Updated stb libs to latest version

src/external/stb_image.h

@@ -1,4 +1,4 @@
-/* stb_image - v2.15 - public domain image loader - http://nothings.org/stb_image.h
+/* stb_image - v2.16 - public domain image loader - http://nothings.org/stb_image.h
                                      no warranty implied; use at your own risk
 
    Do this:
@@ -48,6 +48,7 @@ LICENSE
 
 RECENT REVISION HISTORY:
 
+      2.16  (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes
       2.15  (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC
       2.14  (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs
       2.13  (2016-12-04) experimental 16-bit API, only for PNG so far; fixes
@@ -58,10 +59,6 @@ RECENT REVISION HISTORY:
                          correct channel count for PNG & BMP
       2.10  (2016-01-22) avoid warning introduced in 2.09
       2.09  (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED
-      2.08  (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA
-      2.07  (2015-09-13) partial animated GIF support
-                         limited 16-bit PSD support
-                         minor bugs, code cleanup, and compiler warnings
 
    See end of file for full revision history.
 
@@ -83,6 +80,7 @@ RECENT REVISION HISTORY:
  Optimizations & bugfixes
     Fabian "ryg" Giesen
     Arseny Kapoulkine
+    John-Mark Allen
 
  Bug & warning fixes
     Marc LeBlanc            David Woo          Guillaume George   Martins Mozeiko
@@ -98,7 +96,7 @@ RECENT REVISION HISTORY:
     Michaelangel007@github  Philipp Wiesemann  Dale Weiler        github:grim210
     Oriol Ferrer Mesia      Josh Tobin         Matthew Gregan     github:sammyhw
     Blazej Dariusz Roszkowski                  Gregory Mullen     github:phprus
-
+    Christian Floisand      Kevin Schmidt                         github:poppolopoppo
 */
 
 #ifndef STBI_INCLUDE_STB_IMAGE_H
@@ -134,11 +132,12 @@ RECENT REVISION HISTORY:
 // with each pixel consisting of N interleaved 8-bit components; the first
 // pixel pointed to is top-left-most in the image. There is no padding between
 // image scanlines or between pixels, regardless of format. The number of
-// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise.
-// If req_comp is non-zero, *comp has the number of components that _would_
-// have been output otherwise. E.g. if you set req_comp to 4, you will always
-// get RGBA output, but you can check *comp to see if it's trivially opaque
-// because e.g. there were only 3 channels in the source image.
+// components N is 'desired_channels' if desired_channels is non-zero, or
+// *channels_in_file otherwise. If desired_channels is non-zero,
+// *channels_in_file has the number of components that _would_ have been
+// output otherwise. E.g. if you set desired_channels to 4, you will always
+// get RGBA output, but you can check *channels_in_file to see if it's trivially
+// opaque because e.g. there were only 3 channels in the source image.
 //
 // An output image with N components has the following components interleaved
 // in this order in each pixel:
@@ -150,10 +149,10 @@ RECENT REVISION HISTORY:
 //       4           red, green, blue, alpha
 //
 // If image loading fails for any reason, the return value will be NULL,
-// and *x, *y, *comp will be unchanged. The function stbi_failure_reason()
-// can be queried for an extremely brief, end-user unfriendly explanation
-// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid
-// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly
+// and *x, *y, *channels_in_file will be unchanged. The function
+// stbi_failure_reason() can be queried for an extremely brief, end-user
+// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS
+// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly
 // more user-friendly ones.
 //
 // Paletted PNG, BMP, GIF, and PIC images are automatically depalettized.
@@ -310,7 +309,7 @@ RECENT REVISION HISTORY:
 
 enum
 {
-   STBI_default = 0, // only used for req_comp
+   STBI_default = 0, // only used for desired_channels
 
    STBI_grey       = 1,
    STBI_grey_alpha = 2,
@@ -352,12 +351,12 @@ typedef struct
 // 8-bits-per-channel interface
 //
 
-STBIDEF stbi_uc *stbi_load               (char              const *filename,           int *x, int *y, int *channels_in_file, int desired_channels);
 STBIDEF stbi_uc *stbi_load_from_memory   (stbi_uc           const *buffer, int len   , int *x, int *y, int *channels_in_file, int desired_channels);
 STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk  , void *user, int *x, int *y, int *channels_in_file, int desired_channels);
 
 #ifndef STBI_NO_STDIO
-STBIDEF stbi_uc *stbi_load_from_file   (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load            (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_uc *stbi_load_from_file  (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
 // for stbi_load_from_file, file pointer is left pointing immediately after image
 #endif
 
@@ -366,22 +365,24 @@ STBIDEF stbi_uc *stbi_load_from_file   (FILE *f, int *x, int *y, int *channels_i
 // 16-bits-per-channel interface
 //
 
-STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_16_from_memory   (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels);
+
 #ifndef STBI_NO_STDIO
+STBIDEF stbi_us *stbi_load_16          (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
 STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
 #endif
-// @TODO the other variants
 
 ////////////////////////////////////
 //
 // float-per-channel interface
 //
 #ifndef STBI_NO_LINEAR
-   STBIDEF float *stbi_loadf                 (char const *filename,           int *x, int *y, int *channels_in_file, int desired_channels);
    STBIDEF float *stbi_loadf_from_memory     (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels);
    STBIDEF float *stbi_loadf_from_callbacks  (stbi_io_callbacks const *clbk, void *user, int *x, int *y,  int *channels_in_file, int desired_channels);
 
    #ifndef STBI_NO_STDIO
+   STBIDEF float *stbi_loadf            (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels);
    STBIDEF float *stbi_loadf_from_file  (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels);
    #endif
 #endif
@@ -639,7 +640,7 @@ static int stbi__cpuid3(void)
 
 #define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name
 
-static int stbi__sse2_available()
+static int stbi__sse2_available(void)
 {
    int info3 = stbi__cpuid3();
    return ((info3 >> 26) & 1) != 0;
@@ -647,7 +648,7 @@ static int stbi__sse2_available()
 #else // assume GCC-style if not VC++
 #define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16)))
 
-static int stbi__sse2_available()
+static int stbi__sse2_available(void)
 {
    // If we're even attempting to compile this on GCC/Clang, that means
    // -msse2 is on, which means the compiler is allowed to use SSE2
@@ -1029,6 +1030,30 @@ static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int chan
    return enlarged;
 }
 
+static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel)
+{
+   int row;
+   size_t bytes_per_row = (size_t)w * bytes_per_pixel;
+   stbi_uc temp[2048];
+   stbi_uc *bytes = (stbi_uc *)image;
+
+   for (row = 0; row < (h>>1); row++) {
+      stbi_uc *row0 = bytes + row*bytes_per_row;
+      stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row;
+      // swap row0 with row1
+      size_t bytes_left = bytes_per_row;
+      while (bytes_left) {
+         size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp);
+         memcpy(temp, row0, bytes_copy);
+         memcpy(row0, row1, bytes_copy);
+         memcpy(row1, temp, bytes_copy);
+         row0 += bytes_copy;
+         row1 += bytes_copy;
+         bytes_left -= bytes_copy;
+      }
+   }
+}
+
 static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp)
 {
    stbi__result_info ri;
@@ -1046,21 +1071,8 @@ static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x,
    // @TODO: move stbi__convert_format to here
 
    if (stbi__vertically_flip_on_load) {
-      int w = *x, h = *y;
       int channels = req_comp ? req_comp : *comp;
-      int row,col,z;
-      stbi_uc *image = (stbi_uc *) result;
-
-      // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once
-      for (row = 0; row < (h>>1); row++) {
-         for (col = 0; col < w; col++) {
-            for (z = 0; z < channels; z++) {
-               stbi_uc temp = image[(row * w + col) * channels + z];
-               image[(row * w + col) * channels + z] = image[((h - row - 1) * w + col) * channels + z];
-               image[((h - row - 1) * w + col) * channels + z] = temp;
-            }
-         }
-      }
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc));
    }
 
    return (unsigned char *) result;
@@ -1084,21 +1096,8 @@ static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x,
    // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision
 
    if (stbi__vertically_flip_on_load) {
-      int w = *x, h = *y;
       int channels = req_comp ? req_comp : *comp;
-      int row,col,z;
-      stbi__uint16 *image = (stbi__uint16 *) result;
-
-      // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once
-      for (row = 0; row < (h>>1); row++) {
-         for (col = 0; col < w; col++) {
-            for (z = 0; z < channels; z++) {
-               stbi__uint16 temp = image[(row * w + col) * channels + z];
-               image[(row * w + col) * channels + z] = image[((h - row - 1) * w + col) * channels + z];
-               image[((h - row - 1) * w + col) * channels + z] = temp;
-            }
-         }
-      }
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16));
    }
 
    return (stbi__uint16 *) result;
@@ -1108,21 +1107,8 @@ static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x,
 static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp)
 {
    if (stbi__vertically_flip_on_load && result != NULL) {
-      int w = *x, h = *y;
-      int depth = req_comp ? req_comp : *comp;
-      int row,col,z;
-      float temp;
-
-      // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once
-      for (row = 0; row < (h>>1); row++) {
-         for (col = 0; col < w; col++) {
-            for (z = 0; z < depth; z++) {
-               temp = result[(row * w + col) * depth + z];
-               result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z];
-               result[((h - row - 1) * w + col) * depth + z] = temp;
-            }
-         }
-      }
+      int channels = req_comp ? req_comp : *comp;
+      stbi__vertical_flip(result, *x, *y, channels * sizeof(float));
    }
 }
 #endif
@@ -1191,6 +1177,20 @@ STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, i
 
 #endif //!STBI_NO_STDIO
 
+STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+   stbi__context s;
+   stbi__start_mem(&s,buffer,len);
+   return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
+STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels)
+{
+   stbi__context s;
+   stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user);
+   return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels);
+}
+
 STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp)
 {
    stbi__context s;
@@ -2806,7 +2806,7 @@ static int stbi__process_marker(stbi__jpeg *z, int m)
             if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG");
 
             for (i=0; i < 64; ++i)
-               z->dequant[t][stbi__jpeg_dezigzag[i]] = sixteen ? stbi__get16be(z->s) : stbi__get8(z->s);
+               z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s));
             L -= (sixteen ? 129 : 65);
          }
          return L==0;
@@ -3611,20 +3611,20 @@ static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp
             } else if (z->s->img_n == 4) {
                if (z->app14_color_transform == 0) { // CMYK
                   for (i=0; i < z->s->img_x; ++i) {
-                     stbi_uc k = coutput[3][i];
-                     out[0] = stbi__blinn_8x8(coutput[0][i], k);
-                     out[1] = stbi__blinn_8x8(coutput[1][i], k);
-                     out[2] = stbi__blinn_8x8(coutput[2][i], k);
+                     stbi_uc m = coutput[3][i];
+                     out[0] = stbi__blinn_8x8(coutput[0][i], m);
+                     out[1] = stbi__blinn_8x8(coutput[1][i], m);
+                     out[2] = stbi__blinn_8x8(coutput[2][i], m);
                      out[3] = 255;
                      out += n;
                   }
                } else if (z->app14_color_transform == 2) { // YCCK
                   z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n);
                   for (i=0; i < z->s->img_x; ++i) {
-                     stbi_uc k = coutput[3][i];
-                     out[0] = stbi__blinn_8x8(255 - out[0], k);
-                     out[1] = stbi__blinn_8x8(255 - out[1], k);
-                     out[2] = stbi__blinn_8x8(255 - out[2], k);
+                     stbi_uc m = coutput[3][i];
+                     out[0] = stbi__blinn_8x8(255 - out[0], m);
+                     out[1] = stbi__blinn_8x8(255 - out[1], m);
+                     out[2] = stbi__blinn_8x8(255 - out[2], m);
                      out += n;
                   }
                } else { // YCbCr + alpha?  Ignore the fourth channel for now
@@ -3649,10 +3649,10 @@ static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp
                }
             } else if (z->s->img_n == 4 && z->app14_color_transform == 0) {
                for (i=0; i < z->s->img_x; ++i) {
-                  stbi_uc k = coutput[3][i];
-                  stbi_uc r = stbi__blinn_8x8(coutput[0][i], k);
-                  stbi_uc g = stbi__blinn_8x8(coutput[1][i], k);
-                  stbi_uc b = stbi__blinn_8x8(coutput[2][i], k);
+                  stbi_uc m = coutput[3][i];
+                  stbi_uc r = stbi__blinn_8x8(coutput[0][i], m);
+                  stbi_uc g = stbi__blinn_8x8(coutput[1][i], m);
+                  stbi_uc b = stbi__blinn_8x8(coutput[2][i], m);
                   out[0] = stbi__compute_y(r, g, b);
                   out[1] = 255;
                   out += n;
@@ -4297,11 +4297,10 @@ static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 r
 
    img_width_bytes = (((img_n * x * depth) + 7) >> 3);
    img_len = (img_width_bytes + 1) * y;
-   if (s->img_x == x && s->img_y == y) {
-      if (raw_len != img_len) return stbi__err("not enough pixels","Corrupt PNG");
-   } else { // interlaced:
-      if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");
-   }
+   // we used to check for exact match between raw_len and img_len on non-interlaced PNGs,
+   // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros),
+   // so just check for raw_len < img_len always.
+   if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");
 
    for (j=0; j < y; ++j) {
       stbi_uc *cur = a->out + stride*j;
@@ -4654,9 +4653,10 @@ static void stbi__de_iphone(stbi__png *z)
             stbi_uc a = p[3];
             stbi_uc t = p[0];
             if (a) {
-               p[0] = p[2] * 255 / a;
-               p[1] = p[1] * 255 / a;
-               p[2] =  t   * 255 / a;
+               stbi_uc half = a / 2;
+               p[0] = (p[2] * 255 + half) / a;
+               p[1] = (p[1] * 255 + half) / a;
+               p[2] = ( t   * 255 + half) / a;
             } else {
                p[0] = p[2];
                p[2] = t;
@@ -4819,6 +4819,9 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp)
                if (req_comp >= 3) s->img_out_n = req_comp;
                if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n))
                   return 0;
+            } else if (has_trans) {
+               // non-paletted image with tRNS -> source image has (constant) alpha
+               ++s->img_n;
             }
             STBI_FREE(z->expanded); z->expanded = NULL;
             return 1;
@@ -6966,6 +6969,13 @@ STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int
 
 /*
    revision history:
+      2.16  (2017-07-23) all functions have 16-bit variants;
+                         STBI_NO_STDIO works again;
+                         compilation fixes;
+                         fix rounding in unpremultiply;
+                         optimize vertical flip;
+                         disable raw_len validation;
+                         documentation fixes
       2.15  (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode;
                          warning fixes; disable run-time SSE detection on gcc;
                          uniform handling of optional "return" values;

src/external/stb_image_resize.h

@@ -1,4 +1,4 @@
-/* stb_image_resize - v0.94 - public domain image resizing
+/* stb_image_resize - v0.95 - public domain image resizing
    by Jorge L Rodriguez (@VinoBS) - 2014
    http://github.com/nothings/stb
 
@@ -156,8 +156,10 @@
       Jorge L Rodriguez: Implementation
       Sean Barrett: API design, optimizations
       Aras Pranckevicius: bugfix
-         
+      Nathan Reed: warning fixes
+
    REVISIONS
+      0.95 (2017-07-23) fixed warnings
       0.94 (2017-03-18) fixed warnings
       0.93 (2017-03-03) fixed bug with certain combinations of heights
       0.92 (2017-01-02) fix integer overflow on large (>2GB) images
@@ -393,8 +395,9 @@ STBIRDEF int stbir_resize_region(  const void *input_pixels , int input_w , int
 
 #ifndef STBIR_MALLOC
 #include <stdlib.h>
-#define STBIR_MALLOC(size,c) malloc(size)
-#define STBIR_FREE(ptr,c)    free(ptr)
+// use comma operator to evaluate c, to avoid "unused parameter" warnings
+#define STBIR_MALLOC(size,c) ((void)(c), malloc(size))
+#define STBIR_FREE(ptr,c)    ((void)(c), free(ptr))
 #endif
 
 #ifndef _MSC_VER
@@ -983,7 +986,7 @@ static int stbir__edge_wrap_slow(stbir_edge edge, int n, int max)
 
             return (m);
         }
-        return n;  // NOTREACHED
+        // NOTREACHED
 
     default:
         STBIR_ASSERT(!"Unimplemented edge type");

src/external/stb_image_write.h

@@ -1,5 +1,5 @@
-/* stb_image_write - v1.05 - public domain - http://nothings.org/stb/stb_image_write.h
-   writes out PNG/BMP/TGA images to C stdio - Sean Barrett 2010-2015
+/* stb_image_write - v1.07 - public domain - http://nothings.org/stb/stb_image_write.h
+   writes out PNG/BMP/TGA/JPEG/HDR images to C stdio - Sean Barrett 2010-2015
                                      no warranty implied; use at your own risk
 
    Before #including,
@@ -35,6 +35,7 @@ USAGE:
      int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data);
      int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data);
      int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data);
+     int stbi_write_jpg(char const *filename, int w, int h, int comp, const float *data);
 
    There are also four equivalent functions that use an arbitrary write function. You are
    expected to open/close your file-equivalent before and after calling these:
@@ -43,6 +44,7 @@ USAGE:
      int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
      int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
      int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data);
+     int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality);
 
    where the callback is:
       void stbi_write_func(void *context, void *data, int size);
@@ -79,6 +81,10 @@ USAGE:
 
    TGA supports RLE or non-RLE compressed data. To use non-RLE-compressed
    data, set the global variable 'stbi_write_tga_with_rle' to 0.
+   
+   JPEG does ignore alpha channels in input data; quality is between 1 and 100.
+   Higher quality looks better but results in a bigger image.
+   JPEG baseline (no JPEG progressive).
 
 CREDITS:
 
@@ -94,6 +100,9 @@ CREDITS:
       Alan Hickman
    initial file IO callback implementation
       Emmanuel Julien
+   JPEG
+      Jon Olick (original jo_jpeg.cpp code)
+      Daniel Gibson
    bugfixes:
       github:Chribba
       Guillaume Chereau
@@ -131,6 +140,7 @@ STBIWDEF int stbi_write_png(char const *filename, int w, int h, int comp, const
 STBIWDEF int stbi_write_bmp(char const *filename, int w, int h, int comp, const void  *data);
 STBIWDEF int stbi_write_tga(char const *filename, int w, int h, int comp, const void  *data);
 STBIWDEF int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data);
+STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void  *data, int quality);
 #endif
 
 typedef void stbi_write_func(void *context, void *data, int size);
@@ -139,6 +149,7 @@ STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int w,
 STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
 STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void  *data);
 STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data);
+STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void  *data, int quality);
 
 #ifdef __cplusplus
 }
@@ -277,6 +288,11 @@ static void stbiw__writef(stbi__write_context *s, const char *fmt, ...)
    va_end(v);
 }
 
+static void stbiw__putc(stbi__write_context *s, unsigned char c)
+{
+   s->func(s->context, &c, 1);
+}
+
 static void stbiw__write3(stbi__write_context *s, unsigned char a, unsigned char b, unsigned char c)
 {
    unsigned char arr[3];
@@ -450,7 +466,7 @@ static int stbi_write_tga_core(stbi__write_context *s, int x, int y, int comp, v
    return 1;
 }
 
-int stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data)
+STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data)
 {
    stbi__write_context s;
    stbi__start_write_callbacks(&s, func, context);
@@ -458,7 +474,7 @@ int stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, i
 }
 
 #ifndef STBI_WRITE_NO_STDIO
-int stbi_write_tga(char const *filename, int x, int y, int comp, const void *data)
+STBIWDEF int stbi_write_tga(char const *filename, int x, int y, int comp, const void *data)
 {
    stbi__write_context s;
    if (stbi__start_write_file(&s,filename)) {
@@ -620,7 +636,7 @@ static int stbi_write_hdr_core(stbi__write_context *s, int x, int y, int comp, f
    }
 }
 
-int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const float *data)
+STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const float *data)
 {
    stbi__write_context s;
    stbi__start_write_callbacks(&s, func, context);
@@ -628,7 +644,7 @@ int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, i
 }
 
 #ifndef STBI_WRITE_NO_STDIO
-int stbi_write_hdr(char const *filename, int x, int y, int comp, const float *data)
+STBIWDEF int stbi_write_hdr(char const *filename, int x, int y, int comp, const float *data)
 {
    stbi__write_context s;
    if (stbi__start_write_file(&s,filename)) {
@@ -1013,9 +1029,359 @@ STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int x,
    return 1;
 }
 
+
+/* ***************************************************************************
+ *
+ * JPEG writer
+ *
+ * This is based on Jon Olick's jo_jpeg.cpp:
+ * public domain Simple, Minimalistic JPEG writer - http://www.jonolick.com/code.html
+ */
+
+static const unsigned char stbiw__jpg_ZigZag[] = { 0,1,5,6,14,15,27,28,2,4,7,13,16,26,29,42,3,8,12,17,25,30,41,43,9,11,18,
+      24,31,40,44,53,10,19,23,32,39,45,52,54,20,22,33,38,46,51,55,60,21,34,37,47,50,56,59,61,35,36,48,49,57,58,62,63 };
+
+static void stbiw__jpg_writeBits(stbi__write_context *s, int *bitBufP, int *bitCntP, const unsigned short *bs) {
+   int bitBuf = *bitBufP, bitCnt = *bitCntP;
+   bitCnt += bs[1];
+   bitBuf |= bs[0] << (24 - bitCnt);
+   while(bitCnt >= 8) {
+      unsigned char c = (bitBuf >> 16) & 255;
+      stbiw__putc(s, c);
+      if(c == 255) {
+         stbiw__putc(s, 0);
+      }
+      bitBuf <<= 8;
+      bitCnt -= 8;
+   }
+   *bitBufP = bitBuf;
+   *bitCntP = bitCnt;
+}
+
+static void stbiw__jpg_DCT(float *d0p, float *d1p, float *d2p, float *d3p, float *d4p, float *d5p, float *d6p, float *d7p) {
+   float d0 = *d0p, d1 = *d1p, d2 = *d2p, d3 = *d3p, d4 = *d4p, d5 = *d5p, d6 = *d6p, d7 = *d7p;
+   float z1, z2, z3, z4, z5, z11, z13;
+
+   float tmp0 = d0 + d7;
+   float tmp7 = d0 - d7;
+   float tmp1 = d1 + d6;
+   float tmp6 = d1 - d6;
+   float tmp2 = d2 + d5;
+   float tmp5 = d2 - d5;
+   float tmp3 = d3 + d4;
+   float tmp4 = d3 - d4;
+
+   // Even part
+   float tmp10 = tmp0 + tmp3;   // phase 2
+   float tmp13 = tmp0 - tmp3;
+   float tmp11 = tmp1 + tmp2;
+   float tmp12 = tmp1 - tmp2;
+
+   d0 = tmp10 + tmp11;       // phase 3
+   d4 = tmp10 - tmp11;
+
+   z1 = (tmp12 + tmp13) * 0.707106781f; // c4
+   d2 = tmp13 + z1;       // phase 5
+   d6 = tmp13 - z1;
+
+   // Odd part
+   tmp10 = tmp4 + tmp5;       // phase 2
+   tmp11 = tmp5 + tmp6;
+   tmp12 = tmp6 + tmp7;
+
+   // The rotator is modified from fig 4-8 to avoid extra negations.
+   z5 = (tmp10 - tmp12) * 0.382683433f; // c6
+   z2 = tmp10 * 0.541196100f + z5; // c2-c6
+   z4 = tmp12 * 1.306562965f + z5; // c2+c6
+   z3 = tmp11 * 0.707106781f; // c4
+
+   z11 = tmp7 + z3;      // phase 5
+   z13 = tmp7 - z3;
+
+   *d5p = z13 + z2;         // phase 6
+   *d3p = z13 - z2;
+   *d1p = z11 + z4;
+   *d7p = z11 - z4;
+
+   *d0p = d0;  *d2p = d2;  *d4p = d4;  *d6p = d6;
+}
+
+static void stbiw__jpg_calcBits(int val, unsigned short bits[2]) {
+   int tmp1 = val < 0 ? -val : val;
+   val = val < 0 ? val-1 : val;
+   bits[1] = 1;
+   while(tmp1 >>= 1) {
+      ++bits[1];
+   }
+   bits[0] = val & ((1<<bits[1])-1);
+}
+
+static int stbiw__jpg_processDU(stbi__write_context *s, int *bitBuf, int *bitCnt, float *CDU, float *fdtbl, int DC, const unsigned short HTDC[256][2], const unsigned short HTAC[256][2]) {
+   const unsigned short EOB[2] = { HTAC[0x00][0], HTAC[0x00][1] };
+   const unsigned short M16zeroes[2] = { HTAC[0xF0][0], HTAC[0xF0][1] };
+   int dataOff, i, diff, end0pos;
+   int DU[64];
+
+   // DCT rows
+   for(dataOff=0; dataOff<64; dataOff+=8) {
+      stbiw__jpg_DCT(&CDU[dataOff], &CDU[dataOff+1], &CDU[dataOff+2], &CDU[dataOff+3], &CDU[dataOff+4], &CDU[dataOff+5], &CDU[dataOff+6], &CDU[dataOff+7]);
+   }
+   // DCT columns
+   for(dataOff=0; dataOff<8; ++dataOff) {
+      stbiw__jpg_DCT(&CDU[dataOff], &CDU[dataOff+8], &CDU[dataOff+16], &CDU[dataOff+24], &CDU[dataOff+32], &CDU[dataOff+40], &CDU[dataOff+48], &CDU[dataOff+56]);
+   }
+   // Quantize/descale/zigzag the coefficients
+   for(i=0; i<64; ++i) {
+      float v = CDU[i]*fdtbl[i];
+      // DU[stbiw__jpg_ZigZag[i]] = (int)(v < 0 ? ceilf(v - 0.5f) : floorf(v + 0.5f));
+      // ceilf() and floorf() are C99, not C89, but I /think/ they're not needed here anyway?
+      DU[stbiw__jpg_ZigZag[i]] = (int)(v < 0 ? v - 0.5f : v + 0.5f);
+   }
+
+   // Encode DC
+   diff = DU[0] - DC;
+   if (diff == 0) {
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTDC[0]);
+   } else {
+      unsigned short bits[2];
+      stbiw__jpg_calcBits(diff, bits);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTDC[bits[1]]);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, bits);
+   }
+   // Encode ACs
+   end0pos = 63;
+   for(; (end0pos>0)&&(DU[end0pos]==0); --end0pos) {
+   }
+   // end0pos = first element in reverse order !=0
+   if(end0pos == 0) {
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB);
+      return DU[0];
+   }
+   for(i = 1; i <= end0pos; ++i) {
+      int startpos = i;
+      int nrzeroes;
+      unsigned short bits[2];
+      for (; DU[i]==0 && i<=end0pos; ++i) {
+      }
+      nrzeroes = i-startpos;
+      if ( nrzeroes >= 16 ) {
+         int lng = nrzeroes>>4;
+         int nrmarker;
+         for (nrmarker=1; nrmarker <= lng; ++nrmarker)
+            stbiw__jpg_writeBits(s, bitBuf, bitCnt, M16zeroes);
+         nrzeroes &= 15;
+      }
+      stbiw__jpg_calcBits(DU[i], bits);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTAC[(nrzeroes<<4)+bits[1]]);
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, bits);
+   }
+   if(end0pos != 63) {
+      stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB);
+   }
+   return DU[0];
+}
+
+static int stbi_write_jpg_core(stbi__write_context *s, int width, int height, int comp, const void* data, int quality) {
+   // Constants that don't pollute global namespace
+   static const unsigned char std_dc_luminance_nrcodes[] = {0,0,1,5,1,1,1,1,1,1,0,0,0,0,0,0,0};
+   static const unsigned char std_dc_luminance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11};
+   static const unsigned char std_ac_luminance_nrcodes[] = {0,0,2,1,3,3,2,4,3,5,5,4,4,0,0,1,0x7d};
+   static const unsigned char std_ac_luminance_values[] = {
+      0x01,0x02,0x03,0x00,0x04,0x11,0x05,0x12,0x21,0x31,0x41,0x06,0x13,0x51,0x61,0x07,0x22,0x71,0x14,0x32,0x81,0x91,0xa1,0x08,
+      0x23,0x42,0xb1,0xc1,0x15,0x52,0xd1,0xf0,0x24,0x33,0x62,0x72,0x82,0x09,0x0a,0x16,0x17,0x18,0x19,0x1a,0x25,0x26,0x27,0x28,
+      0x29,0x2a,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59,
+      0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x83,0x84,0x85,0x86,0x87,0x88,0x89,
+      0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6,
+      0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe1,0xe2,
+      0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa
+   };
+   static const unsigned char std_dc_chrominance_nrcodes[] = {0,0,3,1,1,1,1,1,1,1,1,1,0,0,0,0,0};
+   static const unsigned char std_dc_chrominance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11};
+   static const unsigned char std_ac_chrominance_nrcodes[] = {0,0,2,1,2,4,4,3,4,7,5,4,4,0,1,2,0x77};
+   static const unsigned char std_ac_chrominance_values[] = {
+      0x00,0x01,0x02,0x03,0x11,0x04,0x05,0x21,0x31,0x06,0x12,0x41,0x51,0x07,0x61,0x71,0x13,0x22,0x32,0x81,0x08,0x14,0x42,0x91,
+      0xa1,0xb1,0xc1,0x09,0x23,0x33,0x52,0xf0,0x15,0x62,0x72,0xd1,0x0a,0x16,0x24,0x34,0xe1,0x25,0xf1,0x17,0x18,0x19,0x1a,0x26,
+      0x27,0x28,0x29,0x2a,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,
+      0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x82,0x83,0x84,0x85,0x86,0x87,
+      0x88,0x89,0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,
+      0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,
+      0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa
+   };
+   // Huffman tables
+   static const unsigned short YDC_HT[256][2] = { {0,2},{2,3},{3,3},{4,3},{5,3},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9}};
+   static const unsigned short UVDC_HT[256][2] = { {0,2},{1,2},{2,2},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9},{1022,10},{2046,11}};
+   static const unsigned short YAC_HT[256][2] = {
+      {10,4},{0,2},{1,2},{4,3},{11,4},{26,5},{120,7},{248,8},{1014,10},{65410,16},{65411,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {12,4},{27,5},{121,7},{502,9},{2038,11},{65412,16},{65413,16},{65414,16},{65415,16},{65416,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {28,5},{249,8},{1015,10},{4084,12},{65417,16},{65418,16},{65419,16},{65420,16},{65421,16},{65422,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {58,6},{503,9},{4085,12},{65423,16},{65424,16},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {59,6},{1016,10},{65430,16},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {122,7},{2039,11},{65438,16},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {123,7},{4086,12},{65446,16},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {250,8},{4087,12},{65454,16},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {504,9},{32704,15},{65462,16},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {505,9},{65470,16},{65471,16},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {506,9},{65479,16},{65480,16},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {1017,10},{65488,16},{65489,16},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {1018,10},{65497,16},{65498,16},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {2040,11},{65506,16},{65507,16},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {65515,16},{65516,16},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {2041,11},{65525,16},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0}
+   };
+   static const unsigned short UVAC_HT[256][2] = {
+      {0,2},{1,2},{4,3},{10,4},{24,5},{25,5},{56,6},{120,7},{500,9},{1014,10},{4084,12},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {11,4},{57,6},{246,8},{501,9},{2038,11},{4085,12},{65416,16},{65417,16},{65418,16},{65419,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {26,5},{247,8},{1015,10},{4086,12},{32706,15},{65420,16},{65421,16},{65422,16},{65423,16},{65424,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {27,5},{248,8},{1016,10},{4087,12},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{65430,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {58,6},{502,9},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{65438,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {59,6},{1017,10},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{65446,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {121,7},{2039,11},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{65454,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {122,7},{2040,11},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{65462,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {249,8},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{65470,16},{65471,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {503,9},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{65479,16},{65480,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {504,9},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{65488,16},{65489,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {505,9},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{65497,16},{65498,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {506,9},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{65506,16},{65507,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {2041,11},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{65515,16},{65516,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {16352,14},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{65525,16},{0,0},{0,0},{0,0},{0,0},{0,0},
+      {1018,10},{32707,15},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0}
+   };
+   static const int YQT[] = {16,11,10,16,24,40,51,61,12,12,14,19,26,58,60,55,14,13,16,24,40,57,69,56,14,17,22,29,51,87,80,62,18,22,
+                             37,56,68,109,103,77,24,35,55,64,81,104,113,92,49,64,78,87,103,121,120,101,72,92,95,98,112,100,103,99};
+   static const int UVQT[] = {17,18,24,47,99,99,99,99,18,21,26,66,99,99,99,99,24,26,56,99,99,99,99,99,47,66,99,99,99,99,99,99,
+                              99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99};
+   static const float aasf[] = { 1.0f * 2.828427125f, 1.387039845f * 2.828427125f, 1.306562965f * 2.828427125f, 1.175875602f * 2.828427125f, 
+                                 1.0f * 2.828427125f, 0.785694958f * 2.828427125f, 0.541196100f * 2.828427125f, 0.275899379f * 2.828427125f };
+
+   int row, col, i, k;
+   float fdtbl_Y[64], fdtbl_UV[64];
+   unsigned char YTable[64], UVTable[64];
+
+   if(!data || !width || !height || comp > 4 || comp < 1) {
+      return 0;
+   }
+
+   quality = quality ? quality : 90;
+   quality = quality < 1 ? 1 : quality > 100 ? 100 : quality;
+   quality = quality < 50 ? 5000 / quality : 200 - quality * 2;
+
+   for(i = 0; i < 64; ++i) {
+      int uvti, yti = (YQT[i]*quality+50)/100;
+      YTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (yti < 1 ? 1 : yti > 255 ? 255 : yti);
+      uvti = (UVQT[i]*quality+50)/100;
+      UVTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (uvti < 1 ? 1 : uvti > 255 ? 255 : uvti);
+   }
+
+   for(row = 0, k = 0; row < 8; ++row) {
+      for(col = 0; col < 8; ++col, ++k) {
+         fdtbl_Y[k]  = 1 / (YTable [stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]);
+         fdtbl_UV[k] = 1 / (UVTable[stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]);
+      }
+   }
+
+   // Write Headers
+   {
+      static const unsigned char head0[] = { 0xFF,0xD8,0xFF,0xE0,0,0x10,'J','F','I','F',0,1,1,0,0,1,0,1,0,0,0xFF,0xDB,0,0x84,0 };
+      static const unsigned char head2[] = { 0xFF,0xDA,0,0xC,3,1,0,2,0x11,3,0x11,0,0x3F,0 };
+      const unsigned char head1[] = { 0xFF,0xC0,0,0x11,8,(unsigned char)(height>>8),STBIW_UCHAR(height),(unsigned char)(width>>8),STBIW_UCHAR(width),
+                                      3,1,0x11,0,2,0x11,1,3,0x11,1,0xFF,0xC4,0x01,0xA2,0 };
+      s->func(s->context, (void*)head0, sizeof(head0));
+      s->func(s->context, (void*)YTable, sizeof(YTable));
+      stbiw__putc(s, 1);
+      s->func(s->context, UVTable, sizeof(UVTable));
+      s->func(s->context, (void*)head1, sizeof(head1));
+      s->func(s->context, (void*)(std_dc_luminance_nrcodes+1), sizeof(std_dc_luminance_nrcodes)-1);
+      s->func(s->context, (void*)std_dc_luminance_values, sizeof(std_dc_luminance_values));
+      stbiw__putc(s, 0x10); // HTYACinfo
+      s->func(s->context, (void*)(std_ac_luminance_nrcodes+1), sizeof(std_ac_luminance_nrcodes)-1);
+      s->func(s->context, (void*)std_ac_luminance_values, sizeof(std_ac_luminance_values));
+      stbiw__putc(s, 1); // HTUDCinfo
+      s->func(s->context, (void*)(std_dc_chrominance_nrcodes+1), sizeof(std_dc_chrominance_nrcodes)-1);
+      s->func(s->context, (void*)std_dc_chrominance_values, sizeof(std_dc_chrominance_values));
+      stbiw__putc(s, 0x11); // HTUACinfo
+      s->func(s->context, (void*)(std_ac_chrominance_nrcodes+1), sizeof(std_ac_chrominance_nrcodes)-1);
+      s->func(s->context, (void*)std_ac_chrominance_values, sizeof(std_ac_chrominance_values));
+      s->func(s->context, (void*)head2, sizeof(head2));
+   }
+
+   // Encode 8x8 macroblocks
+   {
+      static const unsigned short fillBits[] = {0x7F, 7};
+      const unsigned char *imageData = (const unsigned char *)data;
+      int DCY=0, DCU=0, DCV=0;
+      int bitBuf=0, bitCnt=0;
+      // comp == 2 is grey+alpha (alpha is ignored)
+      int ofsG = comp > 2 ? 1 : 0, ofsB = comp > 2 ? 2 : 0;
+      int x, y, pos;
+      for(y = 0; y < height; y += 8) {
+         for(x = 0; x < width; x += 8) {
+            float YDU[64], UDU[64], VDU[64];
+            for(row = y, pos = 0; row < y+8; ++row) {
+               for(col = x; col < x+8; ++col, ++pos) {
+                  int p = row*width*comp + col*comp;
+                  float r, g, b;
+                  if(row >= height) {
+                     p -= width*comp*(row+1 - height);
+                  }
+                  if(col >= width) {
+                     p -= comp*(col+1 - width);
+                  }
+
+                  r = imageData[p+0];
+                  g = imageData[p+ofsG];
+                  b = imageData[p+ofsB];
+                  YDU[pos]=+0.29900f*r+0.58700f*g+0.11400f*b-128;
+                  UDU[pos]=-0.16874f*r-0.33126f*g+0.50000f*b;
+                  VDU[pos]=+0.50000f*r-0.41869f*g-0.08131f*b;
+               }
+            }
+
+            DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, YDU, fdtbl_Y, DCY, YDC_HT, YAC_HT);
+            DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, UDU, fdtbl_UV, DCU, UVDC_HT, UVAC_HT);
+            DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, VDU, fdtbl_UV, DCV, UVDC_HT, UVAC_HT);
+         }
+      }
+
+      // Do the bit alignment of the EOI marker
+      stbiw__jpg_writeBits(s, &bitBuf, &bitCnt, fillBits);
+   }
+
+   // EOI
+   stbiw__putc(s, 0xFF);
+   stbiw__putc(s, 0xD9);
+
+   return 1;
+}
+
+STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality)
+{
+   stbi__write_context s;
+   stbi__start_write_callbacks(&s, func, context);
+   return stbi_write_jpg_core(&s, x, y, comp, (void *) data, quality);
+}
+
+
+#ifndef STBI_WRITE_NO_STDIO
+STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void *data, int quality)
+{
+   stbi__write_context s;
+   if (stbi__start_write_file(&s,filename)) {
+      int r = stbi_write_jpg_core(&s, x, y, comp, data, quality);
+      stbi__end_write_file(&s);
+      return r;
+   } else
+      return 0;
+}
+#endif
+
 #endif // STB_IMAGE_WRITE_IMPLEMENTATION
 
 /* Revision history
+      1.07  (2017-07-24)
+             doc fix
+      1.06 (2017-07-23)
+             writing JPEG (using Jon Olick's code)
+      1.05   ???
       1.04 (2017-03-03)
              monochrome BMP expansion
       1.03   ???

src/external/stb_rect_pack.h

@@ -524,17 +524,6 @@ static int rect_height_compare(const void *a, const void *b)
    return (p->w > q->w) ? -1 : (p->w < q->w);
 }
 
-static int rect_width_compare(const void *a, const void *b)
-{
-   const stbrp_rect *p = (const stbrp_rect *) a;
-   const stbrp_rect *q = (const stbrp_rect *) b;
-   if (p->w > q->w)
-      return -1;
-   if (p->w < q->w)
-      return  1;
-   return (p->h > q->h) ? -1 : (p->h < q->h);
-}
-
 static int rect_original_order(const void *a, const void *b)
 {
    const stbrp_rect *p = (const stbrp_rect *) a;

src/external/stb_truetype.h

@@ -1,4 +1,4 @@
-// stb_truetype.h - v1.15 - public domain
+// stb_truetype.h - v1.17 - public domain
 // authored from 2009-2016 by Sean Barrett / RAD Game Tools
 //
 //   This library processes TrueType files:
@@ -6,6 +6,7 @@
 //        extract glyph metrics
 //        extract glyph shapes
 //        render glyphs to one-channel bitmaps with antialiasing (box filter)
+//        render glyphs to one-channel SDF bitmaps (signed-distance field/function)
 //
 //   Todo:
 //        non-MS cmaps
@@ -26,9 +27,10 @@
 //       Ryan Gordon
 //       Simon Glass
 //       github:IntellectualKitty
+//       Imanol Celaya
 //
 //   Bug/warning reports/fixes:
-//       "Zer" on mollyrocket (with fix)
+//       "Zer" on mollyrocket
 //       Cass Everitt
 //       stoiko (Haemimont Games)
 //       Brian Hook 
@@ -51,9 +53,12 @@
 //       Thomas Fields
 //       Derek Vinyard
 //       Cort Stratton
+//       github:oyvindjam
 //
 // VERSION HISTORY
 //
+//   1.17 (2017-07-23) make more arguments const; doc fix
+//   1.16 (2017-07-12) SDF support
 //   1.15 (2017-03-03) make more arguments const
 //   1.14 (2017-01-16) num-fonts-in-TTC function
 //   1.13 (2017-01-02) support OpenType fonts, certain Apple fonts
@@ -92,7 +97,7 @@
 //   Improved 3D API (more shippable):
 //           #include "stb_rect_pack.h"           -- optional, but you really want it
 //           stbtt_PackBegin()
-//           stbtt_PackSetOversample()            -- for improved quality on small fonts
+//           stbtt_PackSetOversampling()          -- for improved quality on small fonts
 //           stbtt_PackFontRanges()               -- pack and renders
 //           stbtt_PackEnd()
 //           stbtt_GetPackedQuad()
@@ -110,6 +115,7 @@
 //   Character advance/positioning
 //           stbtt_GetCodepointHMetrics()
 //           stbtt_GetFontVMetrics()
+//           stbtt_GetFontVMetricsOS2()
 //           stbtt_GetCodepointKernAdvance()
 //
 //   Starting with version 1.06, the rasterizer was replaced with a new,
@@ -407,6 +413,18 @@ int main(int arg, char **argv)
    #ifndef STBTT_sqrt
    #include <math.h>
    #define STBTT_sqrt(x)      sqrt(x)
+   #define STBTT_pow(x,y)     pow(x,y)
+   #endif
+
+   #ifndef STBTT_cos
+   #include <math.h>
+   #define STBTT_cos(x)       cos(x)
+   #define STBTT_acos(x)      acos(x)
+   #endif
+
+   #ifndef STBTT_fabs
+   #include <math.h>
+   #define STBTT_fabs(x)      fabs(x)
    #endif
 
    #ifndef STBTT_fabs
@@ -432,7 +450,7 @@ int main(int arg, char **argv)
    #endif
 
    #ifndef STBTT_memcpy
-   #include <memory.h>
+   #include <string.h>
    #define STBTT_memcpy       memcpy
    #define STBTT_memset       memset
    #endif
@@ -548,7 +566,7 @@ STBTT_DEF void stbtt_PackEnd  (stbtt_pack_context *spc);
 
 #define STBTT_POINT_SIZE(x)   (-(x))
 
-STBTT_DEF int  stbtt_PackFontRange(stbtt_pack_context *spc, unsigned char *fontdata, int font_index, float font_size,
+STBTT_DEF int  stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size,
                                 int first_unicode_char_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range);
 // Creates character bitmaps from the font_index'th font found in fontdata (use
 // font_index=0 if you don't know what that is). It creates num_chars_in_range
@@ -573,7 +591,7 @@ typedef struct
    unsigned char h_oversample, v_oversample; // don't set these, they're used internally
 } stbtt_pack_range;
 
-STBTT_DEF int  stbtt_PackFontRanges(stbtt_pack_context *spc, unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges);
+STBTT_DEF int  stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges);
 // Creates character bitmaps from multiple ranges of characters stored in
 // ranges. This will usually create a better-packed bitmap than multiple
 // calls to stbtt_PackFontRange. Note that you can call this multiple
@@ -715,6 +733,12 @@ STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, in
 //   these are expressed in unscaled coordinates, so you must multiply by
 //   the scale factor for a given size
 
+STBTT_DEF int  stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap);
+// analogous to GetFontVMetrics, but returns the "typographic" values from the OS/2
+// table (specific to MS/Windows TTF files).
+//
+// Returns 1 on success (table present), 0 on failure.
+
 STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1);
 // the bounding box around all possible characters
 
@@ -809,6 +833,10 @@ STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, uns
 // same as stbtt_MakeCodepointBitmap, but you can specify a subpixel
 // shift for the character
 
+STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint);
+// same as stbtt_MakeCodepointBitmapSubpixel, but prefiltering
+// is performed (see stbtt_PackSetOversampling)
+
 STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1);
 // get the bbox of the bitmap centered around the glyph origin; so the
 // bitmap width is ix1-ix0, height is iy1-iy0, and location to place
@@ -826,6 +854,7 @@ STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float
 STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff);
 STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph);
 STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph);
+STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int glyph);
 STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1);
 STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1);
 
@@ -848,6 +877,64 @@ STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result,        // 1-channel bitmap
                                int invert,                   // if non-zero, vertically flip shape
                                void *userdata);              // context for to STBTT_MALLOC
 
+//////////////////////////////////////////////////////////////////////////////
+//
+// Signed Distance Function (or Field) rendering
+
+STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata);
+// frees the SDF bitmap allocated below
+
+STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff);
+STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff);
+// These functions compute a discretized SDF field for a single character, suitable for storing
+// in a single-channel texture, sampling with bilinear filtering, and testing against
+// larger than some threshhold to produce scalable fonts.
+//        info              --  the font
+//        scale             --  controls the size of the resulting SDF bitmap, same as it would be creating a regular bitmap
+//        glyph/codepoint   --  the character to generate the SDF for
+//        padding           --  extra "pixels" around the character which are filled with the distance to the character (not 0),
+//                                 which allows effects like bit outlines
+//        onedge_value      --  value 0-255 to test the SDF against to reconstruct the character (i.e. the isocontour of the character)
+//        pixel_dist_scale  --  what value the SDF should increase by when moving one SDF "pixel" away from the edge (on the 0..255 scale)
+//                                 if positive, > onedge_value is inside; if negative, < onedge_value is inside
+//        width,height      --  output height & width of the SDF bitmap (including padding)
+//        xoff,yoff         --  output origin of the character
+//        return value      --  a 2D array of bytes 0..255, width*height in size
+//
+// pixel_dist_scale & onedge_value are a scale & bias that allows you to make
+// optimal use of the limited 0..255 for your application, trading off precision
+// and special effects. SDF values outside the range 0..255 are clamped to 0..255.
+//
+// Example:
+//      scale = stbtt_ScaleForPixelHeight(22)
+//      padding = 5
+//      onedge_value = 180
+//      pixel_dist_scale = 180/5.0 = 36.0
+//
+//      This will create an SDF bitmap in which the character is about 22 pixels
+//      high but the whole bitmap is about 22+5+5=32 pixels high. To produce a filled
+//      shape, sample the SDF at each pixel and fill the pixel if the SDF value
+//      is greater than or equal to 180/255. (You'll actually want to antialias,
+//      which is beyond the scope of this example.) Additionally, you can compute
+//      offset outlines (e.g. to stroke the character border inside & outside,
+//      or only outside). For example, to fill outside the character up to 3 SDF
+//      pixels, you would compare against (180-36.0*3)/255 = 72/255. The above
+//      choice of variables maps a range from 5 pixels outside the shape to
+//      2 pixels inside the shape to 0..255; this is intended primarily for apply
+//      outside effects only (the interior range is needed to allow proper
+//      antialiasing of the font at *smaller* sizes)
+//
+// The function computes the SDF analytically at each SDF pixel, not by e.g.
+// building a higher-res bitmap and approximating it. In theory the quality
+// should be as high as possible for an SDF of this size & representation, but
+// unclear if this is true in practice (perhaps building a higher-res bitmap
+// and computing from that can allow drop-out prevention).
+//
+// The algorithm has not been optimized at all, so expect it to be slow
+// if computing lots of characters or very large sizes. 
+
+
+
 //////////////////////////////////////////////////////////////////////////////
 //
 // Finding the right font...
@@ -2201,6 +2288,17 @@ STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, in
    if (lineGap) *lineGap = ttSHORT(info->data+info->hhea + 8);
 }
 
+STBTT_DEF int  stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap)
+{
+   int tab = stbtt__find_table(info->data, info->fontstart, "OS/2");
+   if (!tab)
+      return 0;
+   if (typoAscent ) *typoAscent  = ttSHORT(info->data+tab + 68);
+   if (typoDescent) *typoDescent = ttSHORT(info->data+tab + 70);
+   if (typoLineGap) *typoLineGap = ttSHORT(info->data+tab + 72);
+   return 1;
+}
+
 STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1)
 {
    *x0 = ttSHORT(info->data + info->head + 36);
@@ -2693,19 +2791,18 @@ static void stbtt__fill_active_edges_new(float *scanline, float *scanline_fill,
                // from the other y segment, and it might ignored as an empty segment. to avoid
                // that, we need to explicitly produce segments based on x positions.
 
-               // rename variables to clear pairs
+               // rename variables to clearly-defined pairs
                float y0 = y_top;
                float x1 = (float) (x);
                float x2 = (float) (x+1);
                float x3 = xb;
                float y3 = y_bottom;
-               float y1,y2;
 
                // x = e->x + e->dx * (y-y_top)
                // (y-y_top) = (x - e->x) / e->dx
                // y = (x - e->x) / e->dx + y_top
-               y1 = (x - x0) / dx + y_top;
-               y2 = (x+1 - x0) / dx + y_top;
+               float y1 = (x - x0) / dx + y_top;
+               float y2 = (x+1 - x0) / dx + y_top;
 
                if (x0 < x1 && x3 > x2) {         // three segments descending down-right
                   stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1);
@@ -3600,6 +3697,29 @@ STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stb
    return k;
 }
 
+STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int prefilter_x, int prefilter_y, float *sub_x, float *sub_y, int glyph)
+{
+   stbtt_MakeGlyphBitmapSubpixel(info,
+                                 output,
+                                 out_w - (prefilter_x - 1),
+                                 out_h - (prefilter_y - 1),
+                                 out_stride,
+                                 scale_x,
+                                 scale_y,
+                                 shift_x,
+                                 shift_y,
+                                 glyph);
+
+   if (prefilter_x > 1)
+      stbtt__h_prefilter(output, out_w, out_h, out_stride, prefilter_x);
+
+   if (prefilter_y > 1)
+      stbtt__v_prefilter(output, out_w, out_h, out_stride, prefilter_y);
+
+   *sub_x = stbtt__oversample_shift(prefilter_x);
+   *sub_y = stbtt__oversample_shift(prefilter_y);
+}
+
 // rects array must be big enough to accommodate all characters in the given ranges
 STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects)
 {
@@ -3688,7 +3808,7 @@ STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect
    stbrp_pack_rects((stbrp_context *) spc->pack_info, rects, num_rects);
 }
 
-STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges)
+STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges)
 {
    stbtt_fontinfo info;
    int i,j,n, return_value = 1;
@@ -3724,7 +3844,7 @@ STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, unsigned char *fontd
    return return_value;
 }
 
-STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, unsigned char *fontdata, int font_index, float font_size,
+STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size,
             int first_unicode_codepoint_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range)
 {
    stbtt_pack_range range;
@@ -3763,6 +3883,387 @@ STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int
    *xpos += b->xadvance;
 }
 
+//////////////////////////////////////////////////////////////////////////////
+//
+// sdf computation
+//
+
+#define STBTT_min(a,b)  ((a) < (b) ? (a) : (b))
+#define STBTT_max(a,b)  ((a) < (b) ? (b) : (a))
+
+static int stbtt__ray_intersect_bezier(float orig[2], float ray[2], float q0[2], float q1[2], float q2[2], float hits[2][2])
+{
+   float q0perp = q0[1]*ray[0] - q0[0]*ray[1];
+   float q1perp = q1[1]*ray[0] - q1[0]*ray[1];
+   float q2perp = q2[1]*ray[0] - q2[0]*ray[1];
+   float roperp = orig[1]*ray[0] - orig[0]*ray[1];
+
+   float a = q0perp - 2*q1perp + q2perp;
+   float b = q1perp - q0perp;
+   float c = q0perp - roperp;
+
+   float s0 = 0., s1 = 0.;
+   int num_s = 0;
+
+   if (a != 0.0) {
+      float discr = b*b - a*c;
+      if (discr > 0.0) {
+         float rcpna = -1 / a;
+         float d = (float) sqrt(discr);
+         s0 = (b+d) * rcpna;
+         s1 = (b-d) * rcpna;
+         if (s0 >= 0.0 && s0 <= 1.0)
+            num_s = 1;
+         if (d > 0.0 && s1 >= 0.0 && s1 <= 1.0) {
+            if (num_s == 0) s0 = s1;
+            ++num_s;
+         }
+      }
+   } else {
+      // 2*b*s + c = 0
+      // s = -c / (2*b)
+      s0 = c / (-2 * b);
+      if (s0 >= 0.0 && s0 <= 1.0)
+         num_s = 1;
+   }
+
+   if (num_s == 0)
+      return 0;
+   else {
+      float rcp_len2 = 1 / (ray[0]*ray[0] + ray[1]*ray[1]);
+      float rayn_x = ray[0] * rcp_len2, rayn_y = ray[1] * rcp_len2;
+
+      float q0d =   q0[0]*rayn_x +   q0[1]*rayn_y;
+      float q1d =   q1[0]*rayn_x +   q1[1]*rayn_y;
+      float q2d =   q2[0]*rayn_x +   q2[1]*rayn_y;
+      float rod = orig[0]*rayn_x + orig[1]*rayn_y;
+
+      float q10d = q1d - q0d;
+      float q20d = q2d - q0d;
+      float q0rd = q0d - rod;
+
+      hits[0][0] = q0rd + s0*(2.0f - 2.0f*s0)*q10d + s0*s0*q20d;
+      hits[0][1] = a*s0+b;
+
+      if (num_s > 1) {
+         hits[1][0] = q0rd + s1*(2.0f - 2.0f*s1)*q10d + s1*s1*q20d;
+         hits[1][1] = a*s1+b;
+         return 2;
+      } else {
+         return 1;
+      }
+   }
+}
+
+static int equal(float *a, float *b)
+{
+   return (a[0] == b[0] && a[1] == b[1]);
+}
+
+static int stbtt__compute_crossings_x(float x, float y, int nverts, stbtt_vertex *verts)
+{
+   int i;
+   float orig[2], ray[2] = { 1, 0 };
+   float y_frac;
+   int winding = 0;
+
+   orig[0] = x;
+   orig[1] = y;
+
+   // make sure y never passes through a vertex of the shape
+   y_frac = (float) fmod(y, 1.0f);
+   if (y_frac < 0.01f)
+      y += 0.01f;
+   else if (y_frac > 0.99f)
+      y -= 0.01f;
+   orig[1] = y;
+
+   // test a ray from (-infinity,y) to (x,y)
+   for (i=0; i < nverts; ++i) {
+      if (verts[i].type == STBTT_vline) {
+         int x0 = (int) verts[i-1].x, y0 = (int) verts[i-1].y;
+         int x1 = (int) verts[i  ].x, y1 = (int) verts[i  ].y;
+         if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) {
+            float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0;
+            if (x_inter < x)  
+               winding += (y0 < y1) ? 1 : -1;
+         }
+      }
+      if (verts[i].type == STBTT_vcurve) {
+         int x0 = (int) verts[i-1].x , y0 = (int) verts[i-1].y ;
+         int x1 = (int) verts[i  ].cx, y1 = (int) verts[i  ].cy;
+         int x2 = (int) verts[i  ].x , y2 = (int) verts[i  ].y ;
+         int ax = STBTT_min(x0,STBTT_min(x1,x2)), ay = STBTT_min(y0,STBTT_min(y1,y2));
+         int by = STBTT_max(y0,STBTT_max(y1,y2));
+         if (y > ay && y < by && x > ax) {
+            float q0[2],q1[2],q2[2];
+            float hits[2][2];
+            q0[0] = (float)x0;
+            q0[1] = (float)y0;
+            q1[0] = (float)x1;
+            q1[1] = (float)y1;
+            q2[0] = (float)x2;
+            q2[1] = (float)y2;
+            if (equal(q0,q1) || equal(q1,q2)) {
+               x0 = (int)verts[i-1].x;
+               y0 = (int)verts[i-1].y;
+               x1 = (int)verts[i  ].x;
+               y1 = (int)verts[i  ].y;
+               if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) {
+                  float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0;
+                  if (x_inter < x)  
+                     winding += (y0 < y1) ? 1 : -1;
+               }
+            } else {
+               int num_hits = stbtt__ray_intersect_bezier(orig, ray, q0, q1, q2, hits);
+               if (num_hits >= 1)
+                  if (hits[0][0] < 0)
+                     winding += (hits[0][1] < 0 ? -1 : 1);
+               if (num_hits >= 2)
+                  if (hits[1][0] < 0)
+                     winding += (hits[1][1] < 0 ? -1 : 1);
+            }
+         } 
+      }
+   }
+   return winding;
+}
+
+static float stbtt__cuberoot( float x )
+{
+   if (x<0)
+      return -(float) STBTT_pow(-x,1.0f/3.0f);
+   else
+      return  (float) STBTT_pow( x,1.0f/3.0f);
+}
+
+// x^3 + c*x^2 + b*x + a = 0
+static int stbtt__solve_cubic(float a, float b, float c, float* r)
+{
+	float s = -a / 3;
+	float p = b - a*a / 3;
+	float q = a * (2*a*a - 9*b) / 27 + c;
+   float p3 = p*p*p;
+	float d = q*q + 4*p3 / 27;
+	if (d >= 0) {
+		float z = (float) STBTT_sqrt(d);
+		float u = (-q + z) / 2;
+		float v = (-q - z) / 2;
+		u = stbtt__cuberoot(u);
+		v = stbtt__cuberoot(v);
+		r[0] = s + u + v;
+		return 1;
+	} else {
+	   float u = (float) STBTT_sqrt(-p/3);
+	   float v = (float) STBTT_acos(-STBTT_sqrt(-27/p3) * q / 2) / 3; // p3 must be negative, since d is negative
+	   float m = (float) STBTT_cos(v);
+      float n = (float) STBTT_cos(v-3.141592/2)*1.732050808f;
+	   r[0] = s + u * 2 * m;
+	   r[1] = s - u * (m + n);
+	   r[2] = s - u * (m - n);
+
+      //STBTT_assert( STBTT_fabs(((r[0]+a)*r[0]+b)*r[0]+c) < 0.05f);  // these asserts may not be safe at all scales, though they're in bezier t parameter units so maybe?
+      //STBTT_assert( STBTT_fabs(((r[1]+a)*r[1]+b)*r[1]+c) < 0.05f);
+      //STBTT_assert( STBTT_fabs(((r[2]+a)*r[2]+b)*r[2]+c) < 0.05f);
+   	return 3;
+   }
+}
+
+STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff)
+{
+   float scale_x = scale, scale_y = scale;
+   int ix0,iy0,ix1,iy1;
+   int w,h;
+   unsigned char *data;
+
+   // if one scale is 0, use same scale for both
+   if (scale_x == 0) scale_x = scale_y;
+   if (scale_y == 0) {
+      if (scale_x == 0) return NULL;  // if both scales are 0, return NULL
+      scale_y = scale_x;
+   }
+
+   stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale, scale, 0.0f,0.0f, &ix0,&iy0,&ix1,&iy1);
+
+   // if empty, return NULL
+   if (ix0 == ix1 || iy0 == iy1)
+      return NULL;
+
+   ix0 -= padding;
+   iy0 -= padding;
+   ix1 += padding;
+   iy1 += padding;
+
+   w = (ix1 - ix0);
+   h = (iy1 - iy0);
+
+   if (width ) *width  = w;
+   if (height) *height = h;
+   if (xoff  ) *xoff   = ix0;
+   if (yoff  ) *yoff   = iy0;
+
+   // invert for y-downwards bitmaps
+   scale_y = -scale_y;
+      
+   {
+      int x,y,i,j;
+      float *precompute;
+      stbtt_vertex *verts;
+      int num_verts = stbtt_GetGlyphShape(info, glyph, &verts);
+      data = (unsigned char *) STBTT_malloc(w * h, info->userdata);
+      precompute = (float *) STBTT_malloc(num_verts * sizeof(float), info->userdata);
+
+      for (i=0,j=num_verts-1; i < num_verts; j=i++) {
+         if (verts[i].type == STBTT_vline) {
+            float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y;
+            float x1 = verts[j].x*scale_x, y1 = verts[j].y*scale_y;
+            float dist = (float) STBTT_sqrt((x1-x0)*(x1-x0) + (y1-y0)*(y1-y0));
+            precompute[i] = (dist == 0) ? 0.0f : 1.0f / dist;
+         } else if (verts[i].type == STBTT_vcurve) {
+            float x2 = verts[j].x *scale_x, y2 = verts[j].y *scale_y;
+            float x1 = verts[i].cx*scale_x, y1 = verts[i].cy*scale_y;
+            float x0 = verts[i].x *scale_x, y0 = verts[i].y *scale_y;
+            float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2;
+            float len2 = bx*bx + by*by;
+            if (len2 != 0.0f)
+               precompute[i] = 1.0f / (bx*bx + by*by);
+            else
+               precompute[i] = 0.0f;
+         } else
+            precompute[i] = 0.0f;
+      }
+
+      for (y=iy0; y < iy1; ++y) {
+         for (x=ix0; x < ix1; ++x) {
+            float val;
+            float min_dist = 999999.0f;
+            float sx = (float) x + 0.5f;
+            float sy = (float) y + 0.5f;
+            float x_gspace = (sx / scale_x);
+            float y_gspace = (sy / scale_y);
+
+            int winding = stbtt__compute_crossings_x(x_gspace, y_gspace, num_verts, verts); // @OPTIMIZE: this could just be a rasterization, but needs to be line vs. non-tesselated curves so a new path
+
+            for (i=0; i < num_verts; ++i) {
+               float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y;
+
+               // check against every point here rather than inside line/curve primitives -- @TODO: wrong if multiple 'moves' in a row produce a garbage point, and given culling, probably more efficient to do within line/curve
+               float dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy);
+               if (dist2 < min_dist*min_dist)
+                  min_dist = (float) STBTT_sqrt(dist2);
+
+               if (verts[i].type == STBTT_vline) {
+                  float x1 = verts[i-1].x*scale_x, y1 = verts[i-1].y*scale_y;
+
+                  // coarse culling against bbox
+                  //if (sx > STBTT_min(x0,x1)-min_dist && sx < STBTT_max(x0,x1)+min_dist &&
+                  //    sy > STBTT_min(y0,y1)-min_dist && sy < STBTT_max(y0,y1)+min_dist)
+                  float dist = (float) STBTT_fabs((x1-x0)*(y0-sy) - (y1-y0)*(x0-sx)) * precompute[i];
+                  STBTT_assert(i != 0);
+                  if (dist < min_dist) {
+                     // check position along line
+                     // x' = x0 + t*(x1-x0), y' = y0 + t*(y1-y0)
+                     // minimize (x'-sx)*(x'-sx)+(y'-sy)*(y'-sy)
+                     float dx = x1-x0, dy = y1-y0;
+                     float px = x0-sx, py = y0-sy;
+                     // minimize (px+t*dx)^2 + (py+t*dy)^2 = px*px + 2*px*dx*t + t^2*dx*dx + py*py + 2*py*dy*t + t^2*dy*dy
+                     // derivative: 2*px*dx + 2*py*dy + (2*dx*dx+2*dy*dy)*t, set to 0 and solve
+                     float t = -(px*dx + py*dy) / (dx*dx + dy*dy);
+                     if (t >= 0.0f && t <= 1.0f)
+                        min_dist = dist;
+                  }
+               } else if (verts[i].type == STBTT_vcurve) {
+                  float x2 = verts[i-1].x *scale_x, y2 = verts[i-1].y *scale_y;
+                  float x1 = verts[i  ].cx*scale_x, y1 = verts[i  ].cy*scale_y;
+                  float box_x0 = STBTT_min(STBTT_min(x0,x1),x2);
+                  float box_y0 = STBTT_min(STBTT_min(y0,y1),y2);
+                  float box_x1 = STBTT_max(STBTT_max(x0,x1),x2);
+                  float box_y1 = STBTT_max(STBTT_max(y0,y1),y2);
+                  // coarse culling against bbox to avoid computing cubic unnecessarily
+                  if (sx > box_x0-min_dist && sx < box_x1+min_dist && sy > box_y0-min_dist && sy < box_y1+min_dist) {
+                     int num=0;
+                     float ax = x1-x0, ay = y1-y0;
+                     float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2;
+                     float mx = x0 - sx, my = y0 - sy;
+                     float res[3],px,py,t,it;
+                     float a_inv = precompute[i];
+                     if (a_inv == 0.0) { // if a_inv is 0, it's 2nd degree so use quadratic formula
+                        float a = 3*(ax*bx + ay*by);
+                        float b = 2*(ax*ax + ay*ay) + (mx*bx+my*by);
+                        float c = mx*ax+my*ay;
+                        if (a == 0.0) { // if a is 0, it's linear
+                           if (b != 0.0) {
+                              res[num++] = -c/b;
+                           }
+                        } else {
+                           float discriminant = b*b - 4*a*c;
+                           if (discriminant < 0)
+                              num = 0;
+                           else {
+                              float root = (float) STBTT_sqrt(discriminant);
+                              res[0] = (-b - root)/(2*a);
+                              res[1] = (-b + root)/(2*a);
+                              num = 2; // don't bother distinguishing 1-solution case, as code below will still work
+                           }
+                        }
+                     } else {
+                        float b = 3*(ax*bx + ay*by) * a_inv; // could precompute this as it doesn't depend on sample point
+                        float c = (2*(ax*ax + ay*ay) + (mx*bx+my*by)) * a_inv;
+                        float d = (mx*ax+my*ay) * a_inv;
+                        num = stbtt__solve_cubic(b, c, d, res);
+                     }
+                     if (num >= 1 && res[0] >= 0.0f && res[0] <= 1.0f) {
+                        t = res[0], it = 1.0f - t;
+                        px = it*it*x0 + 2*t*it*x1 + t*t*x2;
+                        py = it*it*y0 + 2*t*it*y1 + t*t*y2;
+                        dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy);
+                        if (dist2 < min_dist * min_dist)
+                           min_dist = (float) STBTT_sqrt(dist2);
+                     }
+                     if (num >= 2 && res[1] >= 0.0f && res[1] <= 1.0f) {
+                        t = res[1], it = 1.0f - t;
+                        px = it*it*x0 + 2*t*it*x1 + t*t*x2;
+                        py = it*it*y0 + 2*t*it*y1 + t*t*y2;
+                        dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy);
+                        if (dist2 < min_dist * min_dist)
+                           min_dist = (float) STBTT_sqrt(dist2);
+                     }
+                     if (num >= 3 && res[2] >= 0.0f && res[2] <= 1.0f) {
+                        t = res[2], it = 1.0f - t;
+                        px = it*it*x0 + 2*t*it*x1 + t*t*x2;
+                        py = it*it*y0 + 2*t*it*y1 + t*t*y2;
+                        dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy);
+                        if (dist2 < min_dist * min_dist)
+                           min_dist = (float) STBTT_sqrt(dist2);
+                     }
+                  }
+               }
+            }
+            if (winding == 0)
+               min_dist = -min_dist;  // if outside the shape, value is negative
+            val = onedge_value + pixel_dist_scale * min_dist;
+            if (val < 0)
+               val = 0;
+            else if (val > 255)
+               val = 255;
+            data[(y-iy0)*w+(x-ix0)] = (unsigned char) val;
+         }
+      }
+      STBTT_free(precompute, info->userdata);
+      STBTT_free(verts, info->userdata);
+   }
+   return data;
+}   
+
+STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff)
+{
+   return stbtt_GetGlyphSDF(info, scale, stbtt_FindGlyphIndex(info, codepoint), padding, onedge_value, pixel_dist_scale, width, height, xoff, yoff);
+}
+
+STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata)
+{
+   STBTT_free(bitmap, userdata);
+}
 
 //////////////////////////////////////////////////////////////////////////////
 //
@@ -3970,6 +4471,10 @@ STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const
 
 // FULL VERSION HISTORY
 //
+//   1.16 (2017-07-12) SDF support
+//   1.15 (2017-03-03) make more arguments const
+//   1.14 (2017-01-16) num-fonts-in-TTC function
+//   1.13 (2017-01-02) support OpenType fonts, certain Apple fonts
 //   1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual
 //   1.11 (2016-04-02) fix unused-variable warning
 //   1.10 (2016-04-02) allow user-defined fabs() replacement

src/external/stb_vorbis.c

@@ -193,6 +193,7 @@
    #undef __forceinline
    #endif
    #define __forceinline
+   #define alloca __builtin_alloca
 #elif !defined(_MSC_VER)
    #if __GNUC__
       #define __forceinline inline

src/external/stb_vorbis.h

@@ -1,4 +1,4 @@
-// Ogg Vorbis audio decoder - v1.10 - public domain
+// Ogg Vorbis audio decoder - v1.11 - public domain
 // http://nothings.org/stb_vorbis/
 //
 // Original version written by Sean Barrett in 2007.
@@ -29,9 +29,10 @@
 //    Bernhard Wodo      Evan Balster        alxprd@github
 //    Tom Beaumont       Ingo Leitgeb        Nicolas Guillemot
 //    Phillip Bennefall  Rohit               Thiago Goulart
-//    manxorist@github   saga musix
+//    manxorist@github   saga musix          github:infatum
 //
 // Partial history:
+//    1.11    - 2017/07/23 - fix MinGW compilation 
 //    1.10    - 2017/03/03 - more robust seeking; fix negative ilog(); clear error in open_memory
 //    1.09    - 2016/04/04 - back out 'truncation of last frame' fix from previous version
 //    1.08    - 2016/04/02 - warnings; setup memory leaks; truncation of last frame