The start of support for situations like this: the user appends an AVI to an AVI and VobSubs to VobSubs. In such a case simply using the previous file's max timecode won't work for the subtitle-only files. Therefore mkvmerge has to look for an "associated" file whose max timecode it has to use. This has not been finished yet.

src/merge/cluster_helper.cpp

@@ -138,7 +138,7 @@ cluster_helper_c::add_packet(packet_t *packet) {
   if (splitting && (file_num <= split_max_num_files) &&
       (packet->bref == -1) &&
       ((packet->source->get_track_type() == track_video) ||
-       !video_track_present)) {
+       (video_packetizer == NULL))) {
     split = false;
     c = clusters[num_clusters - 1];
     if (first_timecode_in_file == -1)
@@ -536,7 +536,8 @@ cluster_helper_c::render_cluster(ch_contents_t *clstr) {
           // ... or if this is an audio track, there is no video track and the
           // last cue entry was created more than 2s ago.
           ((source->get_cue_creation() == CUES_SPARSE) &&
-           (source->get_track_type() == track_audio) && !video_track_present &&
+           (source->get_track_type() == track_audio) &&
+           (video_packetizer == NULL) &&
            ((source->get_last_cue_timecode() < 0) ||
             ((pack->assigned_timecode - source->get_last_cue_timecode()) >=
              2000000000)))) {

src/merge/mkvmerge.cpp

@@ -323,8 +323,6 @@ identify(const string &filename) {
             file.name);
 
   file.appending = false;
-  file.fp = NULL;
-  file.status = EMOREDATA;
   file.pack = NULL;
   file.ti = new track_info_c(ti);
 
@@ -1725,11 +1723,10 @@ parse_args(vector<string> &args) {
                   "your file type is supported but not recognized "
                   "properly.\n"), file.name);
 
-      file.fp = NULL;
       if (file.type != TYPECHAPTERS) {
-        file.status = EMOREDATA;
         file.pack = NULL;
         file.ti = ti;
+        file.done = false;
 
         files.push_back(file);
       } else {

src/merge/output_control.cpp

@@ -113,7 +113,7 @@ int64_t file_sizes = 0;
 int max_blocks_per_cluster = 65535;
 int64_t max_ns_per_cluster = 2000000000;
 bool write_cues = true, cue_writing_requested = false;
-bool video_track_present = false;
+generic_packetizer_c *video_packetizer = NULL;
 bool write_meta_seek_for_clusters = true;
 bool no_lacing = false, no_linking = true;
 int64_t split_after = -1;
@@ -464,7 +464,7 @@ render_headers(mm_io_c *rout) {
 
     kax_infos = &GetChild<KaxInfo>(*kax_segment);
 
-    if (!video_track_present ||
+    if ((video_packetizer == NULL) ||
         (timecode_scale_mode == timecode_scale_mode_auto))
       kax_duration = new KaxMyDuration(EbmlFloat::FLOAT_64);
     else
@@ -1511,9 +1511,36 @@ append_track(packetizer_t &ptzr,
   // If we're dealing with a subtitle track or if the appending file contains
   // chapters then we have to suck the previous file dry. See below for the
   // reason (short version: we need all max_timecode_seen values).
-  if (((*gptzr)->get_track_type() == track_subtitle) ||
-      (src_file.reader->chapters != NULL))
+  if ((((*gptzr)->get_track_type() == track_subtitle) ||
+       (src_file.reader->chapters != NULL)) &&
+      !dst_file.done) {
+    vector<deferred_connection_t>::const_iterator def_con;
+    
     dst_file.reader->read_all();
+    dst_file.done = true;
+    foreach(def_con, dst_file.deferred_connections)
+      append_track(*def_con->ptzr, def_con->amap);
+    dst_file.deferred_connections.clear();
+  }
+
+  if (((*gptzr)->get_track_type() == track_subtitle) &&
+      (dst_file.reader->num_video_tracks == 0) &&
+      (video_packetizer != NULL) && !ptzr.deferred) {
+    vector<filelist_t>::iterator file;
+
+    foreach(file, files) {
+      if (mxfind(video_packetizer, file->reader->reader_packetizers) !=
+          file->reader->reader_packetizers.end()) {
+        deferred_connection_t new_def_con;
+
+        ptzr.deferred = true;
+        new_def_con.amap = amap;
+        new_def_con.ptzr = &ptzr;
+        file->deferred_connections.push_back(new_def_con);
+        return;
+      }
+    }
+  }
 
   mxinfo("Appending track %lld from file no. %lld ('%s') to track %lld from "
          "file no. %lld ('%s').\n",
@@ -1550,8 +1577,9 @@ append_track(packetizer_t &ptzr,
   // But then again I don't expect that people will try to concatenate such
   // files if they've been split before.
   timecode_adjustment = dst_file.reader->max_timecode_seen;
-  if ((ptzr.packetizer->get_track_type() == track_subtitle) ||
-      (src_file.reader->chapters != NULL)) {
+  if (((ptzr.packetizer->get_track_type() == track_subtitle) ||
+       (src_file.reader->chapters != NULL)) &&
+      !ptzr.deferred) {
     vector<append_spec_t>::const_iterator cmp_amap;
 
     if (src_file.reader->ptzr_first_packet == NULL)
@@ -1587,6 +1615,8 @@ append_track(packetizer_t &ptzr,
     delete src_file.reader->chapters;
     src_file.reader->chapters = NULL;
   }
+
+  ptzr.deferred = false;
 }
 
 /** \brief Decide if packetizers have to be appended
@@ -1605,6 +1635,8 @@ append_tracks_maybe() {
 
   appended_a_track = false;
   foreach(ptzr, packetizers) {
+    if (ptzr->deferred)
+      continue;
     if (!files[ptzr->orig_file].appended_to)
       continue;
     if ((ptzr->status == EMOREDATA) || (ptzr->status == EHOLDING))

src/merge/output_control.h

@@ -38,26 +38,40 @@ class generic_packetizer_c;
 class generic_reader_c;
 
 typedef struct {
-  char *name;
-  mm_io_c *fp;
-
-  int type, status;
-
-  packet_t *pack;
-
-  generic_reader_c *reader;
-
-  track_info_c *ti;
-  bool appending, appended_to;
-} filelist_t;
+  int64_t src_file_id;
+  int64_t src_track_id;
+  int64_t dst_file_id;
+  int64_t dst_track_id;
+} append_spec_t;
 
 typedef struct packetizer_t {
   int status;
   packet_t *pack;
   generic_packetizer_c *packetizer, *orig_packetizer;
   int64_t file, orig_file;
+  bool deferred;
 } packetizer_t;
 
+typedef struct {
+  append_spec_t amap;
+  packetizer_t *ptzr;
+} deferred_connection_t;
+
+typedef struct {
+  char *name;
+
+  int type;
+
+  packet_t *pack;
+
+  generic_reader_c *reader;
+
+  track_info_c *ti;
+  bool appending, appended_to, done;
+
+  vector<deferred_connection_t> deferred_connections;
+} filelist_t;
+
 typedef struct {
   char *name, *mime_type, *description;
   int64_t size;
@@ -69,13 +83,6 @@ typedef struct {
   int64_t track_id;
 } track_order_t;
 
-typedef struct {
-  int64_t src_file_id;
-  int64_t src_track_id;
-  int64_t dst_file_id;
-  int64_t dst_track_id;
-} append_spec_t;
-
 enum timecode_scale_mode_t {
   timecode_scale_mode_normal = 0,
   timecode_scale_mode_fixed,
@@ -116,7 +123,8 @@ extern string default_language;
 
 extern float video_fps;
 
-extern bool write_cues, cue_writing_requested, video_track_present;
+extern generic_packetizer_c *video_packetizer;
+extern bool write_cues, cue_writing_requested;
 extern bool no_lacing, no_linking, use_durations;
 
 extern bool identifying, identify_verbose;

src/merge/pr_generic.cpp

@@ -324,6 +324,14 @@ generic_packetizer_c::set_track_type(int type) {
 
   if ((type == track_audio) && (ti->cues == CUES_UNSPECIFIED))
     ti->cues = CUES_SPARSE;
+  if (type == track_audio)
+    reader->num_audio_tracks++;
+  else if (type == track_video) {
+    reader->num_video_tracks++;
+    if (video_packetizer == NULL)
+      video_packetizer = this;
+  } else
+    reader->num_subtitle_tracks++;
 }
 
 void
@@ -1023,6 +1031,9 @@ generic_reader_c::generic_reader_c(track_info_c *nti) {
   appending = false;
   chapters = NULL;
   ptzr_first_packet = NULL;
+  num_video_tracks = 0;
+  num_audio_tracks = 0;
+  num_subtitle_tracks = 0;
 
   add_all_requested_track_ids2(atracks);
   add_all_requested_track_ids2(vtracks);

src/merge/pr_generic.h

@@ -253,6 +253,7 @@ public:
   int64_t max_timecode_seen;
   KaxChapters *chapters;
   bool appending;
+  int num_video_tracks, num_audio_tracks, num_subtitle_tracks;
 
 public:
   generic_reader_c(track_info_c *nti);

src/output/p_video.cpp

@@ -50,7 +50,6 @@ video_packetizer_c::video_packetizer_c(generic_reader_c *nreader,
   ref_timecode = -1;
   if (get_cue_creation() == CUES_UNSPECIFIED)
     set_cue_creation(CUES_IFRAMES);
-  video_track_present = true;
   duration_shift = 0;
   bref_frame.type = '?';
   fref_frame.type = '?';