Merge remote-tracking branch 'qatar/master'

* qatar/master:
  asfdec: read attached pictures.
  apetag: reindent
  apetag: export attached covers as video streams.
  apetag: fix the amount of data read from binary tags.
  apetag: make sure avio_get_str() doesn't read more than it should.
  mov: read itunes cover art.
  snow: remove VLA in mc_block()
  intfloat: Don't use designated initializers in the public headers
  snow: remove a VLA.
  doc: Remind devs to check return values, especially for malloc() et al
  MS ATC Screen (aka MSS3) decoder
  vf_yadif: move x86 init code to x86/yadif.c
  vf_gradfun: move x86 init code to x86/gradfun.c
  roqvideo: Remove a totally unused dspcontext
  smacker: remove some unused code
  dsicin: remove dead assignment
  aacdec: remove dead assignment
  rl2: remove dead assignment
  proresenc: make a variable local to the loop where it is used
  alsdec: remove dead assignments

Conflicts:
	Changelog
	doc/developer.texi
	libavcodec/allcodecs.c
	libavcodec/avcodec.h
	libavcodec/version.h
	libavfilter/gradfun.h
	libavfilter/x86/gradfun.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Changelog

@@ -14,6 +14,7 @@ version next:
 - Microsoft Screen 1 decoder
 - join audio filter
 - audio channel mapping filter
+- Microsoft ATC Screen decoder
 - showwaves filter
 - LucasArts SMUSH playback support
 - SAMI demuxer and decoder

doc/developer.texi

@@ -489,6 +489,10 @@ send a reminder by email. Your patch should eventually be dealt with.
     Consider to add a regression test for your code.
 @item
     If you added YASM code please check that things still work with --disable-yasm
+@item
+    Make sure you check the return values of function and return appropriate
+    error codes. Especially memory allocation functions like @code{malloc()}
+    are notoriously left unchecked, which is a serious problem.
 @end enumerate
 
 @section Patch review process

doc/general.texi

@@ -565,6 +565,8 @@ following image formats are supported:
 @item LucasArts Smush        @tab     @tab  X
     @tab Used in LucasArts games.
 @item lossless MJPEG         @tab  X  @tab  X
+@item Microsoft ATC Screen   @tab     @tab  X
+    @tab Also known as Microsoft Screen 3.
 @item Microsoft RLE          @tab     @tab  X
 @item Microsoft Screen 1     @tab     @tab  X
     @tab Also known as Windows Media Video V7 Screen.

libavcodec/Makefile

@@ -327,6 +327,7 @@ OBJS-$(CONFIG_MSMPEG4V3_ENCODER)       += msmpeg4.o msmpeg4enc.o msmpeg4data.o \
                                           h263dec.o h263.o ituh263dec.o        \
                                           mpeg4videodec.o
 OBJS-$(CONFIG_MSRLE_DECODER)           += msrle.o msrledec.o
+OBJS-$(CONFIG_MSA1_DECODER)            += mss3.o
 OBJS-$(CONFIG_MSS1_DECODER)            += mss1.o
 OBJS-$(CONFIG_MSVIDEO1_DECODER)        += msvideo1.o
 OBJS-$(CONFIG_MSVIDEO1_ENCODER)        += msvideo1enc.o elbg.o

libavcodec/aacdec.c

@@ -1308,7 +1308,7 @@ static inline float *VMUL4S(float *dst, const float *v, unsigned idx,
     t.i = s.i ^ (sign & 1U<<31);
     *dst++ = v[idx>>4 & 3] * t.f;
 
-    sign <<= nz & 1; nz >>= 1;
+    sign <<= nz & 1;
     t.i = s.i ^ (sign & 1U<<31);
     *dst++ = v[idx>>6 & 3] * t.f;
 

libavcodec/allcodecs.c

@@ -165,6 +165,7 @@ void avcodec_register_all(void)
     REGISTER_DECODER (MPEG_VDPAU, mpeg_vdpau);
     REGISTER_DECODER (MPEG1_VDPAU, mpeg1_vdpau);
     REGISTER_DECODER (MPEG2_CRYSTALHD, mpeg2_crystalhd);
+    REGISTER_DECODER (MSA1, msa1);
     REGISTER_DECODER (MSMPEG4_CRYSTALHD, msmpeg4_crystalhd);
     REGISTER_DECODER (MSMPEG4V1, msmpeg4v1);
     REGISTER_ENCDEC  (MSMPEG4V2, msmpeg4v2);

libavcodec/alsdec.c

@@ -770,7 +770,7 @@ static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd)
         int          delta[8];
         unsigned int k    [8];
         unsigned int b = av_clip((av_ceil_log2(bd->block_length) - 3) >> 1, 0, 5);
-        unsigned int i = start;
+        unsigned int i;
 
         // read most significant bits
         unsigned int high;
@@ -781,7 +781,7 @@ static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd)
 
         current_res = bd->raw_samples + start;
 
-        for (sb = 0; sb < sub_blocks; sb++, i = 0) {
+        for (sb = 0; sb < sub_blocks; sb++) {
             k    [sb] = s[sb] > b ? s[sb] - b : 0;
             delta[sb] = 5 - s[sb] + k[sb];
 

libavcodec/avcodec.h

@@ -258,6 +258,7 @@ enum CodecID {
     CODEC_ID_XBM,
     CODEC_ID_ZEROCODEC,
     CODEC_ID_MSS1,
+    CODEC_ID_MSA1,
     CODEC_ID_Y41P       = MKBETAG('Y','4','1','P'),
     CODEC_ID_ESCAPE130  = MKBETAG('E','1','3','0'),
     CODEC_ID_EXR        = MKBETAG('0','E','X','R'),

libavcodec/dsicinav.c

@@ -122,7 +122,7 @@ static int cin_decode_huffman(const unsigned char *src, int src_size, unsigned c
     unsigned char *dst_end = dst + dst_size;
     const unsigned char *src_end = src + src_size;
 
-    memcpy(huff_code_table, src, 15); src += 15; src_size -= 15;
+    memcpy(huff_code_table, src, 15); src += 15;
 
     while (src < src_end) {
         huff_code = *src++;

libavcodec/proresenc_kostya.c

@@ -726,7 +726,6 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
     int sizes[4] = { 0 };
     int slice_hdr_size = 2 + 2 * (ctx->num_planes - 1);
     int frame_size, picture_size, slice_size;
-    int mbs_per_slice = ctx->mbs_per_slice;
     int pkt_size, ret;
 
     *avctx->coded_frame           = *pic;
@@ -792,7 +791,7 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
     }
 
     for (y = 0; y < ctx->mb_height; y++) {
-        mbs_per_slice = ctx->mbs_per_slice;
+        int mbs_per_slice = ctx->mbs_per_slice;
         for (x = mb = 0; x < ctx->mb_width; x += mbs_per_slice, mb++) {
             q = ctx->force_quant ? ctx->force_quant
                                  : ctx->slice_q[mb + y * ctx->slices_width];

libavcodec/rl2.c

@@ -64,7 +64,7 @@ static void rl2_rle_decode(Rl2Context *s,const unsigned char* in,int size,
     const unsigned char* back_frame = s->back_frame;
     const unsigned char* in_end = in + size;
     const unsigned char* out_end = out + stride * s->avctx->height;
-    unsigned char* line_end = out + s->avctx->width;
+    unsigned char* line_end;
 
     /** copy start of the background frame */
     for(i=0;i<=base_y;i++){

libavcodec/roqvideo.h

@@ -45,7 +45,6 @@ struct RoqTempData;
 typedef struct RoqContext {
 
     AVCodecContext *avctx;
-    DSPContext dsp;
     AVFrame frames[2];
     AVFrame *last_frame;
     AVFrame *current_frame;

libavcodec/smacker.c

@@ -137,13 +137,9 @@ static int smacker_decode_bigtree(GetBitContext *gb, HuffContext *hc, DBCtx *ctx
         return -1;
     }
     if(!get_bits1(gb)){ //Leaf
-        int val, i1, i2, b1, b2;
-        b1 = get_bits_count(gb);
+        int val, i1, i2;
         i1 = ctx->v1->table ? get_vlc2(gb, ctx->v1->table, SMKTREE_BITS, 3) : 0;
-        b1 = get_bits_count(gb) - b1;
-        b2 = get_bits_count(gb);
         i2 = ctx->v2->table ? get_vlc2(gb, ctx->v2->table, SMKTREE_BITS, 3) : 0;
-        b2 = get_bits_count(gb) - b2;
         if (i1 < 0 || i2 < 0)
             return -1;
         val = ctx->recode1[i1] | (ctx->recode2[i2] << 8);

libavcodec/snow.c

@@ -140,7 +140,7 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int
 
     int x, y, b, r, l;
     int16_t tmpIt   [64*(32+HTAPS_MAX)];
-    uint8_t tmp2t[3][stride*(32+HTAPS_MAX)];
+    uint8_t tmp2t[3][64*(32+HTAPS_MAX)];
     int16_t *tmpI= tmpIt;
     uint8_t *tmp2= tmp2t[0];
     const uint8_t *hpel[11];
@@ -178,7 +178,7 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int
                 tmp2[x]= am;
             }
             tmpI+= 64;
-            tmp2+= stride;
+            tmp2+= 64;
             src += stride;
         }
         src -= stride*y;
@@ -207,7 +207,7 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int
                 tmp2[x]= am;
             }
             src += stride;
-            tmp2+= stride;
+            tmp2+= 64;
         }
         src -= stride*y;
     }
@@ -234,12 +234,12 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int
                 tmp2[x]= am;
             }
             tmpI+= 64;
-            tmp2+= stride;
+            tmp2+= 64;
         }
     }
 
     hpel[ 0]= src;
-    hpel[ 1]= tmp2t[0] + stride*(HTAPS_MAX/2-1);
+    hpel[ 1]= tmp2t[0] + 64*(HTAPS_MAX/2-1);
     hpel[ 2]= src + 1;
 
     hpel[ 4]= tmp2t[1];
@@ -247,14 +247,21 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int
     hpel[ 6]= tmp2t[1] + 1;
 
     hpel[ 8]= src + stride;
-    hpel[ 9]= hpel[1] + stride;
+    hpel[ 9]= hpel[1] + 64;
     hpel[10]= hpel[8] + 1;
 
+#define MC_STRIDE(x) (needs[x] ? 64 : stride)
+
     if(b==15){
-        const uint8_t *src1= hpel[dx/8 + dy/8*4  ];
-        const uint8_t *src2= hpel[dx/8 + dy/8*4+1];
-        const uint8_t *src3= hpel[dx/8 + dy/8*4+4];
-        const uint8_t *src4= hpel[dx/8 + dy/8*4+5];
+        int dxy = dx / 8 + dy / 8 * 4;
+        const uint8_t *src1 = hpel[dxy    ];
+        const uint8_t *src2 = hpel[dxy + 1];
+        const uint8_t *src3 = hpel[dxy + 4];
+        const uint8_t *src4 = hpel[dxy + 5];
+        int stride1 = MC_STRIDE(dxy);
+        int stride2 = MC_STRIDE(dxy + 1);
+        int stride3 = MC_STRIDE(dxy + 4);
+        int stride4 = MC_STRIDE(dxy + 5);
         dx&=7;
         dy&=7;
         for(y=0; y < b_h; y++){
@@ -262,23 +269,25 @@ static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int
                 dst[x]= ((8-dx)*(8-dy)*src1[x] + dx*(8-dy)*src2[x]+
                          (8-dx)*   dy *src3[x] + dx*   dy *src4[x]+32)>>6;
             }
-            src1+=stride;
-            src2+=stride;
-            src3+=stride;
-            src4+=stride;
+            src1+=stride1;
+            src2+=stride2;
+            src3+=stride3;
+            src4+=stride4;
             dst +=stride;
         }
     }else{
         const uint8_t *src1= hpel[l];
         const uint8_t *src2= hpel[r];
+        int stride1 = MC_STRIDE(l);
+        int stride2 = MC_STRIDE(r);
         int a= weight[((dx&7) + (8*(dy&7)))];
         int b= 8-a;
         for(y=0; y < b_h; y++){
             for(x=0; x < b_w; x++){
                 dst[x]= (a*src1[x] + b*src2[x] + 4)>>3;
             }
-            src1+=stride;
-            src2+=stride;
+            src1+=stride1;
+            src2+=stride2;
             dst +=stride;
         }
     }

libavcodec/snowenc.c

@@ -666,7 +666,7 @@ static inline int get_block_bits(SnowContext *s, int x, int y, int w){
     }
 }
 
-static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, const uint8_t *obmc_edged){
+static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uint8_t (*obmc_edged)[MB_SIZE * 2]){
     Plane *p= &s->plane[plane_index];
     const int block_size = MB_SIZE >> s->block_max_depth;
     const int block_w    = plane_index ? block_size>>s->chroma_h_shift : block_size;
@@ -698,7 +698,7 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, con
     ff_snow_pred_block(s, cur, tmp, ref_stride, sx, sy, block_w*2, block_h*2, &s->block[mb_x + mb_y*b_stride], plane_index, w, h);
 
     for(y=y0; y<y1; y++){
-        const uint8_t *obmc1= obmc_edged + y*obmc_stride;
+        const uint8_t *obmc1= obmc_edged[y];
         const IDWTELEM *pred1 = pred + y*obmc_stride;
         uint8_t *cur1 = cur + y*ref_stride;
         uint8_t *dst1 = dst + sx + (sy+y)*ref_stride;
@@ -962,7 +962,7 @@ static int encode_subband(SnowContext *s, SubBand *b, const IDWTELEM *src, const
 //    encode_subband_dzr(s, b, src, parent, stride, orientation);
 }
 
-static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, const uint8_t *obmc_edged, int *best_rd){
+static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd){
     const int b_stride= s->b_width << s->block_max_depth;
     BlockNode *block= &s->block[mb_x + mb_y * b_stride];
     BlockNode backup= *block;
@@ -1003,7 +1003,7 @@ static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int
 
 /* special case for int[2] args we discard afterwards,
  * fixes compilation problem with gcc 2.95 */
-static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, const uint8_t *obmc_edged, int *best_rd){
+static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd){
     int p[2] = {p0, p1};
     return check_block(s, mb_x, mb_y, p, 0, obmc_edged, best_rd);
 }
@@ -1083,7 +1083,7 @@ static void iterative_me(SnowContext *s){
                 BlockNode *blb= mb_x           && mb_y+1<b_height ? &s->block[index+b_stride-1] : NULL;
                 BlockNode *brb= mb_x+1<b_width && mb_y+1<b_height ? &s->block[index+b_stride+1] : NULL;
                 const int b_w= (MB_SIZE >> s->block_max_depth);
-                uint8_t obmc_edged[b_w*2][b_w*2];
+                uint8_t obmc_edged[MB_SIZE * 2][MB_SIZE * 2];
 
                 if(pass && (block->type & BLOCK_OPT))
                     continue;
@@ -1098,7 +1098,8 @@ static void iterative_me(SnowContext *s){
                 //FIXME precalculate
                 {
                     int x, y;
-                    memcpy(obmc_edged, ff_obmc_tab[s->block_max_depth], b_w*b_w*4);
+                    for (y = 0; y < b_w * 2; y++)
+                        memcpy(obmc_edged[y], ff_obmc_tab[s->block_max_depth] + y * b_w * 2, b_w * 2);
                     if(mb_x==0)
                         for(y=0; y<b_w*2; y++)
                             memset(obmc_edged[y], obmc_edged[y][0] + obmc_edged[y][b_w-1], b_w);
@@ -1153,9 +1154,9 @@ static void iterative_me(SnowContext *s){
                 // get previous score (cannot be cached due to OBMC)
                 if(pass > 0 && (block->type&BLOCK_INTRA)){
                     int color0[3]= {block->color[0], block->color[1], block->color[2]};
-                    check_block(s, mb_x, mb_y, color0, 1, *obmc_edged, &best_rd);
+                    check_block(s, mb_x, mb_y, color0, 1, obmc_edged, &best_rd);
                 }else
-                    check_block_inter(s, mb_x, mb_y, block->mx, block->my, *obmc_edged, &best_rd);
+                    check_block_inter(s, mb_x, mb_y, block->mx, block->my, obmc_edged, &best_rd);
 
                 ref_b= *block;
                 ref_rd= best_rd;
@@ -1166,16 +1167,16 @@ static void iterative_me(SnowContext *s){
                     block->ref= ref;
                     best_rd= INT_MAX;
 
-                    check_block_inter(s, mb_x, mb_y, mvr[0][0], mvr[0][1], *obmc_edged, &best_rd);
-                    check_block_inter(s, mb_x, mb_y, 0, 0, *obmc_edged, &best_rd);
+                    check_block_inter(s, mb_x, mb_y, mvr[0][0], mvr[0][1], obmc_edged, &best_rd);
+                    check_block_inter(s, mb_x, mb_y, 0, 0, obmc_edged, &best_rd);
                     if(tb)
-                        check_block_inter(s, mb_x, mb_y, mvr[-b_stride][0], mvr[-b_stride][1], *obmc_edged, &best_rd);
+                        check_block_inter(s, mb_x, mb_y, mvr[-b_stride][0], mvr[-b_stride][1], obmc_edged, &best_rd);
                     if(lb)
-                        check_block_inter(s, mb_x, mb_y, mvr[-1][0], mvr[-1][1], *obmc_edged, &best_rd);
+                        check_block_inter(s, mb_x, mb_y, mvr[-1][0], mvr[-1][1], obmc_edged, &best_rd);
                     if(rb)
-                        check_block_inter(s, mb_x, mb_y, mvr[1][0], mvr[1][1], *obmc_edged, &best_rd);
+                        check_block_inter(s, mb_x, mb_y, mvr[1][0], mvr[1][1], obmc_edged, &best_rd);
                     if(bb)
-                        check_block_inter(s, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], *obmc_edged, &best_rd);
+                        check_block_inter(s, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], obmc_edged, &best_rd);
 
                     /* fullpel ME */
                     //FIXME avoid subpel interpolation / round to nearest integer
@@ -1183,10 +1184,10 @@ static void iterative_me(SnowContext *s){
                         dia_change=0;
                         for(i=0; i<FFMAX(s->avctx->dia_size, 1); i++){
                             for(j=0; j<i; j++){
-                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+4*(i-j), block->my+(4*j), *obmc_edged, &best_rd);
-                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx-4*(i-j), block->my-(4*j), *obmc_edged, &best_rd);
-                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+4*(i-j), block->my-(4*j), *obmc_edged, &best_rd);
-                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx-4*(i-j), block->my+(4*j), *obmc_edged, &best_rd);
+                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+4*(i-j), block->my+(4*j), obmc_edged, &best_rd);
+                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx-4*(i-j), block->my-(4*j), obmc_edged, &best_rd);
+                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+4*(i-j), block->my-(4*j), obmc_edged, &best_rd);
+                                dia_change |= check_block_inter(s, mb_x, mb_y, block->mx-4*(i-j), block->my+(4*j), obmc_edged, &best_rd);
                             }
                         }
                     }while(dia_change);
@@ -1195,7 +1196,7 @@ static void iterative_me(SnowContext *s){
                         static const int square[8][2]= {{+1, 0},{-1, 0},{ 0,+1},{ 0,-1},{+1,+1},{-1,-1},{+1,-1},{-1,+1},};
                         dia_change=0;
                         for(i=0; i<8; i++)
-                            dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+square[i][0], block->my+square[i][1], *obmc_edged, &best_rd);
+                            dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+square[i][0], block->my+square[i][1], obmc_edged, &best_rd);
                     }while(dia_change);
                     //FIXME or try the standard 2 pass qpel or similar
 
@@ -1208,7 +1209,7 @@ static void iterative_me(SnowContext *s){
                 }
                 best_rd= ref_rd;
                 *block= ref_b;
-                check_block(s, mb_x, mb_y, color, 1, *obmc_edged, &best_rd);
+                check_block(s, mb_x, mb_y, color, 1, obmc_edged, &best_rd);
                 //FIXME RD style color selection
                 if(!same_block(block, &backup)){
                     if(tb ) tb ->type &= ~BLOCK_OPT;

libavcodec/version.h

@@ -27,7 +27,7 @@
  */
 
 #define LIBAVCODEC_VERSION_MAJOR 54
-#define LIBAVCODEC_VERSION_MINOR  31
+#define LIBAVCODEC_VERSION_MINOR  32
 #define LIBAVCODEC_VERSION_MICRO 100
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \

libavfilter/gradfun.h

@@ -37,12 +37,9 @@ typedef struct {
     void (*blur_line) (uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width);
 } GradFunContext;
 
+void ff_gradfun_init_x86(GradFunContext *gf);
+
 void ff_gradfun_filter_line_c(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers);
 void ff_gradfun_blur_line_c(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width);
 
-void ff_gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers);
-void ff_gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers);
-
-void ff_gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width);
-
 #endif /* AVFILTER_GRADFUN_H */

libavfilter/vf_gradfun.c

@@ -123,7 +123,6 @@ static av_cold int init(AVFilterContext *ctx, const char *args)
     GradFunContext *gf = ctx->priv;
     float thresh = 1.2;
     int radius = 16;
-    int cpu_flags = av_get_cpu_flags();
 
     if (args)
         sscanf(args, "%f:%d", &thresh, &radius);
@@ -135,12 +134,8 @@ static av_cold int init(AVFilterContext *ctx, const char *args)
     gf->blur_line = ff_gradfun_blur_line_c;
     gf->filter_line = ff_gradfun_filter_line_c;
 
-    if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX2)
-        gf->filter_line = ff_gradfun_filter_line_mmx2;
-    if (HAVE_SSSE3 && cpu_flags & AV_CPU_FLAG_SSSE3)
-        gf->filter_line = ff_gradfun_filter_line_ssse3;
-    if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
-        gf->blur_line = ff_gradfun_blur_line_sse2;
+    if (HAVE_MMX)
+        ff_gradfun_init_x86(gf);
 
     av_log(ctx, AV_LOG_VERBOSE, "threshold:%.2f radius:%d\n", thresh, gf->radius);
 

libavfilter/vf_yadif.c

@@ -30,42 +30,6 @@
 #undef NDEBUG
 #include <assert.h>
 
-typedef struct {
-    /**
-     * 0: send 1 frame for each frame
-     * 1: send 1 frame for each field
-     * 2: like 0 but skips spatial interlacing check
-     * 3: like 1 but skips spatial interlacing check
-     */
-    int mode;
-
-    /**
-     *  0: top field first
-     *  1: bottom field first
-     * -1: auto-detection
-     */
-    int parity;
-
-    int frame_pending;
-
-    /**
-     *  0: deinterlace all frames
-     *  1: only deinterlace frames marked as interlaced
-     */
-    int auto_enable;
-
-    AVFilterBufferRef *cur;
-    AVFilterBufferRef *next;
-    AVFilterBufferRef *prev;
-    AVFilterBufferRef *out;
-    void (*filter_line)(uint8_t *dst,
-                        uint8_t *prev, uint8_t *cur, uint8_t *next,
-                        int w, int prefs, int mrefs, int parity, int mode);
-
-    const AVPixFmtDescriptor *csp;
-    int eof;
-} YADIFContext;
-
 #define CHECK(j)\
     {   int score = FFABS(cur[mrefs-1+(j)] - cur[prefs-1-(j)])\
                   + FFABS(cur[mrefs  +(j)] - cur[prefs  -(j)])\
@@ -400,7 +364,6 @@ static int query_formats(AVFilterContext *ctx)
 static av_cold int init(AVFilterContext *ctx, const char *args)
 {
     YADIFContext *yadif = ctx->priv;
-    int cpu_flags = av_get_cpu_flags();
 
     yadif->mode = 0;
     yadif->parity = -1;
@@ -410,12 +373,9 @@ static av_cold int init(AVFilterContext *ctx, const char *args)
     if (args) sscanf(args, "%d:%d:%d", &yadif->mode, &yadif->parity, &yadif->auto_enable);
 
     yadif->filter_line = filter_line_c;
-    if (HAVE_SSSE3 && cpu_flags & AV_CPU_FLAG_SSSE3)
-        yadif->filter_line = ff_yadif_filter_line_ssse3;
-    else if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
-        yadif->filter_line = ff_yadif_filter_line_sse2;
-    else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX)
-        yadif->filter_line = ff_yadif_filter_line_mmx;
+
+    if (HAVE_MMX)
+        ff_yadif_init_x86(yadif);
 
     av_log(ctx, AV_LOG_VERBOSE, "mode:%d parity:%d auto_enable:%d\n", yadif->mode, yadif->parity, yadif->auto_enable);
 

libavfilter/x86/gradfun.c

@@ -18,6 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/attributes.h"
 #include "libavutil/cpu.h"
 #include "libavutil/x86_cpu.h"
 #include "libavfilter/gradfun.h"
@@ -25,9 +26,9 @@
 DECLARE_ALIGNED(16, static const uint16_t, pw_7f)[8] = {0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F};
 DECLARE_ALIGNED(16, static const uint16_t, pw_ff)[8] = {0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF};
 
-void ff_gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
+#if HAVE_MMX2
+static void gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
 {
-#if HAVE_MMX
     intptr_t x;
     if (width & 3) {
         x = width & ~3;
@@ -70,12 +71,12 @@ void ff_gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_
          "rm"(thresh), "m"(*dithers), "m"(*pw_7f)
         :"memory"
     );
-#endif
 }
+#endif
 
-void ff_gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
-{
 #if HAVE_SSSE3
+static void gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
+{
     intptr_t x;
     if (width & 7) {
         // could be 10% faster if I somehow eliminated this
@@ -117,12 +118,12 @@ void ff_gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16
          "rm"(thresh), "m"(*dithers), "m"(*pw_7f)
         :"memory"
     );
-#endif // HAVE_SSSE3
 }
+#endif // HAVE_SSSE3
 
-void ff_gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width)
-{
 #if HAVE_SSE
+static void gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, const uint8_t *src, int src_linesize, int width)
+{
 #define BLURV(load)\
     intptr_t x = -2*width;\
     __asm__ volatile(\
@@ -160,5 +161,17 @@ void ff_gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1
     } else {
         BLURV("movdqa");
     }
+}
 #endif // HAVE_SSE
+
+av_cold void ff_gradfun_init_x86(GradFunContext *gf)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2)
+        gf->filter_line = gradfun_filter_line_mmx2;
+    if (HAVE_SSSE3 && cpu_flags & AV_CPU_FLAG_SSSE3)
+        gf->filter_line = gradfun_filter_line_ssse3;
+    if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
+        gf->blur_line = gradfun_blur_line_sse2;
 }

libavfilter/x86/yadif.c

@@ -18,6 +18,7 @@
  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  */
 
+#include "libavutil/attributes.h"
 #include "libavutil/cpu.h"
 #include "libavutil/x86_cpu.h"
 #include "libavcodec/x86/dsputil_mmx.h"
@@ -47,3 +48,15 @@ DECLARE_ASM_CONST(16, const xmm_reg, pw_1) = {0x0001000100010001ULL, 0x000100010
 #define RENAME(a) a ## _mmx
 #include "yadif_template.c"
 #endif
+
+av_cold void ff_yadif_init_x86(YADIFContext *yadif)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX)
+        yadif->filter_line = yadif_filter_line_mmx;
+    if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
+        yadif->filter_line = yadif_filter_line_sse2;
+    if (HAVE_SSSE3 && cpu_flags & AV_CPU_FLAG_SSSE3)
+        yadif->filter_line = yadif_filter_line_ssse3;
+}

libavfilter/x86/yadif_template.c

@@ -103,9 +103,9 @@
             "por       "MM"5, "MM"3 \n\t"\
             MOVQ"      "MM"3, "MM"1 \n\t"
 
-void RENAME(ff_yadif_filter_line)(uint8_t *dst,
-                                  uint8_t *prev, uint8_t *cur, uint8_t *next,
-                                  int w, int prefs, int mrefs, int parity, int mode)
+static void RENAME(yadif_filter_line)(uint8_t *dst, uint8_t *prev, uint8_t *cur,
+                                      uint8_t *next, int w, int prefs,
+                                      int mrefs, int parity, int mode)
 {
     uint8_t tmp[5*16];
     uint8_t *tmpA= (uint8_t*)(((uint64_t)(tmp+15)) & ~15);

libavfilter/yadif.h

@@ -19,18 +19,45 @@
 #ifndef AVFILTER_YADIF_H
 #define AVFILTER_YADIF_H
 
+#include "libavutil/pixdesc.h"
 #include "avfilter.h"
 
-void ff_yadif_filter_line_mmx(uint8_t *dst,
-                              uint8_t *prev, uint8_t *cur, uint8_t *next,
-                              int w, int prefs, int mrefs, int parity, int mode);
+typedef struct {
+    /**
+     * 0: send 1 frame for each frame
+     * 1: send 1 frame for each field
+     * 2: like 0 but skips spatial interlacing check
+     * 3: like 1 but skips spatial interlacing check
+     */
+    int mode;
 
-void ff_yadif_filter_line_sse2(uint8_t *dst,
-                               uint8_t *prev, uint8_t *cur, uint8_t *next,
-                               int w, int prefs, int mrefs, int parity, int mode);
+    /**
+     *  0: top field first
+     *  1: bottom field first
+     * -1: auto-detection
+     */
+    int parity;
 
-void ff_yadif_filter_line_ssse3(uint8_t *dst,
-                                uint8_t *prev, uint8_t *cur, uint8_t *next,
-                                int w, int prefs, int mrefs, int parity, int mode);
+    int frame_pending;
+
+    /**
+     *  0: deinterlace all frames
+     *  1: only deinterlace frames marked as interlaced
+     */
+    int auto_enable;
+
+    AVFilterBufferRef *cur;
+    AVFilterBufferRef *next;
+    AVFilterBufferRef *prev;
+    AVFilterBufferRef *out;
+    void (*filter_line)(uint8_t *dst,
+                        uint8_t *prev, uint8_t *cur, uint8_t *next,
+                        int w, int prefs, int mrefs, int parity, int mode);
+
+    const AVPixFmtDescriptor *csp;
+    int eof;
+} YADIFContext;
+
+void ff_yadif_init_x86(YADIFContext *yadif);
 
 #endif /* AVFILTER_YADIF_H */

libavformat/apetag.c

@@ -24,6 +24,7 @@
 #include "libavutil/dict.h"
 #include "avformat.h"
 #include "apetag.h"
+#include "internal.h"
 
 #define APE_TAG_VERSION               2000
 #define APE_TAG_FOOTER_BYTES          32
@@ -56,20 +57,47 @@ static int ape_tag_read_field(AVFormatContext *s)
         return -1;
     if (flags & APE_TAG_FLAG_IS_BINARY) {
         uint8_t filename[1024];
+        enum CodecID id;
         AVStream *st = avformat_new_stream(s, NULL);
         if (!st)
             return AVERROR(ENOMEM);
-        avio_get_str(pb, INT_MAX, filename, sizeof(filename));
-        st->codec->extradata = av_malloc(size + FF_INPUT_BUFFER_PADDING_SIZE);
-        if (!st->codec->extradata)
-            return AVERROR(ENOMEM);
-        if (avio_read(pb, st->codec->extradata, size) != size) {
-            av_freep(&st->codec->extradata);
-            return AVERROR(EIO);
+
+        size -= avio_get_str(pb, size, filename, sizeof(filename));
+        if (size <= 0) {
+            av_log(s, AV_LOG_WARNING, "Skipping binary tag '%s'.\n", key);
+            return 0;
         }
-        st->codec->extradata_size = size;
+
         av_dict_set(&st->metadata, key, filename, 0);
-        st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
+
+        if ((id = ff_guess_image2_codec(filename)) != CODEC_ID_NONE) {
+            AVPacket pkt;
+            int ret;
+
+            ret = av_get_packet(s->pb, &pkt, size);
+            if (ret < 0) {
+                av_log(s, AV_LOG_ERROR, "Error reading cover art.\n");
+                return ret;
+            }
+
+            st->disposition      |= AV_DISPOSITION_ATTACHED_PIC;
+            st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
+            st->codec->codec_id   = id;
+
+            st->attached_pic              = pkt;
+            st->attached_pic.stream_index = st->index;
+            st->attached_pic.flags       |= AV_PKT_FLAG_KEY;
+        } else {
+            st->codec->extradata = av_malloc(size + FF_INPUT_BUFFER_PADDING_SIZE);
+            if (!st->codec->extradata)
+                return AVERROR(ENOMEM);
+            if (avio_read(pb, st->codec->extradata, size) != size) {
+                av_freep(&st->codec->extradata);
+                return AVERROR(EIO);
+            }
+            st->codec->extradata_size = size;
+            st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
+        }
     } else {
         value = av_malloc(size+1);
         if (!value)

libavformat/asfdec.c

@@ -30,6 +30,7 @@
 #include "avformat.h"
 #include "internal.h"
 #include "avio_internal.h"
+#include "id3v2.h"
 #include "riff.h"
 #include "asf.h"
 #include "asfcrypt.h"
@@ -163,6 +164,101 @@ static int get_value(AVIOContext *pb, int type){
     }
 }
 
+/* MSDN claims that this should be "compatible with the ID3 frame, APIC",
+ * but in reality this is only loosely similar */
+static int asf_read_picture(AVFormatContext *s, int len)
+{
+    AVPacket pkt = { 0 };
+    const CodecMime *mime = ff_id3v2_mime_tags;
+    enum  CodecID      id = CODEC_ID_NONE;
+    char mimetype[64];
+    uint8_t  *desc = NULL;
+    ASFStream *ast = NULL;
+    AVStream   *st = NULL;
+    int ret, type, picsize, desc_len;
+
+    /* type + picsize + mime + desc */
+    if (len < 1 + 4 + 2 + 2) {
+        av_log(s, AV_LOG_ERROR, "Invalid attached picture size: %d.\n", len);
+        return AVERROR_INVALIDDATA;
+    }
+
+    /* picture type */
+    type = avio_r8(s->pb);
+    len--;
+    if (type >= FF_ARRAY_ELEMS(ff_id3v2_picture_types) || type < 0) {
+        av_log(s, AV_LOG_WARNING, "Unknown attached picture type: %d.\n", type);
+        type = 0;
+    }
+
+    /* picture data size */
+    picsize = avio_rl32(s->pb);
+    len -= 4;
+
+    /* picture MIME type */
+    len -= avio_get_str16le(s->pb, len, mimetype, sizeof(mimetype));
+    while (mime->id != CODEC_ID_NONE) {
+        if (!strncmp(mime->str, mimetype, sizeof(mimetype))) {
+            id = mime->id;
+            break;
+        }
+        mime++;
+    }
+    if (id == CODEC_ID_NONE) {
+        av_log(s, AV_LOG_ERROR, "Unknown attached picture mimetype: %s.\n",
+               mimetype);
+        return 0;
+    }
+
+    if (picsize >= len) {
+        av_log(s, AV_LOG_ERROR, "Invalid attached picture data size: %d >= %d.\n",
+               picsize, len);
+        return AVERROR_INVALIDDATA;
+    }
+
+    /* picture description */
+    desc_len = (len - picsize) * 2 + 1;
+    desc     = av_malloc(desc_len);
+    if (!desc)
+        return AVERROR(ENOMEM);
+    len -= avio_get_str16le(s->pb, len - picsize, desc, desc_len);
+
+    ret = av_get_packet(s->pb, &pkt, picsize);
+    if (ret < 0)
+        goto fail;
+
+    st = avformat_new_stream(s, NULL);
+    ast = av_mallocz(sizeof(*ast));
+    if (!st || !ast) {
+        ret = AVERROR(ENOMEM);
+        goto fail;
+    }
+    st->priv_data = ast;
+
+    st->disposition      |= AV_DISPOSITION_ATTACHED_PIC;
+    st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
+    st->codec->codec_id   = id;
+
+    st->attached_pic      = pkt;
+    st->attached_pic.stream_index = st->index;
+    st->attached_pic.flags |= AV_PKT_FLAG_KEY;
+
+    if (*desc)
+        av_dict_set(&st->metadata, "title", desc, AV_DICT_DONT_STRDUP_VAL);
+    else
+        av_freep(&desc);
+
+    av_dict_set(&st->metadata, "comment", ff_id3v2_picture_types[type], 0);
+
+    return 0;
+
+fail:
+    av_freep(&ast);
+    av_freep(&desc);
+    av_free_packet(&pkt);
+    return ret;
+}
+
 static void get_tag(AVFormatContext *s, const char *key, int type, int len)
 {
     char *value;
@@ -183,6 +279,9 @@ static void get_tag(AVFormatContext *s, const char *key, int type, int len)
     } else if (type > 1 && type <= 5) {  // boolean or DWORD or QWORD or WORD
         uint64_t num = get_value(s->pb, type);
         snprintf(value, len, "%"PRIu64, num);
+    } else if (type == 1 && !strcmp(key, "WM/Picture")) { // handle cover art
+        asf_read_picture(s, len);
+        goto finish;
     } else {
         av_log(s, AV_LOG_DEBUG, "Unsupported value type %d in tag %s.\n", type, key);
         goto finish;

libavformat/mov.c

@@ -168,6 +168,48 @@ static int mov_read_mac_string(MOVContext *c, AVIOContext *pb, int len,
     return p - dst;
 }
 
+static int mov_read_covr(MOVContext *c, AVIOContext *pb, int type, int len)
+{
+    AVPacket pkt;
+    AVStream *st;
+    MOVStreamContext *sc;
+    enum CodecID id;
+    int ret;
+
+    switch (type) {
+    case 0xd:  id = CODEC_ID_MJPEG; break;
+    case 0xe:  id = CODEC_ID_PNG;   break;
+    case 0x1b: id = CODEC_ID_BMP;   break;
+    default:
+        av_log(c->fc, AV_LOG_WARNING, "Unknown cover type: 0x%x.\n", type);
+        avio_skip(pb, len);
+        return 0;
+    }
+
+    st = avformat_new_stream(c->fc, NULL);
+    if (!st)
+        return AVERROR(ENOMEM);
+    sc = av_mallocz(sizeof(*sc));
+    if (!sc)
+        return AVERROR(ENOMEM);
+    st->priv_data = sc;
+
+    ret = av_get_packet(pb, &pkt, len);
+    if (ret < 0)
+        return ret;
+
+    st->disposition              |= AV_DISPOSITION_ATTACHED_PIC;
+
+    st->attached_pic              = pkt;
+    st->attached_pic.stream_index = st->index;
+    st->attached_pic.flags       |= AV_PKT_FLAG_KEY;
+
+    st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
+    st->codec->codec_id   = id;
+
+    return 0;
+}
+
 static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
 {
 #ifdef MOV_EXPORT_ALL_METADATA
@@ -228,6 +270,14 @@ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
             avio_rb32(pb); // unknown
             str_size = data_size - 16;
             atom.size -= 16;
+
+            if (atom.type == MKTAG('c', 'o', 'v', 'r')) {
+                int ret = mov_read_covr(c, pb, data_type, str_size);
+                if (ret < 0) {
+                    av_log(c->fc, AV_LOG_ERROR, "Error parsing cover art.\n");
+                    return ret;
+                }
+            }
         } else return 0;
     } else if (atom.size > 4 && key && !c->itunes_metadata) {
         str_size = avio_rb16(pb); // string length

libavformat/riff.c

@@ -308,6 +308,7 @@ const AVCodecTag ff_codec_bmp_tags[] = {
     { CODEC_ID_FLIC,         MKTAG('A', 'F', 'L', 'C') },
     { CODEC_ID_EXR,          MKTAG('e', 'x', 'r', ' ') },
     { CODEC_ID_MSS1,         MKTAG('M', 'S', 'S', '1') },
+    { CODEC_ID_MSA1,         MKTAG('M', 'S', 'A', '1') },
     { CODEC_ID_NONE,         0 }
 };
 

libavutil/intfloat.h

@@ -39,7 +39,8 @@ union av_intfloat64 {
  */
 static av_always_inline float av_int2float(uint32_t i)
 {
-    union av_intfloat32 v = { .i = i };
+    union av_intfloat32 v;
+    v.i = i;
     return v.f;
 }
 
@@ -48,7 +49,8 @@ static av_always_inline float av_int2float(uint32_t i)
  */
 static av_always_inline uint32_t av_float2int(float f)
 {
-    union av_intfloat32 v = { .f = f };
+    union av_intfloat32 v;
+    v.f = f;
     return v.i;
 }
 
@@ -57,7 +59,8 @@ static av_always_inline uint32_t av_float2int(float f)
  */
 static av_always_inline double av_int2double(uint64_t i)
 {
-    union av_intfloat64 v = { .i = i };
+    union av_intfloat64 v;
+    v.i = i;
     return v.f;
 }
 
@@ -66,7 +69,8 @@ static av_always_inline double av_int2double(uint64_t i)
  */
 static av_always_inline uint64_t av_double2int(double f)
 {
-    union av_intfloat64 v = { .f = f };
+    union av_intfloat64 v;
+    v.f = f;
     return v.i;
 }