Hi guys, Sean and I have slowly been poking at a bug where MIDI messages "seem" to be disappearing on Linux, and I think we've finally made some progress in at least assessing what's really going on: https://bugs.launchpad.net/mixxx/+bug/342952
The problem is that when we quickly fire a bunch of MIDI messages at a device (eg. SCS.3d) in succession, something weird happens and the device behaves as if some of those messages were lost. For a while we thought it was some ALSA MIDI queue overflowing, so I spent a couple hours reading the API docs and using some calls to increase the ALSA sequencer memory pool size. None of that made any difference, so I asked Sean to get some data to verify that our messages weren't actually being sent, and low and behold, aseqdump reports that our messages _are_ being sent (see the attachment in the bug thread). However, this raises a couple of questions now: 1) Was that aseqdump log captured from the output of Mixxx or the input to the SCS.3d device? I don't know how aseqdump works, so if we are still overflowing some ALSA MIDI queue, I can see it showing up on the SCS.3d side and not the Mixxx side. (Mixxx->ALSA->[overflow]->SCS.3d?) 2) Is there a lower-level buffer in the kernel that's getting full? Alternatively, I wonder if the USB-MIDI kernel module is intentionally limited to sending at the MIDI baudrate. Food for thought. If anyone has any ideas as to what's going on or is an ALSA sequencer guru, now's a great time to chime in... :) Thanks, Albert ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects _______________________________________________ Mixxx-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mixxx-devel
