Hi,
I have a very disturbing problem. When I open PD I sometimes have this message popping out of nowhere :
warning: MIDI timing FIFO overflowed
It even sometimes happens without opening any patch. It isn't related to any specific MIDI device either. When I have this message MIDI starts working in a very erratic way, to say the least. I happened to me on stage and ruined a show.
I am totally clueless.
I run PD on osx 10.7.5 with a RME fireface UC.
Any help would be greatly appreciated.
Lucien

FIFO stands for "First In First Out". In this case it refers to the Midi input data buffer. Are any of your midi devices generating active sense or other continuously sent midi messages? See this thread (read till the end):

http://forum.pdpatchrepo.info/topic/9501/filter-out-midi-clock-information

@LucienR I am reading through an old post in French. It says that it's a bug that was fixed by Miller on the 6th July 2013 (the day after it was discovered). The fix made it into Vanilla 0.45 for osx, but never made it into Pd extended, and probably never will.
So you will have to use vanilla, and add the other libraries that you need. Its not as difficult as it sounds, and there is quite a lot of help around...........
If you are using Ableton Live to send the midi to Pd then apparently you can stop this error by putting Ableton's midi clock source into "pattern" mode instead of "song" so that the head position is not sent. Apparently in ProTools it is not possible to avoid this error, but maybe in some other programs it could be?.
Aparently Pd extended tries to decode 3 octets (thinking? there should always be 3) when there are only 2.
David.

Thanks to both of you.
I've already considered switching to vanilla, I think I will, only time will tell if it actually solves that bug, as it happens unexpectedly.

@whale-av said:

@LucienR I am reading through an old post in French.

Can you send me the link? I'm French actually.

I don't have Whale's French link, but I did notice this stuff on the Pd-dev list:

http://lists.puredata.info/namazu/pd-dev/namazu.cgi?query=midi+FIFO&max=20&result=normal&sort=score

@LucienR Salut, c'est surement sur codelab, peut-être ce lien ?
http://codelab.fr/5420

à+

I switched to Vanilla, and I thought it solved my problem, but it didn't.
I really need some help, I give a concert in two days, and I can't play without MIDI. I still have this message : "warning: MIDI timing FIFO overflowed", and the MIDI connections just stop working. I use PD 0.46-7 on osx 10.7.5 with a RME fireface UC. Any advice will be appreciated.
Lucien

@LucienR Hello Lucien....... What bad news. You have probably visited the post that I talked about previously, that @EEight correctly linked to. The posters were using osx as well.
I wonder if it is an osx problem. If your patch could run under windows or Linux then maybe that would be a solution. I am surprised to see that [midirealtimein] creates without error under windows....... windows not usually being too midi friendly in Pd.
I wonder, if you post your patch with a few ideas about how to do a "burn-in" test upon it, whether some testers could try it on the different platforms and let you know how it goes.
@gsagostinho might already have tested [midirealtimein] to death in Pd-l2ork?
David.

@whale-av Hmm, I never really used [midirealtimein]. By what I read, it works only under Windows anyway, and after trying to test it now and failing I believe the developers of Pd-l2Ork may not have made any modifications to it to make it work under Linux (though I may be completely wrong here, this opinion is based on a 5 minutes atempt...).

That's funny, [midirealtimein] works for me in Linux w/Vanilla. Thought it worked in L2Ork but too but I'd have to boot to my other patition to check ATM. Haven't tested it much on OSX but it creates with no error messages. I don't quite see how it relates to the OP's problem, thou.

@LucienR What is connected to the midi input on the RME? I assume that input is connected to your PD midi input. The first thing I would do is use a midi monitor program to check anything you have routed into PD via midi.

Thanks guys for considering my problem.

@whale-av Yeah I read the thread on Codelab, but in the end it seemed that using Vanilla instead of Extended solved the issue. Apparently it didn't in my case. I could post my patch for someone else to try it, but it's a very complicated one, with a dozen abstractions of my own, it requires specific sound files, and it's configured for specific MIDI devices, so I really doubt anyone would bother to actually run it...

@NoDsp I have two devices connected, one my RME, a FC-300 foot controller, and one on usb, an Akai max 25. I've already had midi monitor running, but it didn't help much, everything seemed normal to me, anyway I quite ignorant of the theory behind midi.

The baffling thing about this is that it happens like once every two months, with otherwise stable patches. I guess I'm just going to hope everything works fine tomorrow...

@LucienR If your patch is so complicated, did you try creating simpler ones to troubleshoot the problem? Have you experienced this problem and these error messages with any other patch of yours as well?

@gsagostinho The point is that as strange as it seems, I don't think this problem is related to my patch. It can happen with a patch that had been working perfectly for days. Besides, as it happens very seldom and unexpectedly, it's very hard to troubleshoot... I suspect this might be a hardware problem, but I'm clueless about it.

@LucienR Hello again Lucien...... I think you might be right that it is not actually related to your patch or even Pd. It is more likely to be drivers than hardware. I have amassed a bit of reading on this subject for Christmas.
I hope your concert is a success. My experience has often been better when things go wrong..... (broken strings, power cuts, stage collapsing, sets falling on heads etc.) and it is always worth having a good joke to tell while you are rebooting. Somehow the public are drawn more into your world, and feel that they are a part of it when they are not just watching you "perform".
David.

@LucienR Crossing my fingers for you tomorrow!

@whale-av @gsagostinho Thanks guys for your kind words. Everything worked fine today, fingers crossed as you say.

Glad things worked out for your performance, but for future considerations...

@LucienR said:

@NoDsp I have two devices connected, one my RME, a FC-300 foot controller, and one on usb, an Akai max 25.

Both those devices would raise my suspicion right away without even looking at the manuals. The Max 25 simply because it is a post-Japanese bankruptcy Akai branded product (too much experience) but also the Roland -- not because it's a Roland, but because it"s a midi product with a battery power option. Midi devices that run on battery power often implement continuous midi transmission of Active Sense messages. According to a quick check of the manual, this appears to be the case. I didn't read through enough to see if there's an "off" switch in the software for that. There may not be.

As far as I can tell the Max 25 does not use Active Sense, but in that case I would wonder if it was sporadically sending any manner of weird malfunctioning gibberish that could confuse the system.

You may not be able to see the A.S. message on a midi monitor. It's part of the System Realtime group of midi commands, so you would have to make sure the monitor isn't filtering those commands out. You still may not see it, as some of the simple monitor programs don't even define it. If you have a midi interface with an input LED it will show up as a dim continuous glow.

As that thread I linked indicates, Pd does not like Active Sense. Letting it in there is known to cause the problem you are having (and others) You will want to filter that stuff out with either external hardware or internal software before you send the output to Pd. If you don't need realtime commands from the Rolland, you can just block them all out. If you do need clock/stop/start etc. then you will need a filter that specifically blocks Active Sense and not the others.

Thank you @NoDSP, this is valuable information. As you assume, my foot controller sends Active Sense messages, and it does appear on MIDI monitor. I had no idea PD didn't like it. I have no clue about how to filter it, could you tell me a little bit more about it?

By the way, I'm in the process of completing a series of recordings with the patch I was mentioning, and to be honest this is the first time I'm rather proud of recordings I"m doing with PD . I'll post the links in the output~ section later, maybe tomorrow.

@LucienR Well here's bit about Active Sense:

http://www.blitter.com/~russtopia/MIDI/~jglatt/tech/midispec/sense.htm

As far as how to filter it, there's probably more than one midi app that can do the job. I've mainly been working with Midi Patchbay, but that only has an "all or none" system realtime filter. I just downloaded this:

http://www.macupdate.com/app/mac/10541/midipipe

Give it a try and see how it works out.