Skip to contentSkip to conversation

amidi: add delay option

PATCH Superseded Felipe F. Tonello 3 participants 6 messages Aug 11, 2016 – Aug 12, 2016

v1 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 amidi: add delay option

This patch adds a new options to amidi tool to add a delay (in milliseconds) to each MIDI message.

This is very useful when sending firmware updates to a remote device via SysEx or any other use that requires this delay in between messages.

---
 amidi/amidi.1 | 14 ++++++++++++
 amidi/amidi.c | 72 ++++++++++++++++++++++++++++++++++++++++++++++++++++++-----
 2 files changed, 81 insertions(+), 5 deletions(-)
amidi/amidi.1 +14 -0
index 1b4cfb1ea956..40402c17e374 100644
@@ -111,6 +111,12 @@ to stop receiving data.
 Does not ignore Active Sensing bytes (FEh) when saving or printing
 received MIDI commands.
 
+.TP
+.I \-D, \-\-delay=mseconds
+Adds a delay in between each MIDI message sent to a device. It is
+useful when sending firmware updates via SysEx messages to a remote
+device.
+
 .SH EXAMPLES
 
 .TP
@@ -121,6 +127,14 @@ to port
 .I hw:0.
 
 .TP
+.B amidi \-p hw:1,0,0 -s firmware.syx \-D 100
+will send the MIDI commands in
+.I firmware.syx
+to port
+.I hw:1,0,0
+with 100 milliseconds delay in between MIDI messages.
+
+.TP
 .B amidi \-S 'F0 43 10 4C 00 00 7E 00 F7'
 sends an XG Reset to the default port.
 
amidi/amidi.c +67 -5

114 lines not on this page · raw message

Signature · 1 line
2.9.2
#1 lore raw

amidi/amidi.c static void parse_data(void) · line 330

Quoted 11 lines · Felipe F. Tonello

This patch adds a new options to amidi tool to add a delay (in milliseconds) to each MIDI message.

This is very useful when sending firmware updates to a remote device via SysEx or any other use that requires this delay in between messages.

amidi/amidi.c +5 -0
+	case 0xf4:
+	case 0xf5:
+		return 0; /* ignore */
+	case 0xf7:
+		return 0; /* ignore */

Silent data loss is bad.

amidi/amidi.c static void parse_data(void) · line 335

Quoted 2 lines
amidi/amidi.c +2 -0
+	case 0x00 ... 0x7f:
+		return rstate_len ?: rstate_len - 1;

rstate_len - 1 == -1

amidi/amidi.c int main(int argc, char *argv[]) · line 594

Quoted 6 lines
amidi/amidi.c +6 -0
+				int len = message_length(*data);
+
+				/* check for variable length */
+				if (len == -1) {
+					int i = 0;
+					while (data[i++] != 0xf7);

Possible buffer overflow.

amidi/amidi.c int main(int argc, char *argv[]) · line 599

Quoted 2 lines
amidi/amidi.c +2 -0
+				if (len > 0 &&
+				    (err = snd_rawmidi_write(output, data, len)) < 0) {

Same here.

The most robust way to find message boundaries probably would be a state machine. However, there is actually no requirement to find all message boundaries; the _actual_ requirement is to insert pauses between SysEx messages. So don't bother to try to parse everything; just search for F0 or F7.

Regards,
Clemens

#2 lore raw

Hi Clemens,

Quoted 7 lines · Clemens Ladisch
Quoted 5 lines · Felipe F. Tonello

This patch adds a new options to amidi tool to add a delay (in milliseconds) to each MIDI message.

This is very useful when sending firmware updates to a remote device via SysEx or any other use that requires this delay in between messages.

Quoted 7 lines

amidi/amidi.c static void parse_data(void) · line 330

Quoted 5 lines
amidi/amidi.c +5 -0
+	case 0xf4:
+	case 0xf5:
+		return 0; /* ignore */
+	case 0xf7:
+		return 0; /* ignore */

Silent data loss is bad.

What do you mean by that?

Quoted 5 lines

amidi/amidi.c static void parse_data(void) · line 335

Quoted 3 lines
amidi/amidi.c +2 -0
+	case 0x00 ... 0x7f:
+		return rstate_len ?: rstate_len - 1;

rstate_len - 1 == -1

?

That's why there is this condition, if rstate_len is zero, than it returns 0.

Quoted 9 lines

amidi/amidi.c int main(int argc, char *argv[]) · line 594

Quoted 7 lines
amidi/amidi.c +6 -0
+				int len = message_length(*data);
+
+				/* check for variable length */
+				if (len == -1) {
+					int i = 0;
+					while (data[i++] != 0xf7);

Possible buffer overflow.

True. I could check while data + i < send_data + send_data_length.

Quoted 5 lines

amidi/amidi.c int main(int argc, char *argv[]) · line 599

Quoted 3 lines
amidi/amidi.c +2 -0
+				if (len > 0 &&
+				    (err = snd_rawmidi_write(output, data, len)) < 0) {

Same here.

This would be solved by previous boundaries check.

Quoted 7 lines

The most robust way to find message boundaries probably would be a state machine. However, there is actually no requirement to find all message boundaries; the _actual_ requirement is to insert pauses between SysEx messages. So don't bother to try to parse everything; just search for F0 or F7.

Well, I tough it would be useful to work in any type of MIDI message not just SysEx. So, do you think only SysEx then?

Signature · 1 line
Felipe
#3 lore raw
Quoted 11 lines · Felipe Ferreri Tonello
Quoted 8 lines · Clemens Ladisch

amidi/amidi.c static void parse_data(void) · line 330

Quoted 5 lines · Felipe F. Tonello
amidi/amidi.c +5 -0
+	case 0xf4:
+	case 0xf5:
+		return 0; /* ignore */
+	case 0xf7:
+		return 0; /* ignore */

Silent data loss is bad.

What do you mean by that?

When you want to send the byte F5, this code will silently drop it.

Quoted 9 lines
Quoted 4 lines

amidi/amidi.c static void parse_data(void) · line 335

Quoted 2 lines
amidi/amidi.c +2 -0
+	case 0x00 ... 0x7f:
+		return rstate_len ?: rstate_len - 1;

rstate_len - 1 == -1

?

That's why there is this condition, if rstate_len is zero, than it returns 0.

There is not need to use rstate_len when its value is already known and does not affect the computation; it could just be written as:

        return rstate_len ?: -1;
Quoted 8 lines
Quoted 5 lines

The most robust way to find message boundaries probably would be a state machine. However, there is actually no requirement to find all message boundaries; the _actual_ requirement is to insert pauses between SysEx messages. So don't bother to try to parse everything; just search for F0 or F7.

Well, I tough it would be useful to work in any type of MIDI message not just SysEx.

But there is no _actual_ use case, where somebody actually needed it. And as shown by all those problems, handling all types of messages makes the code too complex.

Regards,
Clemens

#4 lore raw

Hi Clemens,

Quoted 11 lines · Clemens Ladisch
Quoted 10 lines · Felipe Ferreri Tonello
Quoted 8 lines · Clemens Ladisch

amidi/amidi.c static void parse_data(void) · line 330

Quoted 5 lines · Felipe F. Tonello
amidi/amidi.c +5 -0
+	case 0xf4:
+	case 0xf5:
+		return 0; /* ignore */
+	case 0xf7:
+		return 0; /* ignore */

Silent data loss is bad.

Quoted 3 lines
Quoted 1 line

What do you mean by that?

When you want to send the byte F5, this code will silently drop it.

I see your point, you are right. Although, as we know, the underlying driver won't send this byte anyway, but that is not for the application to decide.

Quoted 8 lines
Quoted 8 lines
Quoted 5 lines

amidi/amidi.c static void parse_data(void) · line 335

Quoted 3 lines
amidi/amidi.c +2 -0
+	case 0x00 ... 0x7f:
+		return rstate_len ?: rstate_len - 1;

rstate_len - 1 == -1

?

Quoted 7 lines
Quoted 2 lines

That's why there is this condition, if rstate_len is zero, than it returns 0.

There is not need to use rstate_len when its value is already known and does not affect the computation; it could just be written as:

        return rstate_len ?: -1;

Right, that was not the original intent, so this code was actually wrong. But I will remove it anyway based on your comment below.

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines

The most robust way to find message boundaries probably would be a state machine. However, there is actually no requirement to find all message boundaries; the _actual_ requirement is to insert pauses between SysEx messages. So don't bother to try to parse everything; just search for F0 or F7.

Quoted 6 lines
Quoted 2 lines

Well, I tough it would be useful to work in any type of MIDI message not just SysEx.

But there is no _actual_ use case, where somebody actually needed it. And as shown by all those problems, handling all types of messages makes the code too complex.

True. So I will make sure to add this to the documentation.

Signature · 1 line
Felipe
#5 lore raw
In reply to Clemens Ladisch

On Fri, 12 Aug 2016, Clemens Ladisch wrote:

Off topic here but: what happened to that patch you proposed (requested by me) to enable/disable filtering of MIDI Clock messages, just like already exists for MIDI Active Sensing messages?

Signature · 1 line
MT