some notes on mem-leaks...

libs/ardour/plugin_manager.cc

@@ -582,6 +582,11 @@ PluginManager::ladspa_discover (string path)
 	DEBUG_TRACE (DEBUG::PluginManager, string_compose ("LADSPA plugin found at %1\n", path));
 
 	for (uint32_t i = 0; ; ++i) {
+		/* if a ladspa plugin allocates memory here
+		 * it is never free()ed (or plugin-dependent only when unloading).
+		 * For some plugins memory allocated is incremental, we should
+		 * avoid re-scanning plugins and file bug reports.
+		 */
 		if ((descriptor = dfunc (i)) == 0) {
 			break;
 		}

libs/ardour/tempo.cc

@@ -284,6 +284,7 @@ TempoMap::TempoMap (framecnt_t fr)
 	start.beats = 1;
 	start.ticks = 0;
 
+	// these leak memory, well Metrics does
 	TempoSection *t = new TempoSection (start, _default_tempo.beats_per_minute(), _default_tempo.note_type());
 	MeterSection *m = new MeterSection (start, _default_meter.divisions_per_bar(), _default_meter.note_divisor());
 

libs/pbd/pbd/abstract_ui.cc

@@ -444,7 +444,7 @@ AbstractUI<RequestObject>::call_slot (InvalidationRecord* invalidation, const bo
 template<typename RequestObject> void*
 AbstractUI<RequestObject>::request_buffer_factory (uint32_t num_requests)
 {
-	RequestBuffer*  mcr = new RequestBuffer (num_requests);
+	RequestBuffer*  mcr = new RequestBuffer (num_requests); // leaks
 	per_thread_request_buffer.set (mcr);
 	return mcr;
 }