This has bothered me for some time, so I did some investigation on it. The unit test for Midi assert-fails perhaps 10% - 20% of the time. Out of 100 or so test runs, I have isolated *what* the failure is, but have no answer to *why* it fails. I suspected that it my be the random buffer size value passed to midi_write_init() in test_midi_writer, but even if I "de - randomize" it (pass a prior failing sequence of values) it does not change the frequency of success or failure. I have only seen the failures occur with 2 different buffers, specifically the packets sysex4_1 (5 octets) sysex5_1 (1024 octets) The 3 symptoms that appears to always be true on the fail case: 1. Fail always occurs on reception of a SND_SEQ_EVENT_SYSEX 2. Received buffer is always exactly 1 octet shorter than what was expected 3. The expected and received buffers are identical except for a missing 0xF7 octet at the end of the received buffer. That's all I have time for, and I might not know enough about MIDI to continue further. But it would be nice if this intermitent Unit Test Failure was fixed.
