--- decoder.c	2004/02/03 02:56:57	1.49.2.31
+++ decoder.c	2004/08/29 11:36:22	1.51.2.8
@@ -20,7 +20,7 @@
  *  along with this program ; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: decoder.c,v 1.49.2.31 2004/02/03 02:56:57 syskin Exp $
+ * $Id: decoder.c,v 1.51.2.8 2004/08/29 11:36:22 edgomez Exp $
  *
  ****************************************************************************/
 
@@ -77,6 +77,8 @@
 		xvid_free(dec->last_mbs);
 	if (dec->mbs)
 		xvid_free(dec->mbs);
+	if (dec->qscale)
+		xvid_free(dec->qscale);
 
 	/* realloc */
 	dec->mb_width = (dec->width + 15) / 16;
@@ -161,6 +163,13 @@
 
 	memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
 
+	/* nothing happens if that fails */
+	dec->qscale =
+		xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE);
+	
+	if (dec->qscale)
+		memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height);
+
 	return 0;
 }
 
@@ -200,9 +209,9 @@
 	/* image based GMC */
 	image_null(&dec->gmc);
 
-
 	dec->mbs = NULL;
 	dec->last_mbs = NULL;
+	dec->qscale = NULL;
 
 	init_timer();
 	init_postproc(&dec->postproc);
@@ -213,6 +222,7 @@
 	dec->time = dec->time_base = dec->last_time_base = 0;
 	dec->low_delay = 0;
 	dec->packed_mode = 0;
+	dec->time_inc_resolution = 1; /* until VOL header says otherwise */
 
 	dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
 
@@ -228,6 +238,7 @@
 {
 	xvid_free(dec->last_mbs);
 	xvid_free(dec->mbs);
+	xvid_free(dec->qscale);
 
 	/* image based GMC */
 	image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
@@ -328,7 +339,7 @@
 		stop_coding_timer();
 
 		start_timer();
-		add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors);
+		add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version);
 		stop_prediction_timer();
 
 		start_timer();
@@ -449,6 +460,42 @@
 	stop_transfer_timer();
 }
 
+static void __inline
+validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec)
+{
+	/* clip a vector to valid range 
+	   prevents crashes if bitstream is broken 
+	*/
+	int shift = 5 + dec->quarterpel;
+	int xborder_high = (int)(dec->mb_width - x_pos) << shift;
+	int xborder_low = (-(int)x_pos-1) << shift;
+	int yborder_high = (int)(dec->mb_height - y_pos) << shift;
+	int yborder_low = (-(int)y_pos-1) << shift;
+
+#define CHECK_MV(mv) \
+	do { \
+	if ((mv).x > xborder_high) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \
+		(mv).x = xborder_high; \
+	} else if ((mv).x < xborder_low) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \
+		(mv).x = xborder_low; \
+	} \
+	if ((mv).y > yborder_high) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \
+		(mv).y = yborder_high; \
+	} else if ((mv).y < yborder_low) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \
+		(mv).y = yborder_low; \
+	} \
+	} while (0)
+
+	CHECK_MV(mv[0]);
+	CHECK_MV(mv[1]);
+	CHECK_MV(mv[2]);
+	CHECK_MV(mv[3]);
+}
+
 /* decode an inter macroblock */
 static void
 decoder_mbinter(DECODER * dec,
@@ -486,6 +533,8 @@
 			mv[i] = pMB->mvs[i];
 	}
 
+	validate_vector(mv, x_pos, y_pos, dec);
+
 	start_timer();
 
 	if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
@@ -779,7 +828,7 @@
 
 	start_timer();
 	image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
-					dec->width, dec->height);
+					dec->width, dec->height, dec->bs_version);
 	stop_edges_timer();
 
 	if (gmc_warp) {
@@ -937,7 +986,9 @@
 get_b_motion_vector(Bitstream * bs,
 					VECTOR * mv,
 					int fcode,
-					const VECTOR pmv)
+					const VECTOR pmv,
+					const DECODER * const dec,
+					const int x, const int y)
 {
 	const int scale_fac = 1 << (fcode - 1);
 	const int high = (32 * scale_fac) - 1;
@@ -969,7 +1020,7 @@
 decoder_bf_interpolate_mbinter(DECODER * dec,
 								IMAGE forward,
 								IMAGE backward,
-								const MACROBLOCK * pMB,
+								MACROBLOCK * pMB,
 								const uint32_t x_pos,
 								const uint32_t y_pos,
 								Bitstream * bs,
@@ -986,6 +1037,9 @@
 	pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
 	pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
 
+	validate_vector(pMB->mvs, x_pos, y_pos, dec);
+	validate_vector(pMB->b_mvs, x_pos, y_pos, dec);
+
 	if (!direct) {
 		uv_dx = pMB->mvs[0].x;
 		uv_dy = pMB->mvs[0].y;
@@ -1052,8 +1106,8 @@
 							pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
 		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
 							pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
-		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8,
-							16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8,
+							pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
 	}
 
 	interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
@@ -1100,32 +1154,32 @@
 	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
 						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
 						dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
 						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
 						dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
 						dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
 						dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
 						dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
 						dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
-						stride, 1, 8);
+						stride, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 1, 8);
+						stride2, 0, 8);
 
 	interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 1, 8);
+						stride2, 0, 8);
 
 	stop_comp_timer();
 
@@ -1175,14 +1229,13 @@
 	uint32_t x, y;
 	VECTOR mv;
 	const VECTOR zeromv = {0,0};
-	const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp;
 	int i;
 
 	start_timer();
 	image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
-					dec->width, dec->height);
+					dec->width, dec->height, dec->bs_version);
 	image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
-					dec->width, dec->height);
+					dec->width, dec->height, dec->bs_version);
 	stop_edges_timer();
 
 	for (y = 0; y < dec->mb_height; y++) {
@@ -1266,18 +1319,19 @@
 
 			switch (mb->mode) {
 			case MODE_DIRECT:
-				get_b_motion_vector(bs, &mv, 1, zeromv);
+				get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y);
 
 			case MODE_DIRECT_NONE_MV:
 				for (i = 0; i < 4; i++) {
-					mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x);
-					mb->b_mvs[i].x = (int32_t) ((mv.x == 0)
-									? ((TRB - TRD) * last_mb->mvs[i].x) / TRD
-									: mb->mvs[i].x - last_mb->mvs[i].x);
-					mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y);
-					mb->b_mvs[i].y = (int32_t) ((mv.y == 0)
-									? ((TRB - TRD) * last_mb->mvs[i].y) / TRD
-									: mb->mvs[i].y - last_mb->mvs[i].y);
+					mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x;
+					mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y;
+					
+					mb->b_mvs[i].x = (mv.x)
+						?  mb->mvs[i].x - last_mb->mvs[i].x
+						: last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp;
+					mb->b_mvs[i].y = (mv.y)
+						? mb->mvs[i].y - last_mb->mvs[i].y
+						: last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp;
 				}
 
 				decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
@@ -1285,10 +1339,10 @@
 				break;
 
 			case MODE_INTERPOLATE:
-				get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv);
+				get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
 				dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] =	mb->mvs[0];
 
-				get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv);
+				get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
 				dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
 
 				decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
@@ -1296,14 +1350,14 @@
 				break;
 
 			case MODE_BACKWARD:
-				get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv);
+				get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
 				dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] =	mb->mvs[0];
 
 				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
 				break;
 
 			case MODE_FORWARD:
-				get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv);
+				get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
 				dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] =	mb->mvs[0];
 
 				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
@@ -1342,10 +1396,17 @@
 		stats->type = coding2type(coding_type);
 		stats->data.vop.time_base = (int)dec->time_base;
 		stats->data.vop.time_increment = 0;	/* XXX: todo */
+		stats->data.vop.qscale_stride = dec->mb_width;
+		stats->data.vop.qscale = dec->qscale;
+		if (stats->data.vop.qscale != NULL && mbs != NULL) {
+			int i;
+			for (i = 0; i < dec->mb_width*dec->mb_height; i++)
+				stats->data.vop.qscale[i] = mbs[i].quant;
+		} else 
+			stats->data.vop.qscale = NULL;
 	}
 }
 
-
 int
 decoder_decode(DECODER * dec,
 				xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
@@ -1354,7 +1415,7 @@
 	Bitstream bs;
 	uint32_t rounding;
 	uint32_t reduced_resolution;
-	uint32_t quant;
+	uint32_t quant = 2;
 	uint32_t fcode_forward;
 	uint32_t fcode_backward;
 	uint32_t intra_dc_threshold;
@@ -1442,6 +1503,11 @@
 		goto repeat;
 	}
 
+	if(dec->frames == 0 && coding_type != I_VOP) {
+		/* 1st frame is not an i-vop */
+		goto repeat;
+	}
+
 	dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0;	/* init pred vector to 0 */
 
 	/* packed_mode: special-N_VOP treament */
@@ -1503,7 +1569,7 @@
 
 		if (dec->low_delay) {
 			DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
-			dec->low_delay = 1;
+			dec->low_delay = 0;
 		}
 
 		if (dec->frames < 2) {
@@ -1526,7 +1592,9 @@
 		dec->frames++;
 	}
 
-	BitstreamByteAlign(&bs);
+#if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */
+	 BitstreamByteAlign(&bs);
+#endif
 
 	/* low_delay_default mode: repeat in packed_mode */
 	if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
@@ -1557,5 +1625,5 @@
 	emms();
 	stop_global_timer();
 
-	return BitstreamPos(&bs) / 8;	/* number of bytes consumed */
+	return (BitstreamPos(&bs) + 7) / 8;	/* number of bytes consumed */
 }