are stored in order for the symbols, so lengths are
+ *   provided for each of the literal/length symbols, and for each of the
+ *   distance symbols.
+ *
+ * - If a symbol is not used in the block, this is represented by a zero as
+ *   as the code length.  This does not mean a zero-length code, but rather
+ *   that no code should be created for this symbol.  There is no way in the
+ *   deflate format to represent a zero-length code.
+ *
+ * - The maximum number of bits in a code is 15, so the possible lengths for
+ *   any code are 1..15.
+ *
+ * - The fact that a length of zero is not permitted for a code has an
+ *   interesting consequence.  Normally if only one symbol is used for a given
+ *   code, then in fact that code could be represented with zero bits.  However
+ *   in deflate, that code has to be at least one bit.  So for example, if
+ *   only a single distance base symbol appears in a block, then it will be
+ *   represented by a single code of length one, in particular one 0 bit.  This
+ *   is an incomplete code, since if a 1 bit is received, it has no meaning,
+ *   and should result in an error.  So incomplete distance codes of one symbol
+ *   should be permitted, and the receipt of invalid codes should be handled.
+ *
+ * - It is also possible to have a single literal/length code, but that code
+ *   must be the end-of-block code, since every dynamic block has one.  This
+ *   is not the most efficient way to create an empty block (an empty fixed
+ *   block is fewer bits), but it is allowed by the format.  So incomplete
+ *   literal/length codes of one symbol should also be permitted.
+ *
+ * - If there are only literal codes and no lengths, then there are no distance
+ *   codes.  This is represented by one distance code with zero bits.
+ *
+ * - The list of up to 286 length/literal lengths and up to 30 distance lengths
+ *   are themselves compressed using Huffman codes and run-length encoding.  In
+ *   the list of code lengths, a 0 symbol means no code, a 1..15 symbol means
+ *   that length, and the symbols 16, 17, and 18 are run-length instructions.
+ *   Each of 16, 17, and 18 are follwed by extra bits to define the length of
+ *   the run.  16 copies the last length 3 to 6 times.  17 represents 3 to 10
+ *   zero lengths, and 18 represents 11 to 138 zero lengths.  Unused symbols
+ *   are common, hence the special coding for zero lengths.
+ *
+ * - The symbols for 0..18 are Huffman coded, and so that code must be
+ *   described first.  This is simply a sequence of up to 19 three-bit values
+ *   representing no code (0) or the code length for that symbol (1..7).
+ *
+ * - A dynamic block starts with three fixed-size counts from which is computed
+ *   the number of literal/length code lengths, the number of distance code
+ *   lengths, and the number of code length code lengths (ok, you come up with
+ *   a better name!) in the code descriptions.  For the literal/length and
+ *   distance codes, lengths after those provided are considered zero, i.e. no
+ *   code.  The code length code lengths are received in a permuted order (see
+ *   the order[] array below) to make a short code length code length list more
+ *   likely.  As it turns out, very short and very long codes are less likely
+ *   to be seen in a dynamic code description, hence what may appear initially
+ *   to be a peculiar ordering.
+ *
+ * - Given the number of literal/length code lengths (nlen) and distance code
+ *   lengths (ndist), then they are treated as one long list of nlen + ndist
+ *   code lengths.  Therefore run-length coding can and often does cross the
+ *   boundary between the two sets of lengths.
+ *
+ * - So to summarize, the code description at the start of a dynamic block is
+ *   three counts for the number of code lengths for the literal/length codes,
+ *   the distance codes, and the code length codes.  This is followed by the
+ *   code length code lengths, three bits each.  This is used to construct the
+ *   code length code which is used to read the remainder of the lengths.  Then
+ *   the literal/length code lengths and distance lengths are read as a single
+ *   set of lengths using the code length codes.  Codes are constructed from
+ *   the resulting two sets of lengths, and then finally you can start
+ *   decoding actual compressed data in the block.
+ *
+ * - For reference, a "typical" size for the code description in a dynamic
+ *   block is around 80 bytes.
+ */
+local int dynamic(struct state *s)
+{
+    int nlen, ndist, ncode;             /* number of lengths in descriptor */
+    int index;                          /* index of lengths[] */
+    int err;                            /* construct() return value */
+    short lengths[MAXCODES];            /* descriptor code lengths */
+    short lencnt[MAXBITS+1], lensym[MAXLCODES];         /* lencode memory */
+    short distcnt[MAXBITS+1], distsym[MAXDCODES];       /* distcode memory */
+    struct huffman lencode, distcode;   /* length and distance codes */
+    static const short order[19] =      /* permutation of code length codes */
+        {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+    /* construct lencode and distcode */
+    lencode.count = lencnt;
+    lencode.symbol = lensym;
+    distcode.count = distcnt;
+    distcode.symbol = distsym;
+
+    /* get number of lengths in each table, check lengths */
+    nlen = bits(s, 5) + 257;
+    ndist = bits(s, 5) + 1;
+    ncode = bits(s, 4) + 4;
+    if (nlen > MAXLCODES || ndist > MAXDCODES)
+        return -3;                      /* bad counts */
+
+    /* read code length code lengths (really), missing lengths are zero */
+    for (index = 0; index < ncode; index++)
+        lengths[order[index]] = bits(s, 3);
+    for (; index < 19; index++)
+        lengths[order[index]] = 0;
+
+    /* build huffman table for code lengths codes (use lencode temporarily) */
+    err = construct(&lencode, lengths, 19);
+    if (err != 0) return -4;            /* require complete code set here */
+
+    /* read length/literal and distance code length tables */
+    index = 0;
+    while (index < nlen + ndist) {
+        int symbol;             /* decoded value */
+        int len;                /* last length to repeat */
+
+        symbol = decode(s, &lencode);
+        if (symbol < 16)                /* length in 0..15 */
+            lengths[index++] = symbol;
+        else {                          /* repeat instruction */
+            len = 0;                    /* assume repeating zeros */
+            if (symbol == 16) {         /* repeat last length 3..6 times */
+                if (index == 0) return -5;      /* no last length! */
+                len = lengths[index - 1];       /* last length */
+                symbol = 3 + bits(s, 2);
+            }
+            else if (symbol == 17)      /* repeat zero 3..10 times */
+                symbol = 3 + bits(s, 3);
+            else                        /* == 18, repeat zero 11..138 times */
+                symbol = 11 + bits(s, 7);
+            if (index + symbol > nlen + ndist)
+                return -6;              /* too many lengths! */
+            while (symbol--)            /* repeat last or zero symbol times */
+                lengths[index++] = len;
+        }
+    }
+
+    /* check for end-of-block code -- there better be one! */
+    if (lengths[256] == 0)
+        return -9;
+
+    /* build huffman table for literal/length codes */
+    err = construct(&lencode, lengths, nlen);
+    if (err < 0 || (err > 0 && nlen - lencode.count[0] != 1))
+        return -7;      /* only allow incomplete codes if just one code */
+
+    /* build huffman table for distance codes */
+    err = construct(&distcode, lengths + nlen, ndist);
+    if (err < 0 || (err > 0 && ndist - distcode.count[0] != 1))
+        return -8;      /* only allow incomplete codes if just one code */
+
+    /* decode data until end-of-block code */
+    return codes(s, &lencode, &distcode);
+}
+
+/*
+ * Inflate source to dest.  On return, destlen and sourcelen are updated to the
+ * size of the uncompressed data and the size of the deflate data respectively.
+ * On success, the return value of puff() is zero.  If there is an error in the
+ * source data, i.e. it is not in the deflate format, then a negative value is
+ * returned.  If there is not enough input available or there is not enough
+ * output space, then a positive error is returned.  In that case, destlen and
+ * sourcelen are not updated to facilitate retrying from the beginning with the
+ * provision of more input data or more output space.  In the case of invalid
+ * inflate data (a negative error), the dest and source pointers are updated to
+ * facilitate the debugging of deflators.
+ *
+ * puff() also has a mode to determine the size of the uncompressed output with
+ * no output written.  For this dest must be (unsigned char *)0.  In this case,
+ * the input value of *destlen is ignored, and on return *destlen is set to the
+ * size of the uncompressed output.
+ *
+ * The return codes are:
+ *
+ *   2:  available inflate data did not terminate
+ *   1:  output space exhausted before completing inflate
+ *   0:  successful inflate
+ *  -1:  invalid block type (type == 3)
+ *  -2:  stored block length did not match one's complement
+ *  -3:  dynamic block code description: too many length or distance codes
+ *  -4:  dynamic block code description: code lengths codes incomplete
+ *  -5:  dynamic block code description: repeat lengths with no first length
+ *  -6:  dynamic block code description: repeat more than specified lengths
+ *  -7:  dynamic block code description: invalid literal/length code lengths
+ *  -8:  dynamic block code description: invalid distance code lengths
+ *  -9:  dynamic block code description: missing end-of-block code
+ * -10:  invalid literal/length or distance code in fixed or dynamic block
+ * -11:  distance is too far back in fixed or dynamic block
+ *
+ * Format notes:
+ *
+ * - Three bits are read for each block to determine the kind of block and
+ *   whether or not it is the last block.  Then the block is decoded and the
+ *   process repeated if it was not the last block.
+ *
+ * - The leftover bits in the last byte of the deflate data after the last
+ *   block (if it was a fixed or dynamic block) are undefined and have no
+ *   expected values to check.
+ */
+int puff(unsigned char *dest,           /* pointer to destination pointer */
+         unsigned long *destlen,        /* amount of output space */
+         const unsigned char *source,         /* pointer to source data pointer */
+         unsigned long *sourcelen)      /* amount of input available */
+{
+    struct state s;             /* input/output state */
+    int last, type;             /* block information */
+    int err;                    /* return value */
+
+    /* initialize output state */
+    s.out = dest;
+    s.outlen = *destlen;                /* ignored if dest is NIL */
+    s.outcnt = 0;
+
+    /* initialize input state */
+    s.in = source;
+    s.inlen = *sourcelen;
+    s.incnt = 0;
+    s.bitbuf = 0;
+    s.bitcnt = 0;
+
+    /* return if bits() or decode() tries to read past available input */
+    if (setjmp(s.env) != 0)             /* if came back here via longjmp() */
+        err = 2;                        /* then skip do-loop, return error */
+    else {
+        /* process blocks until last block or error */
+        do {
+            last = bits(&s, 1);         /* one if last block */
+            type = bits(&s, 2);         /* block type 0..3 */
+            err = type == 0 ? stored(&s) :
+                  (type == 1 ? fixed(&s) :
+                   (type == 2 ? dynamic(&s) :
+                    -1));               /* type == 3, invalid */
+            if (err != 0) break;        /* return with error */
+        } while (!last);
+    }
+
+    /* update the lengths and return */
+    if (err <= 0) {
+        *destlen = s.outcnt;
+        *sourcelen = s.incnt;
+    }
+    return err;
+}
+
+#ifdef TEST
+/* Examples of how to use puff().
+
+   Usage: puff [-w] [-nnn] file
+          ... | puff [-w] [-nnn]
+
+   where file is the input file with deflate data, nnn is the number of bytes
+   of input to skip before inflating (e.g. to skip a zlib or gzip header), and
+   -w is used to write the decompressed data to stdout */
+
+#include <stdio.h>
+#include <stdlib.h>
+
+/* Return size times approximately the cube root of 2, keeping the result as 1,
+   3, or 5 times a power of 2 -- the result is always > size, until the result
+   is the maximum value of an unsigned long, where it remains.  This is useful
+   to keep reallocations less than ~33% over the actual data. */
+local size_t bythirds(size_t size)
+{
+    int n;
+    size_t m;
+
+    m = size;
+    for (n = 0; m; n++)
+        m >>= 1;
+    if (n < 3)
+        return size + 1;
+    n -= 3;
+    m = size >> n;
+    m += m == 6 ? 2 : 1;
+    m <<= n;
+    return m > size ? m : (size_t)(-1);
+}
+
+/* Read the input file *name, or stdin if name is NULL, into allocated memory.
+   Reallocate to larger buffers until the entire file is read in.  Return a
+   pointer to the allocated data, or NULL if there was a memory allocation
+   failure.  *len is the number of bytes of data read from the input file (even
+   if load() returns NULL).  If the input file was empty or could not be opened
+   or read, *len is zero. */
+local void *load(char *name, size_t *len)
+{
+    size_t size;
+    void *buf, *swap;
+    FILE *in;
+
+    *len = 0;
+    buf = malloc(size = 4096);
+    if (buf == NULL)
+        return NULL;
+    in = name == NULL ? stdin : fopen(name, "rb");
+    if (in != NULL) {
+        for (;;) {
+            *len += fread((char *)buf + *len, 1, size - *len, in);
+            if (*len < size) break;
+            size = bythirds(size);
+            if (size == *len || (swap = realloc(buf, size)) == NULL) {
+                free(buf);
+                buf = NULL;
+                break;
+            }
+            buf = swap;
+        }
+        fclose(in);
+    }
+    return buf;
+}
+
+int main(int argc, char **argv)
+{
+    int ret, put = 0;
+    unsigned skip = 0;
+    char *arg, *name = NULL;
+    unsigned char *source = NULL, *dest;
+    size_t len = 0;
+    unsigned long sourcelen, destlen;
+
+    /* process arguments */
+    while (arg = *++argv, --argc)
+        if (arg[0] == '-') {
+            if (arg[1] == 'w' && arg[2] == 0)
+                put = 1;
+            else if (arg[1] >= '0' && arg[1] <= '9')
+                skip = (unsigned)atoi(arg + 1);
+            else {
+                fprintf(stderr, "invalid option %s\n", arg);
+                return 3;
+            }
+        }
+        else if (name != NULL) {
+            fprintf(stderr, "only one file name allowed\n");
+            return 3;
+        }
+        else
+            name = arg;
+    source = load(name, &len);
+    if (source == NULL) {
+        fprintf(stderr, "memory allocation failure\n");
+        return 4;
+    }
+    if (len == 0) {
+        fprintf(stderr, "could not read %s, or it was empty\n",
+                name == NULL ? "<stdin>" : name);
+        free(source);
+        return 3;
+    }
+    if (skip >= len) {
+        fprintf(stderr, "skip request of %d leaves no input\n", skip);
+        free(source);
+        return 3;
+    }
+
+    /* test inflate data with offset skip */
+    len -= skip;
+    sourcelen = (unsigned long)len;
+    ret = puff(NIL, &destlen, source + skip, &sourcelen);
+    if (ret)
+        fprintf(stderr, "puff() failed with return code %d\n", ret);
+    else {
+        fprintf(stderr, "puff() succeeded uncompressing %lu bytes\n", destlen);
+        if (sourcelen < len) fprintf(stderr, "%lu compressed bytes unused\n",
+                                     len - sourcelen);
+    }
+
+    /* if requested, inflate again and write decompressd data to stdout */
+    if (put) {
+        dest = malloc(destlen);
+        if (dest == NULL) {
+            fprintf(stderr, "memory allocation failure\n");
+            free(source);
+            return 4;
+        }
+        puff(dest, &destlen, source + skip, &sourcelen);
+        fwrite(dest, 1, destlen, stdout);
+        free(dest);
+    }
+
+    /* clean up */
+    free(source);
+    return ret;
+}
+#endif
diff --git a/redist/puff.h b/redist/puff.h
new file mode 100644
index 0000000..eef7139
--- /dev/null
+++ b/redist/puff.h
@@ -0,0 +1,31 @@
+/* puff.h
+  Copyright (C) 2002-2010 Mark Adler, all rights reserved
+  version 2.1, 4 Apr 2010
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the author be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Mark Adler    madler@alumni.caltech.edu
+ */
+
+
+/*
+ * See puff.c for purpose and usage.
+ */
+int puff(unsigned char *dest,           /* pointer to destination pointer */
+         unsigned long *destlen,        /* amount of output space */
+         const unsigned char *source,         /* pointer to source data pointer */
+         unsigned long *sourcelen);     /* amount of input available */
\ No newline at end of file
diff --git a/redist/symbol_enumerator.c b/redist/symbol_enumerator.c
new file mode 100644
index 0000000..eddddb9
--- /dev/null
+++ b/redist/symbol_enumerator.c
@@ -0,0 +1,241 @@
+#include <stdio.h>
+#include "symbol_enumerator.h"
+
+#if defined( WIN32 ) || defined( WINDOWS ) || defined( USE_WINDOWS )
+
+#include <windows.h>
+
+typedef struct _SYMBOL_INFO {
+  ULONG   SizeOfStruct;
+  ULONG   TypeIndex;
+  ULONG64 Reserved[2];
+  ULONG   Index;
+  ULONG   Size;
+  ULONG64 ModBase;
+  ULONG   Flags;
+  ULONG64 Value;
+  ULONG64 Address;
+  ULONG   Register;
+  ULONG   Scope;
+  ULONG   Tag;
+  ULONG   NameLen;
+  ULONG   MaxNameLen;
+  TCHAR   Name[1];
+} SYMBOL_INFO, *PSYMBOL_INFO;
+typedef struct _IMAGEHLP_STACK_FRAME {
+  ULONG64 InstructionOffset;
+  ULONG64 ReturnOffset;
+  ULONG64 FrameOffset;
+  ULONG64 StackOffset;
+  ULONG64 BackingStoreOffset;
+  ULONG64 FuncTableEntry;
+  ULONG64 Params[4];
+  ULONG64 Reserved[5];
+  BOOL    Virtual;
+  ULONG   Reserved2;
+} IMAGEHLP_STACK_FRAME, *PIMAGEHLP_STACK_FRAME;
+
+
+typedef BOOL (*PSYM_ENUMERATESYMBOLS_CALLBACK)(
+  PSYMBOL_INFO pSymInfo,
+  ULONG        SymbolSize,
+  PVOID        UserContext
+  );
+  
+BOOL WINAPI SymEnumSymbols(
+  HANDLE                         hProcess,
+  ULONG64                        BaseOfDll,
+  PCTSTR                         Mask,
+  PSYM_ENUMERATESYMBOLS_CALLBACK EnumSymbolsCallback,
+  const PVOID                          UserContext
+);
+
+BOOL WINAPI SymInitialize(
+  HANDLE hProcess,
+  PCTSTR UserSearchPath,
+  BOOL   fInvadeProcess
+);
+
+BOOL CALLBACK __cdecl mycb(
+  PSYMBOL_INFO pSymInfo,
+  ULONG        SymbolSize,
+  PVOID        UserContext
+  ){
+	  SymEnumeratorCallback cb = (SymEnumeratorCallback)UserContext;
+	  return !cb( "", &pSymInfo->Name[0], (void*)pSymInfo->Address, (long) pSymInfo->Size );
+  }
+  
+int EnumerateSymbols( SymEnumeratorCallback cb )
+{
+	HANDLE proc = GetCurrentProcess();
+	if( !SymInitialize( proc, 0, 1 ) ) return -1;
+	if( !SymEnumSymbols( proc, 0, "*!*", &mycb, (void*)cb ) )
+	{
+		SymCleanup(proc);
+		return -2;
+	}
+	SymCleanup(proc);
+	return 0;
+}
+
+#else
+	
+
+#include <stdio.h>
+#include <dlfcn.h>
+#include <stdint.h>
+#include <limits.h>
+#include <string.h>
+
+#ifndef __GNUC__
+#define __int128_t long long long
+#endif
+
+#include <link.h>
+#include <elf.h>
+
+#define UINTS_PER_WORD (__WORDSIZE / (CHAR_BIT * sizeof (unsigned int)))
+
+
+    struct dl_phdr_info {
+        ElfW(Addr)        dlpi_addr;  /* Base address of object */
+        const char       *dlpi_name;  /* (Null-terminated) name of
+                                         object */
+        const ElfW(Phdr) *dlpi_phdr;  /* Pointer to array of
+                                         ELF program headers
+                                         for this object */
+        ElfW(Half)        dlpi_phnum; /* # of items in dlpi_phdr */
+    };
+
+
+
+void dl_iterate_phdr( void*, void*);
+
+
+static ElfW(Word) gnu_hashtab_symbol_count(const unsigned int *const table)
+{
+    const unsigned int *const bucket = table + 4 + table[2] * (unsigned int)(UINTS_PER_WORD);
+    unsigned int              b = table[0];
+    unsigned int              max = 0U;
+
+    while (b-->0U)
+        if (bucket[b] > max)
+            max = bucket[b];
+
+    return (ElfW(Word))max;
+}
+
+static static void *dynamic_pointer(const ElfW(Addr) addr,
+                             const ElfW(Addr) base, const ElfW(Phdr) *const header, const ElfW(Half) headers)
+{
+    if (addr) {
+        ElfW(Half) h;
+
+        for (h = 0; h < headers; h++)
+            if (header[h].p_type == PT_LOAD)
+                if (addr >= base + header[h].p_vaddr &&
+                    addr <  base + header[h].p_vaddr + header[h].p_memsz)
+                    return (void *)addr;
+    }
+
+    return NULL;
+}
+
+//Mostly based off of http://stackoverflow.com/questions/29903049/get-names-and-addresses-of-exported-functions-from-in-linux
+static int callback(struct dl_phdr_info *info,
+               size_t size, void *data)
+{
+	SymEnumeratorCallback cb = (SymEnumeratorCallback)data;
+	const ElfW(Addr)                 base = info->dlpi_addr;
+	const ElfW(Phdr) *const          header = info->dlpi_phdr;
+	const ElfW(Half)                 headers = info->dlpi_phnum;
+	const char                      *libpath, *libname;
+	ElfW(Half)                       h;
+
+	if (info->dlpi_name && info->dlpi_name[0])
+		libpath = info->dlpi_name;
+	else
+		libpath = "";
+
+	libname = strrchr(libpath, '/');
+
+	if (libname && libname[0] == '/' && libname[1])
+		libname++;
+	else
+		libname = libpath;
+
+	for (h = 0; h < headers; h++)
+	{
+		if (header[h].p_type == PT_DYNAMIC)
+		{
+			const ElfW(Dyn)  *entry = (const ElfW(Dyn) *)(base + header[h].p_vaddr);
+			const ElfW(Word) *hashtab;
+			const ElfW(Sym)  *symtab = NULL;
+			const char       *strtab = NULL;
+			ElfW(Word)        symbol_count = 0;
+
+			for (; entry->d_tag != DT_NULL; entry++)
+			{
+				switch (entry->d_tag)
+				{
+				case DT_HASH:
+					hashtab = dynamic_pointer(entry->d_un.d_ptr, base, header, headers);
+					if (hashtab)
+						symbol_count = hashtab[1];
+					break;
+				case DT_GNU_HASH:
+					hashtab = dynamic_pointer(entry->d_un.d_ptr, base, header, headers);
+					if (hashtab)
+					{
+						ElfW(Word) count = gnu_hashtab_symbol_count(hashtab);
+						if (count > symbol_count)
+							symbol_count = count;
+					}
+					break;
+				case DT_STRTAB:
+					strtab = dynamic_pointer(entry->d_un.d_ptr, base, header, headers);
+					break;
+				case DT_SYMTAB:
+					symtab = dynamic_pointer(entry->d_un.d_ptr, base, header, headers);
+					break;
+				}
+			}
+
+			if (symtab && strtab && symbol_count > 0) {
+				ElfW(Word)  s;
+
+				for (s = 0; s < symbol_count; s++) {
+					const char *name;
+					void *const ptr = dynamic_pointer(base + symtab[s].st_value, base, header, headers);
+					int         result;
+
+					if (!ptr)
+						continue;
+
+					if (symtab[s].st_name)
+						name = strtab + symtab[s].st_name;
+					else
+						name = "";
+
+					result = cb( libpath, name, ptr, symtab[s].st_size );
+					if( result ) return result; //Bail early.
+				}
+			}
+		}
+	}
+	return 0;
+}
+
+
+
+
+
+int EnumerateSymbols( SymEnumeratorCallback cb )
+{
+	dl_iterate_phdr( callback, cb );
+}
+
+
+
+
+#endif
diff --git a/redist/symbol_enumerator.h b/redist/symbol_enumerator.h
new file mode 100644
index 0000000..87e033b
--- /dev/null
+++ b/redist/symbol_enumerator.h
@@ -0,0 +1,12 @@
+#ifndef _SYMBOL_ENUMERATOR_H
+#define _SYMBOL_ENUMERATOR_H
+
+//Enumerates all symbols in the currently loaded excutable.
+//Don't forget to compile with -rdynamic!
+
+//Return 0 to continue search.  1 to stop.
+typedef int (*SymEnumeratorCallback)( const char * path, const char * name, void * location, long size );
+
+int EnumerateSymbols( SymEnumeratorCallback cb );
+
+#endif