+template<typename T>
+size_t
+MidiRingBuffer<T>::skip_to(framepos_t start)
+{
+ if (this->read_space() == 0) {
+ return 0;
+ }
+
+ T ev_time;
+ uint32_t ev_size;
+ size_t count = 0;
+ const size_t prefix_size = sizeof(T) + sizeof(Evoral::EventType) + sizeof(uint32_t);
+
+ while (this->read_space() >= prefix_size) {
+
+ uint8_t peekbuf[prefix_size];
+ this->peek (peekbuf, prefix_size);
+
+ ev_time = *(reinterpret_cast<T*>((uintptr_t)peekbuf));
+ ev_size = *(reinterpret_cast<uint32_t*>((uintptr_t)(peekbuf + sizeof(T) + sizeof (Evoral::EventType))));
+
+ if (ev_time >= start) {
+ return count;
+ }
+
+ if (this->read_space() < ev_size) {
+ continue;
+ }
+
+ this->increment_read_ptr (prefix_size);
+
+ uint8_t status;
+ bool r = this->peek (&status, sizeof(uint8_t));
+ assert (r); // If this failed, buffer is corrupt, all hope is lost
+
+ ++count;
+
+ /* TODO investigate and think:
+ *
+ * Does it makes sense to keep track of notes
+ * that are skipped (because they're either too late
+ * (underrun) or never used (read-ahead, loop) ?
+ *
+ * skip_to() is called on the rinbuffer between
+ * disk and process. it seems wrong to track them
+ * (a potential synth never sees skipped notes, either)
+ * but there may be more to this.
+ */
+
+ if (ev_size >= 8) {
+ this->increment_read_ptr (ev_size);
+ } else {
+ // we only track note on/off, 8 bytes are plenty.
+ uint8_t write_loc[8];
+ bool success = read_contents (ev_size, write_loc);
+ if (success) {
+ _tracker.track(write_loc);
+ }
+ }
+ }
+ return count;
+}
+
+
+