rbdannenberg
commented
4 months ago Wow, this sounds significant -- great find! This stuff is tricky and amazing this hasn't come up after so many years of use, but maybe sysex is pretty unusual so this is definitely an rare case. I want to look at this carefully and have some other things to merge that I put off due to travel, but I think I can do this within a week. Your contribution, especially taking time to track down the problem, is greatly appreciated.
during a midi stress tests i noticed that sometimes sysex messages got stuck in PortMidi. what was specific to the error cases was that there was always a real time event just before the last byte of a sysex message. ie . [0xf0, 0x1d, 0x51, 0xf8, 0xf7]
coremidi sends one packet containing 0xf8 and 0xf7 and the old code did not handle that case correctly.