Skip to contentSkip to conversation

amidi.c.patch review request

PATCH Martin Tarenskeen 2 participants 2 messages Mar 20, 2016 – Mar 21, 2016

lore
Share & tools
lore
#0 lore raw

Hi,

I improved my own patch to amidi.c

As a result I can now succesfully fetch a SysEx dump from my Yamaha Reface DX, and it exits succesfully after the timeout that I specify using the -t option.

Problem: My synth sends continuously not only 0xFE (Active Sensing) but also 0xF8 (Timing Clock). 24 times per quarter note to be exactly. With a fast tempo like 250 BPM this gives a 0xF8 byte every 10 ms. Until now when I tried to fetch a SysEx dump from my synth using amidi the data were polluted with lots of 0xF8 bytes and the -t option would not work. A bug I would say.

The attached patch seems to solve this issue, but please comment and improve where needed. Please try with other synths also.

Can someone submit my patch to the official sources? I don't know how to. See attached.

Signature · 1 line
Martin
amidi.c renamed from amidi.c.orig +3 -3
@@ -559,7 +559,7 @@
 			int i, length;
 			unsigned short revents;
 
-			err = poll(pfds, npfds, 200);
+			err = poll(pfds, npfds, 5);
 			if (stop || (err < 0 && errno == EINTR))
 				break;
 			if (err < 0) {
@@ -567,7 +567,7 @@
 				break;
 			}
 			if (err == 0) {
-				time += 200;
+				time += 5;
 				if (timeout && time >= timeout)
 					break;
 				continue;
@@ -589,7 +589,7 @@
 			}
 			length = 0;
 			for (i = 0; i < err; ++i)
-				if (!ignore_active_sensing || buf[i] != 0xfe)
+				if (!ignore_active_sensing || buf[i] < 0xf8)
 					buf[length++] = buf[i];
 			if (length == 0)
 				continue;
#1 lore raw

amidi.c line 562

Quoted 12 lines · Martin Tarenskeen

My synth sends continuously not only 0xFE (Active Sensing) but also 0xF8 (Timing Clock). 24 times per quarter note to be exactly. With a fast tempo like 250 BPM this gives a 0xF8 byte every 10 ms. Until now when I tried to fetch a SysEx dump from my synth using amidi the data were polluted with lots of 0xF8 bytes and the -t option would not work. A bug I would say.

The attached patch seems to solve this issue, but please comment and improve where needed.

amidi.c renamed from amidi.c.orig +1 -1
-			err = poll(pfds, npfds, 200);
+			err = poll(pfds, npfds, 5);

This does not help when the clocks are sent faster.

To properly handle this, there needs to be a separate timer for the timeout. I'm working on this ...

Regards,
Clemens