ffmpeg: add support to read/write mp4 'titl' track tag

3 files changed, 297 insertions, 11 deletions

diff --git a/contrib/ffmpeg/A05-mov-read-name-track-tag-written-by-movenc.patch b/contrib/ffmpeg/A05-mov-read-name-track-tag-written-by-movenc.patch
index 514aa79d0..67e4d2868 100644
--- a/contrib/ffmpeg/A05-mov-read-name-track-tag-written-by-movenc.patch
+++ b/contrib/ffmpeg/A05-mov-read-name-track-tag-written-by-movenc.patch
@@ -1,14 +1,14 @@
-From b6ce7b6a4ec0429069f05e63edce907c0f80aaea Mon Sep 17 00:00:00 2001
+From 375d09561cb88a7002a833614dbd399446479cd1 Mon Sep 17 00:00:00 2001
 From: John Stebbins <[email protected]>
 Date: Wed, 31 Jul 2019 15:27:40 -0700
-Subject: [PATCH] mov: read 'name' track tag written by movenc
+Subject: [PATCH 1/3] mov: read 'name' track tag written by movenc
 
 ---
  libavformat/mov.c | 26 +++++++++++++++++---------
  1 file changed, 17 insertions(+), 9 deletions(-)
 
 diff --git a/libavformat/mov.c b/libavformat/mov.c
-index 24de5429d1..fce69129d9 100644
+index 24de5429d1..1f933cccfd 100644
 --- a/libavformat/mov.c
 +++ b/libavformat/mov.c
 @@ -306,6 +306,7 @@ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
@@ -19,14 +19,14 @@ index 24de5429d1..fce69129d9 100644
  
      switch (atom.type) {
      case MKTAG( '@','P','R','M'): key = "premiere_version"; raw = 1; break;
-@@ -384,6 +385,7 @@ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
-     case MKTAG(0xa9,'m','a','k'): key = "make";      break;
-     case MKTAG(0xa9,'m','o','d'): key = "model";     break;
-     case MKTAG(0xa9,'n','a','m'): key = "title";     break;
-+    case MKTAG( 'n','a','m','e'): key = "title";     break;
-     case MKTAG(0xa9,'o','p','e'): key = "original_artist"; break;
-     case MKTAG(0xa9,'p','r','d'): key = "producer";  break;
-     case MKTAG(0xa9,'p','r','f'): key = "performers"; break;
+@@ -338,6 +339,7 @@ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
+         return mov_metadata_loci(c, pb, atom.size);
+     case MKTAG( 'm','a','n','u'): key = "make"; break;
+     case MKTAG( 'm','o','d','l'): key = "model"; break;
++    case MKTAG( 'n','a','m','e'): key = "title"; raw = 1; break;
+     case MKTAG( 'p','c','s','t'): key = "podcast";
+         parse = mov_metadata_int8_no_padding; break;
+     case MKTAG( 'p','g','a','p'): key = "gapless_playback";
 @@ -514,17 +516,23 @@ retry:
              }
              str[str_size] = 0;
diff --git a/contrib/ffmpeg/A06-movenc-write-3gpp-track-titl-tag.patch b/contrib/ffmpeg/A06-movenc-write-3gpp-track-titl-tag.patch
new file mode 100644
index 000000000..5c40e66a0
--- /dev/null
+++ b/contrib/ffmpeg/A06-movenc-write-3gpp-track-titl-tag.patch
@@ -0,0 +1,126 @@
+From e9dde141a5fc27df87eb4b3454e34d725b211708 Mon Sep 17 00:00:00 2001
+From: John Stebbins <[email protected]>
+Date: Sun, 4 Aug 2019 10:52:30 -0700
+Subject: [PATCH 2/3] movenc: write 3gpp track 'titl' tag
+
+Apple software used to use 'name' raw tag for track titles.  When they
+rewrote everything with AVFoundation, they switched to using 3gpp 'titl'
+tag for track titles and 'name' no longer works.
+---
+ libavformat/movenc.c | 78 +++++++++++++++++++++++---------------------
+ 1 file changed, 40 insertions(+), 38 deletions(-)
+
+diff --git a/libavformat/movenc.c b/libavformat/movenc.c
+index a96139077b..b1b6ef0e4e 100644
+--- a/libavformat/movenc.c
++++ b/libavformat/movenc.c
+@@ -3077,6 +3077,35 @@ static int mov_write_udta_sdp(AVIOContext *pb, MOVTrack *track)
+     return len + 24;
+ }
+ 
++static uint16_t language_code(const char *str)
++{
++    return (((str[0] - 0x60) & 0x1F) << 10) +
++           (((str[1] - 0x60) & 0x1F) <<  5) +
++           (( str[2] - 0x60) & 0x1F);
++}
++
++static int mov_write_3gp_udta_tag(AVIOContext *pb, AVDictionary *metadata,
++                                  const char *tag, const char *str)
++{
++    int64_t pos = avio_tell(pb);
++    AVDictionaryEntry *t = av_dict_get(metadata, str, NULL, 0);
++    if (!t || !utf8len(t->value))
++        return 0;
++    avio_wb32(pb, 0);   /* size */
++    ffio_wfourcc(pb, tag); /* type */
++    avio_wb32(pb, 0);   /* version + flags */
++    if (!strcmp(tag, "yrrc"))
++        avio_wb16(pb, atoi(t->value));
++    else {
++        avio_wb16(pb, language_code("eng")); /* language */
++        avio_write(pb, t->value, strlen(t->value) + 1); /* UTF8 string value */
++        if (!strcmp(tag, "albm") &&
++            (t = av_dict_get(metadata, "track", NULL, 0)))
++            avio_w8(pb, atoi(t->value));
++    }
++    return update_size(pb, pos);
++}
++
+ static int mov_write_track_metadata(AVIOContext *pb, AVStream *st,
+                                     const char *tag, const char *str)
+ {
+@@ -3105,8 +3134,10 @@ static int mov_write_track_udta_tag(AVIOContext *pb, MOVMuxContext *mov,
+     if (ret < 0)
+         return ret;
+ 
+-    if (mov->mode & (MODE_MP4|MODE_MOV))
++    if (mov->mode & (MODE_MP4|MODE_MOV)) {
+         mov_write_track_metadata(pb_buf, st, "name", "title");
++        mov_write_3gp_udta_tag(pb_buf, st->metadata, "titl", "title");
++    }
+ 
+     if ((size = avio_close_dyn_buf(pb_buf, &buf)) > 0) {
+         avio_wb32(pb, size + 8);
+@@ -3680,35 +3711,6 @@ static int ascii_to_wc(AVIOContext *pb, const uint8_t *b)
+     return 0;
+ }
+ 
+-static uint16_t language_code(const char *str)
+-{
+-    return (((str[0] - 0x60) & 0x1F) << 10) +
+-           (((str[1] - 0x60) & 0x1F) <<  5) +
+-           (( str[2] - 0x60) & 0x1F);
+-}
+-
+-static int mov_write_3gp_udta_tag(AVIOContext *pb, AVFormatContext *s,
+-                                  const char *tag, const char *str)
+-{
+-    int64_t pos = avio_tell(pb);
+-    AVDictionaryEntry *t = av_dict_get(s->metadata, str, NULL, 0);
+-    if (!t || !utf8len(t->value))
+-        return 0;
+-    avio_wb32(pb, 0);   /* size */
+-    ffio_wfourcc(pb, tag); /* type */
+-    avio_wb32(pb, 0);   /* version + flags */
+-    if (!strcmp(tag, "yrrc"))
+-        avio_wb16(pb, atoi(t->value));
+-    else {
+-        avio_wb16(pb, language_code("eng")); /* language */
+-        avio_write(pb, t->value, strlen(t->value) + 1); /* UTF8 string value */
+-        if (!strcmp(tag, "albm") &&
+-            (t = av_dict_get(s->metadata, "track", NULL, 0)))
+-            avio_w8(pb, atoi(t->value));
+-    }
+-    return update_size(pb, pos);
+-}
+-
+ static int mov_write_chpl_tag(AVIOContext *pb, AVFormatContext *s)
+ {
+     int64_t pos = avio_tell(pb);
+@@ -3747,14 +3749,14 @@ static int mov_write_udta_tag(AVIOContext *pb, MOVMuxContext *mov,
+         return ret;
+ 
+     if (mov->mode & MODE_3GP) {
+-        mov_write_3gp_udta_tag(pb_buf, s, "perf", "artist");
+-        mov_write_3gp_udta_tag(pb_buf, s, "titl", "title");
+-        mov_write_3gp_udta_tag(pb_buf, s, "auth", "author");
+-        mov_write_3gp_udta_tag(pb_buf, s, "gnre", "genre");
+-        mov_write_3gp_udta_tag(pb_buf, s, "dscp", "comment");
+-        mov_write_3gp_udta_tag(pb_buf, s, "albm", "album");
+-        mov_write_3gp_udta_tag(pb_buf, s, "cprt", "copyright");
+-        mov_write_3gp_udta_tag(pb_buf, s, "yrrc", "date");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "perf", "artist");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "titl", "title");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "auth", "author");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "gnre", "genre");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "dscp", "comment");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "albm", "album");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "cprt", "copyright");
++        mov_write_3gp_udta_tag(pb_buf, s->metadata, "yrrc", "date");
+         mov_write_loci_tag(s, pb_buf);
+     } else if (mov->mode == MODE_MOV && !(mov->flags & FF_MOV_FLAG_USE_MDTA)) { // the title field breaks gtkpod with mp4 and my suspicion is that stuff is not valid in mp4
+         mov_write_string_metadata(s, pb_buf, "\251ART", "artist",      0);
+-- 
+2.21.0
+
diff --git a/contrib/ffmpeg/A07-mov-read-3gpp-udta-tags.patch b/contrib/ffmpeg/A07-mov-read-3gpp-udta-tags.patch
new file mode 100644
index 000000000..6368c7180
--- /dev/null
+++ b/contrib/ffmpeg/A07-mov-read-3gpp-udta-tags.patch
@@ -0,0 +1,160 @@
+From 1b02b0afe201cd14a97d83dc08716bf9cc188cb0 Mon Sep 17 00:00:00 2001
+From: John Stebbins <[email protected]>
+Date: Sun, 4 Aug 2019 10:59:38 -0700
+Subject: [PATCH 3/3] mov: read 3gpp udta tags
+
+---
+ libavformat/mov.c | 110 ++++++++++++++++++++++++++++++++++++++++++++--
+ 1 file changed, 107 insertions(+), 3 deletions(-)
+
+diff --git a/libavformat/mov.c b/libavformat/mov.c
+index 1f933cccfd..0c365cde3e 100644
+--- a/libavformat/mov.c
++++ b/libavformat/mov.c
+@@ -50,6 +50,7 @@
+ #include "libavcodec/flac.h"
+ #include "libavcodec/mpegaudiodecheader.h"
+ #include "avformat.h"
++#include "avlanguage.h"
+ #include "internal.h"
+ #include "avio_internal.h"
+ #include "riff.h"
+@@ -295,6 +296,82 @@ static int mov_metadata_hmmt(MOVContext *c, AVIOContext *pb, unsigned len)
+     return 0;
+ }
+ 
++static int mov_read_3gp_udta_tag(MOVContext *c, AVIOContext *pb, MOVAtom atom)
++{
++    const char *key;
++    AVDictionary *metadata;
++    uint16_t langcode = 0;
++    char key2[32], language[4] = {0};
++    uint32_t str_size, version;
++    char *str;
++
++    if (atom.size < 6)
++        return AVERROR_INVALIDDATA;
++
++    switch (atom.type) {
++    case MKTAG( 'a','l','b','m'): key = "album"; break;
++    case MKTAG( 'a','u','t','h'): key = "author"; break;
++    case MKTAG( 'c','p','r','t'): key = "copyright"; break;
++    case MKTAG( 'd','s','c','p'): key = "comment"; break;
++    case MKTAG( 'g','n','r','e'): key = "genre"; break;
++    case MKTAG( 'p','e','r','f'): key = "artist"; break;
++    case MKTAG( 't','i','t','l'): key = "title"; break;
++    case MKTAG( 'y','r','r','c'): key = "date"; break;
++    default: return 0;
++    }
++
++    version = avio_rb32(pb); // version + flags
++    if (version != 0)
++        return AVERROR_INVALIDDATA;
++
++    if (MKTAG( 'y','r','r','c') == atom.type) {
++        int year;
++        year = avio_rb16(pb);
++        str = av_asprintf("%d", year);
++        if (!str)
++            return AVERROR(ENOMEM);
++    } else {
++        int ret;
++        const char *tmp;
++        langcode = avio_rb16(pb);
++        ff_mov_lang_to_iso639(langcode, language);
++        tmp = ff_convert_lang_to(language, AV_LANG_ISO639_2_BIBL);
++        if (!tmp)
++            return AVERROR_INVALIDDATA;
++
++        str_size = atom.size - 6;
++        if (str_size <= 0 || str_size >= INT_MAX/2)
++            return AVERROR_INVALIDDATA;
++
++        str = av_mallocz(str_size + 1);
++        if (!str)
++            return AVERROR(ENOMEM);
++
++        ret = ffio_read_size(pb, str, str_size);
++        if (ret < 0) {
++            av_free(str);
++            return ret;
++        }
++        str[str_size] = 0;
++    }
++
++    if (c->trak_index < 0) {
++        metadata = c->fc->metadata;
++        c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
++    }
++    else {
++        metadata = c->fc->streams[c->trak_index]->metadata;
++    }
++    av_dict_set(&metadata, key, str, 0);
++    if (*language && strcmp(language, "und")) {
++        snprintf(key2, sizeof(key2), "%s-%s", key, language);
++        av_dict_set(&metadata, key2, str, 0);
++    }
++
++    av_freep(&str);
++    return 0;
++}
++
+ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
+ {
+     char tmp_key[5];
+@@ -320,15 +397,33 @@ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
+     case MKTAG( 'c','a','t','g'): key = "category"; break;
+     case MKTAG( 'c','p','i','l'): key = "compilation";
+         parse = mov_metadata_int8_no_padding; break;
+-    case MKTAG( 'c','p','r','t'): key = "copyright"; break;
++    case MKTAG( 'c','p','r','t'):
++        key = "copyright";
++        if (!c->itunes_metadata) {
++            int64_t pos = avio_tell(pb);
++            int ret = mov_read_3gp_udta_tag(c, pb, atom);
++            if (ret != AVERROR_INVALIDDATA)
++                return ret;
++            avio_seek(pb, pos, SEEK_SET);
++        }
++        break;
+     case MKTAG( 'd','e','s','c'): key = "description"; break;
+     case MKTAG( 'd','i','s','k'): key = "disc";
+         parse = mov_metadata_track_or_disc_number; break;
+     case MKTAG( 'e','g','i','d'): key = "episode_uid";
+         parse = mov_metadata_int8_no_padding; break;
+     case MKTAG( 'F','I','R','M'): key = "firmware"; raw = 1; break;
+-    case MKTAG( 'g','n','r','e'): key = "genre";
+-        parse = mov_metadata_gnre; break;
++    case MKTAG( 'g','n','r','e'):
++        key = "genre";
++        parse = mov_metadata_gnre;
++        if (!c->itunes_metadata) {
++            int64_t pos = avio_tell(pb);
++            int ret = mov_read_3gp_udta_tag(c, pb, atom);
++            if (ret != AVERROR_INVALIDDATA)
++                return ret;
++            avio_seek(pb, pos, SEEK_SET);
++        }
++        break;
+     case MKTAG( 'h','d','v','d'): key = "hd_video";
+         parse = mov_metadata_int8_no_padding; break;
+     case MKTAG( 'H','M','M','T'):
+@@ -399,6 +494,15 @@ static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
+     case MKTAG(0xa9,'w','r','n'): key = "warning";   break;
+     case MKTAG(0xa9,'w','r','t'): key = "composer";  break;
+     case MKTAG(0xa9,'x','y','z'): key = "location";  break;
++    case MKTAG( 'a','l','b','m'):
++    case MKTAG( 'a','u','t','h'):
++    case MKTAG( 'd','s','c','p'):
++    case MKTAG( 'p','e','r','f'):
++    case MKTAG( 't','i','t','l'):
++    case MKTAG( 'y','r','r','c'):
++        if (!c->itunes_metadata) {
++            return mov_read_3gp_udta_tag(c, pb, atom);
++        }
+     }
+ retry:
+     if (c->itunes_metadata && atom.size > 8) {
+-- 
+2.21.0
+