Hmm, are you using MPU401 intelligent mode or UART mode (as set in the dosbox.conf)?
There was a thread in "General" section with updated build (MPU401/MIDI improvements + MIDI input), maybe it helps with your (first) problem, especially if your program uses intelligent mode.

Regarding timing....maybe you should try with ALSA / Linux, it has a smaller output buffer than windows (if you are sending it out on MIDI port). Depends on host OS performance (kernel timing resolution/ latency and IPC involved), probably not Dosbox's fault.

im pretty sure its in intellgent mode. i dont know the differnce, but uart may not have worked when i tried it.

so far as that new build:

1. hopefully it will be clear how i use it,
2 I didnt know it adressed any midi isssues - though it just added MIDI input.

Will look now.
thnx

I'm having some issues with MIDi (detalied in another message)

Is there documentation anywhere for the MOPU401 modes I can chose?
none/uart/intellegent

how about for the device choices:
default/alsa/opensound/windows 32 bit/core audio?

Yes, I'm using intellegent mode..

Is there ANY hope of the MIDI clock send timing being fixed in DOSbox??

*** Merging related topics: No need for starting a new one ***

would be nice to know if this timing issue piques anyone's interest, or if it was already known.

thnx!!!

I'm here again 😀

If this works in UART mode (i.e. you set it to uart in config file)...than you don't have to even care about intelligent mode.
Btw. for windows just use win32 "device".

So, questions:
Do your clock messages get lost, or you just don't receive them at exact time when they are expected? What kind of precision do you get and how precise do you want it (the timing drift) to be ?

Do you mix MIDI clock with other MIDI data, and, if so, is your timing better when you don't?

Which MIDI outputs do you use to receive MIDI clock? (by output I mean win32 MIDI output, not necessarilly real device). How do you measure your MIDI clock precision?

I'm afraid this app only seems to work in intellegent mode.

If I run sequence with only MIDI clock, and no notes, its pretty stable. If I send just quarter notes for example, it's still fine. BUT, if I send many notes then the clock will step at some of these notes making the timing competely all over the map.

I'm not talking about subtle timing issues. I mean that you might end up with clock speed doubling when many notes are being sent.

I am running a slaved hardware sequencer with a LED for each quarter note, so its very easy to watch, and hear, massive timing changes.

Yes, I send to win32 output, then to a USB hardware MIDI inferface.

I'd love to get this resolved, so please let me know all i can do to assist.

(If you prefer via phone or email, just PM me.)

If your program is using intelligent mode, then from your description this seems like some kind of race condition bug with MPU-401. It is possible, but not guaranteed that it is corrected in newer build which I posted few months ago here on forums. (I'm not really doing this lately, as I'm not at home where system is properly set up).

So, try with the build, which I hopefully got right, from this thread (unless you already found it before):
does dosbox support MIDI IN yet?

Also, have in mind that MPU401 emulation was developed with somewhat limited number of games and applications and certainly misses some stuff that's not used by those (especially what's in CVS). I guess you use some kind of old sequencer, like e.g. texture ? (I didn't test such applications thoroughly).

I didn't try that build.

It says "mingw binary", which I thought was some kind of linux version, so never gave it a shot.

Just copy that exe over the old one?

Yes, just copy over the old one (or rename it so you can have both).

I did a bunch of tests with the new exe. Same behavior.

If I have a very simple midi file with a note on each quarter note, clock timing is fine.

If I simply add a second note of a new pitch to beat one, clock timing goes crazy.

Been using DOSBOX all month and it's amazing, but these two bugs sure are a drag..

The timig issue with sending MIDI clock makes unable to control an external sequencer, which is what I hoped to do.

Ok, time for an update..

As you know from above, I've had MIDI issues using an old sequencer under DOSBOX.. In the meantime, Don Wells created a new test version for Roland MT32 that FIXED THE CLOCK ISSUES I was having..

The seem unrelated , but i see that its just a question of how running status was handled.

His TEST6MT32.ZIP is what I use. This is amazing in that it solves it, as MIDI clock is now properly handled!!!

The update seem to be for .070 only, so I'm stuck using .70 until his changes get added to a more current version, so thought I throw in my two cents that its working GREAT now!

Dosbos is amazing, and these changes make it 10000% for me!

Dennis,

Thanks for your input and testing. Glad the Dosbox MIDI changes worked out for you and fixed your real-time MIDI message problem. The test build may be found on this VOGONs thread.

When I get all my MT-32 changes finished I'll submit the DIFF for possible review/conclusion. The fixes which helped your real-time MIDI will also be included.

Thanks again,

dvwjr