osx - CoreMIDI: midiReadProc only receives the first packet of 3 packets with the same timestamp -
i working on small program monitor midi output ni maschine 2. (idea add program change messages parts in ni machine , use them trigger vj-efects , other stuff)
the following test case gives problem. test part in ni machine has 1 note , 2 program change messages on first beat , second note 16th later.
when hit start , capture output midi monitor tool see: 
this correct. maschine sending. - notice 2 control changes , next note on packets have same timestamp.
when same simple virtual client (see code below) this: 
- as can see second control change , note on packets missing!
notice second third line (song position , continue) have same timestamp , in both cases received.
if took trouble read until point understand problem. know big difference between simpel virtual client , midi monitor program use of coremidi service plugin 'spying' midi output. limitation of coremidi or missing something?
below code virtual client: stripped down basic needs receive ni mashine. init sets virtual client, destination , sets uniqueid. readproc producing nslog messages missing packets shown above.
any suggestions appreciated
// // virtualclient.m // testmidireadproc // // created rob keeris on 02/06/15. // copyright (c) 2015 connector. rights reserved. // #import "virtualclient.h" @implementation virtualclient sint32 virtualinuniqueid = 1234567893; midiclientref client; midiendpointref virtualin; nsstring * miditypename(byte miditype){ switch (miditype) { case 0x80: return @"note off"; case 0x90: return @"note on"; case 0xb0: return @"controlchange"; case 0xc0: return @"programchange"; case 0xf2: return @"songposition"; case 0xf8: return @"clock"; case 0xfa: return @"start"; case 0xfb: return @"continue"; case 0xfc: return @"stop"; case 0x00: return @"invalidtype"; default: return [nsstring stringwithformat:@"unlisted miditype 0x%02x",miditype]; } } void midireadproc (const midipacketlist *list, void *procref, void *srcref) { const midipacket *packet = &list->packet[0]; // ?defined const avoid compiler warnings? (int = 0; < list->numpackets; i++) { if (packet->data[0] != 0xf8){ // filter out clock messages nslog(@"%llu packet(%i of %i) %@(0x%02x) 0x%02x 0x%02x", packet->timestamp,i+1,list->numpackets,miditypename(packet->data[0]),packet->data[0],packet->data[1],packet->data[2]); } packet = midipacketnext (packet); } } - (id)init{ osstatus result; self = [super init]; if (self) { // create client result = midiclientcreate(cfstr("myvirtualclient"), null, null, &client); if (result !=0) nslog(@"midiclientcreate error %i",result); // create destination result = mididestinationcreate(client, cfstr("myvirtualdestination"), midireadproc,(__bridge void *)(self),&virtualin); if (result !=0) nslog(@"midiclientcreate error %i",result); // set uniqueid dont have toggele output in ni maschine result = midiobjectsetintegerproperty(virtualin, kmidipropertyuniqueid, virtualinuniqueid); if (result !=0) nslog(@"midiclientcreate error %i",result); } return self; } @end extra info put in test in response hints of gene see happens when removing 0xf8 filter , using printf instead of nslog();
output midi monitor: 
output form code:
no solution clock received same timestamp.
tempo in test set 50 bpm (so 1 clock every 50 ms) checked see if there missing clock pulses isn't case. clocks received approx expected timestamp.
minor comment here. since you're doing this:
const midipacket *packet = &list->packet[0]; // ?defined const avoid compiler warnings? this
for (int = 0; < list->numpackets; i++) { should be
for (int = 0; < list->numpackets; ++i) { but doesn't fix problem. happens when remove filtering "if" (besides getting lots of f8s)? shot in dark: try printf see if nslog no good.
Comments
Post a Comment