sub: do not clear subtitle list on seeking

This affects non-ASS text subtitles (those which go through libavcodec's
subtitle converter), which are muxed with video/audio. (Typically srt
subs in mkv.)

The problem is that seeking in the file can send a subtitle packet to
the decoder multiple times. These packets are interlaved with video,
and thus can't be all read when opening the file. Rather, subtitle
packets can essentially be randomly skipped or repeated (by seeking).
Until recently, this was solved by scanning the libass event list for
duplicates. Then our builtin srt-to-ass converter was removed, and
the problem was handled by fully clearing the subtitle list on each
seek.

This resulted in sub-seek not working properly for this type of file.
Since the subtitle list was cleared on seek, it was not possible to
do e.g. sub-seeks to subtitles before the current playback position.

Fix this by not clearing the list, and intead explicitly rejecting
duplicate packets. We use the packet file position was unique ID for
subtitles; this is confirmed working for most file formats (although
it is slightly risky - new demuxers may not necessarily set the file
position to something unique, or at all).

The list of seen packets is sorted, and the lookup uses binary search.
This is to avoid quadratic complexity when subtitles are added in
bulks, such as when opening a text subtitle file.

In some places, the code has to be adjusted to pass through the packet
file position correctly.

sub/dec_sub.c

@@ -273,6 +273,7 @@ static struct demux_packet *recode_packet(struct mp_log *log,
             .buffer = conv.start,
             .len = conv.len,
             .pts = in->pts,
+            .pos = in->pos,
             .duration = in->duration,
             .avpacket = in->avpacket, // questionable, but gives us sidedata
         };
@@ -305,13 +306,6 @@ static void add_sub_list(struct dec_sub *sub, struct packet_list *subs)
 
     for (int n = 0; n < subs->num_packets; n++)
         decode_chain_recode(sub, subs->packets[n]);
-
-    // Hack for broken FFmpeg packet format: make sd_ass keep the subtitle
-    // events on reset(), even if broken FFmpeg ASS packets were received
-    // (from sd_lavc_conv.c). Normally, these events are removed on seek/reset,
-    // but this is obviously unwanted in this case.
-    if (sd->driver->fix_events)
-        sd->driver->fix_events(sd);
 }
 
 static void add_packet(struct packet_list *subs, struct demux_packet *pkt)
@@ -446,7 +440,7 @@ struct sd_conv_buffer {
 };
 
 void sd_conv_add_packet(struct sd *sd, void *data, int data_len, double pts,
-                        double duration)
+                        double duration, int64_t pos)
 {
     if (!sd->sd_conv_buffer)
         sd->sd_conv_buffer = talloc_zero(sd, struct sd_conv_buffer);
@@ -462,6 +456,7 @@ void sd_conv_add_packet(struct sd *sd, void *data, int data_len, double pts,
         .len = data_len,
         .pts = pts,
         .duration = duration,
+        .pos = pos,
     };
     memcpy(pkt->buffer, data, data_len);
     pkt->buffer[data_len] = 0;

sub/sd.h

@@ -58,7 +58,6 @@ struct sd_functions {
     void (*uninit)(struct sd *sd);
 
     bool (*accepts_packet)(struct sd *sd); // implicit default if NULL: true
-    void (*fix_events)(struct sd *sd);
     int (*control)(struct sd *sd, enum sd_ctrl cmd, void *arg);
 
     // decoder
@@ -71,7 +70,7 @@ struct sd_functions {
 };
 
 void sd_conv_add_packet(struct sd *sd, void *data, int data_len, double pts,
-                        double duration);
+                        double duration, int64_t pos);
 struct demux_packet *sd_conv_def_get_converted(struct sd *sd);
 void sd_conv_def_reset(struct sd *sd);
 void sd_conv_def_uninit(struct sd *sd);

sub/sd_ass.c

@@ -42,11 +42,12 @@ struct sd_ass_priv {
     bool is_converted;
     bool on_top;
     struct sub_bitmap *parts;
-    bool flush_on_seek;
     char last_text[500];
     struct mp_image_params video_params;
     struct mp_image_params last_params;
     double sub_speed;
+    int64_t *seen_packets;
+    int num_seen_packets;
 };
 
 static void mangle_colors(struct sd *sd, struct sub_bitmaps *parts);
@@ -130,19 +131,42 @@ static int init(struct sd *sd)
     return 0;
 }
 
+// Test if the packet with the given file position (used as unique ID) was
+// already consumed. Return false if the packet is new (and add it to the
+// internal list), and return true if it was already seen.
+static bool check_packet_seen(struct sd *sd, int64_t pos)
+{
+    struct sd_ass_priv *priv = sd->priv;
+    int a = 0;
+    int b = priv->num_seen_packets;
+    while (a < b) {
+        int mid = a + (b - a) / 2;
+        int64_t val = priv->seen_packets[mid];
+        if (pos == val)
+            return true;
+        if (pos > val) {
+            a = mid + 1;
+        } else {
+            b = mid;
+        }
+    }
+    MP_TARRAY_INSERT_AT(priv, priv->seen_packets, priv->num_seen_packets, a, pos);
+    return false;
+}
+
 static void decode(struct sd *sd, struct demux_packet *packet)
 {
     struct sd_ass_priv *ctx = sd->priv;
     ASS_Track *track = ctx->ass_track;
     if (strcmp(sd->codec, "ass") == 0) {
-        long long ipts = lrint(packet->pts * 1000);
-        long long iduration = lrint(packet->duration * 1000);
-        ass_process_chunk(track, packet->buffer, packet->len, ipts, iduration);
-        return;
+        // Note that for this packet format, libass has an internal mechanism
+        // for discarding duplicate (already seen) packets.
+        ass_process_chunk(track, packet->buffer, packet->len,
+                          lrint(packet->pts * 1000),
+                          lrint(packet->duration * 1000));
     } else {
-        // "ssa" codec ID
-        ctx->flush_on_seek = true;
-        ass_process_data(track, packet->buffer, packet->len);
+        if (!check_packet_seen(sd, packet->pos))
+            ass_process_data(track, packet->buffer, packet->len);
     }
 }
 
@@ -476,18 +500,11 @@ static void fill_plaintext(struct sd *sd, double pts)
         track->styles[track->default_style].Alignment = ctx->on_top ? 6 : 2;
 }
 
-static void fix_events(struct sd *sd)
-{
-    struct sd_ass_priv *ctx = sd->priv;
-    ctx->flush_on_seek = false;
-}
-
 static void reset(struct sd *sd)
 {
     struct sd_ass_priv *ctx = sd->priv;
-    if (ctx->flush_on_seek || sd->opts->sub_clear_on_seek)
+    if (sd->opts->sub_clear_on_seek)
         ass_flush_events(ctx->ass_track);
-    ctx->flush_on_seek = false;
 }
 
 static void uninit(struct sd *sd)
@@ -530,7 +547,6 @@ const struct sd_functions sd_ass = {
     .decode = decode,
     .get_bitmaps = get_bitmaps,
     .get_text = get_text,
-    .fix_events = fix_events,
     .control = control,
     .reset = reset,
     .uninit = uninit,

sub/sd_lavc_conv.c

@@ -250,7 +250,7 @@ static void decode(struct sd *sd, struct demux_packet *packet)
             // ASS packet format, in which case pts/duration might be ignored
             // at a later point.
             sd_conv_add_packet(sd, ass_line, strlen(ass_line),
-                               packet->pts, packet->duration);
+                               packet->pts, packet->duration, packet->pos + i);
         }
     }