Virus TI - OS 5.1.1.0 - causing MIDIServer crash

    Ive now established that 5.1.1.0 is crashing MIDIServer on OS X 10.9.5 and 10.10.3.


    This includes crashing on a brand new Mac Book Pro, which has NO other software/drivers/hardware on it (other than supplied by Apple pre-installed)
    (tested with a TI Keyboard mk1), and also on an iMac which was working perfectly with 5.1.0.0


    Note: this only affects 5.1.1.0, 5.1.0.0 and prior do NOT exhibit this issue.


    Ive reported this to support @ access, but Im repeating here for information purposes, and to see how widespread a problem it is, and why you might see issues with this version of the driver, and Id suggest if you see this issue, then report it to Access (as they told me it was not a known issue), and perhaps mention on this thread.


    Causes of crash:
    a) Rescan MIDI - 100%
    b) Put Virus TI into standby - 100%
    c) adding new USB devices - sometimes
    d) exiting midi applications (DAW) - sometimes


    Implications:
    Crashing the MIDIServer process will halt ALL midi activity, and will not be restarted until a new MIDI enabled application is started. (eg. a DAW)
    In practice this means, you loose midi until you restart you DAW, this is how I first noticed the issue, as I plugged in a new USB midi device (case c) , and lost all midi.
    also some daws may have other issues, apparently unrelated, but caused by suddenly losing all midi devices, which would be pretty 'unexpected'.
    (Ive never seen midi server crash, prior to this driver issue)


    I think it is a pretty serious bug.



    How can you see if its affecting you:
    start the Console App (Utilities/Console)
    then under "Diagnostics and Usage information" ,
    then User Diagnostic reports, look for something beginning with MIDIServer
    e.g. MIDIServer_2015-05-14-165356_bear.crash(the name includes the date and time of the crash)
    If you click on it, if its the virus driver (de.access-music.virusti.driver.midi) crashing you will see the
    Thread 0 crashed, and a line like:
    1 de.access-music.virusti.driver.midi 0x00000001022ad45f 0x1022aa000 + 13407


    (a fuller example of the crash report is shown at the bottom of this post)


    Replacing (just) the midi driver with 5.1.0.0 fixes the issues
    Finally today, I have proved this is the 5.1.1.0 Midi driver....
    I took the virus midi driver (only) supplied in 5.1.0.0, and replaced the one installed in 5.1.1.0
    and surprise, surprise (not!)... it fixes the problem, and still works with VC/audio interface etc :)
    (don't try this ... Im a developer and I know what I'm doing etc., wait for a fix from Access :))


    other notes:
    I noted that @Marc said, in a post on this forum, that 5.1.1.0 and 5.1.0.0 were identical for the Mac ( a little secret, only the pc version changed, he said ;) ) ,
    I think this was mistake caused by a small version management issue by the developers of the midi driver.
    the driver in 5.1.1.0 has been in labelled as 1.8.0.0 (info string), but crucial in areas (mis)labelled as 1.7.0.0 (bundle info), as a change, it should have had its version number increased to 1.8.0.0
    the 1.7.0.0 driver is the driver supplied by 5.1.0.0, and its build tags and binary size are different (and has been demonstrated, its behaviour), so its definitely not the same build/driver version.


    as I mentioned on the other thread, I suspect 1.8 (5.1.1.0) aimed to fix a sleep related bug, but unfortunately has introduced a more severe bug.


    technical info:
    you can see the cause of the crash is a SIGTRAP, caused in CFRelease, due to CFRelease being called with NULL.
    this is completely correct! Apple clearly states in the SDK documentation for CFRelease


    void CFRelease ( CFTypeRef cf );
    ....
    Special Considerations
    If cf is NULL, this will cause a runtime error and your application will crash.


    (https://developer.apple.com/li…pple_ref/c/func/CFRelease)


    there is some additional info, on the original post I made about this on a different thread : Virus TI 5.1.1 Ableton 9.1 and Push


    example crash report (part of)
    trace will vary slightly depending upon which case (as detailed above) causes the crash, but the top two lines for the virus driver are always the same (well the 'code offset' is which is the important bit)
    0 com.apple.CoreFoundation 0x00007fff87e77010 CFRelease + 912
    1 de.access-music.virusti.driver.midi 0x000000011011545f 0x110112000 + 13407
    2 de.access-music.virusti.driver.midi 0x0000000110115d8f 0x110112000 + 15759


    technobear,


    thanks for the report. works fine here. on a mac pro 2013 and on a macbook pro from 2011. i can re-scan in AMS, switch it on and off etc and it doesn't crash. on top, i think we would have had a lot of people writing into support if there would be a general issue. the adoption rate of apple updates is faster than ever and we are not the only ones on 10.10. - so again, thanks for the report but not thanks to a general advise to our users to downgrade. i don't think that's necessary for the most of us.


    marc

    Zitat von Marc

    technobear,


    thanks for the report. works fine here. on a mac pro 2013 and on a macbook pro from 2011. i can re-scan in AMS, switch it on and off etc and it doesn't crash. on top, i think we would have had a lot of people writing into support if there would be a general issue. the adoption rate of apple updates is faster than ever and we are not the only ones on 10.10.


    it doesn't crash, does this mean you checked your console logs?


    I agree, I was also very surprised... which is why I said little until I had looked into it extensively,
    but two separate machines, one of which is completely 'clean room install', how can this be specific to me?


    as for people not reporting, I think its possible most would not notice anything other than 'odd behaviour', as they don't check the console logs,
    but who knows... and as a developer, I try to stick to facts... one clear crash is enough for investigation :)


    I do however, wonder why your not getting it.... perhaps something to do with it not being a 'clean room' install... either an upgraded virus install, or a continually upgrade Mac OS X install? If I were in your QA/Dev department, this is what I would be testing.
    also, why does just replacing the Midi driver fix the issue?


    again, I can't say why it works on other machines, as it doesn't work on any of mine... but as we all know, there are so many variations between machines, working (or not working) does not mean it will be the same on (lots of) others... such is the joy of qa/support for software products ;)



    Zitat von Marc


    so again, thanks for the report but not thanks to a general advise to our users to downgrade. i don't think that's necessary for the most of us.
    marc


    I meant don't upgrade if you haven't already... as it could lead to issues... but I can see it might be misinterpreted, so I'll remove

    Zitat von Marc

    the MIDI driver in 5.1.0.0 and 5.1.1.0 have exactly the same size and were build exactly on the same day.
    marc


    whats the size? (and are you talking about the binary, or the .plugin file, doesn't make any difference, I just need to know which to see which version you have)

    Also, for what it's worth, just in case there is an issue of clean installs:


    My Mac Pro was delivered with OSX 10.10.2 and I updated it to 10.10.3 as almost the first thing I did.


    I then went directly for Virus 5.1.1.0, so I had no intermediate versions on there.


    My MacBook Pro is 2011 vintage, so has been through several OSX versions and Virus versions.


    I've had MIDI crashes on my Mac Pro (from Marc's analysis of the logs), but no problems on my MacBook Pro.

    ok, Ive just (double) checked 5.0.8, 5.1.0 and 5.1.1
    I did this in two ways...
    a) check my Time Machine backup.
    b) took the official packages for each supplied, and unpacked the package manually (i.e. not installing) and checked both the BOM, and also extracted the payload.
    the result: all have drivers with different sizes binaries and dates.


    and as I stated in the original post, they all have not been packaged 'correctly' with different versions....
    basically they are all (except the getinfo string) have 1.7.0.0 (they should have been, at least, 1.7.0.0, 1.7.0.1, 1.7.0.2)
    as i said a developer 'mistake', and makes @Marc and other supports jobs alot harder.


    so Im wondering, perhaps, if you do an upgrade, its comparing the bundle version, says its not changed, and so doesn't bother replacing them.
    - Ive not looked into this, I'll leave this to the 'reader' :)
    but this might explain the why some users are not affected... they are still running the driver from a previous version.


    (btw, I tested, Im thorough :), not only does the 5.1.0 driver work with the rest of 5.1.1.0 but also the driver with 5.0.8 ... both do not exhibit the issue of 5.1.1.0)


    just seen @ChrisCabbage post, your experience would also tend to support the 'upgrading' not installing the new driver theory too

    It's certainly an interesting hypothesis.


    Anything I can do to help confirm it?


    I've not heard anything back from my support emails for some time now...


    My issues may be more complex though. This might be just one of my problems.

    sure.... so your saying
    Mac Book Pro, has been upgraded to 5.1.1.0 and NO problems
    Mac Pro, was a clean install to 5.1.0.0 and has problems (with Midi crashes)


    ok, so go to /Library/Audio/MIDI Drivers/
    In there you will find : Virus TI MIDI Driver.plugin


    tell me that date and size of this for both the MBP and the MP.


    also, as mentioned in the first post, on your MP go check the console (details in the above post) and see if you can find a MIDIserver crash file,
    and if it looks similar the one at the top of my post.


    THEN... if you feeling brave :)
    on the MP, copy the Virus TI MIDI Driver.plugin somewhere safe (you can get it back, but easier to know)
    then copy the Virus TI MIDI Driver.plugin FROM the MBP to the MP, to exactly the same place
    restart the machine, and see if it helps. (in particular the test cases I list at the top... I cannot say this is the cause of ALL issue, just the ones I've encountered)


    (but still let us know time/sizes ... as then we can see if the MBP had the 5.0.8 driver, or the 5.1.0 driver)



    Im assuming Access will be trying some fresh installs, and also some upgrade installs to help determine if this is a possible issue,
    and then advise us... the slight issue is 5.1.0.0 is no longer on the beta downloads, luckily I found a copy on my Time Machine :)

    I get Midi server crash from doing automation with Virus TI:


    Process: MIDIServer [67931]
    Path: /System/Library/Frameworks/CoreMIDI.framework/Versions/A/MIDIServer
    Identifier: MIDIServer
    Version: 89.3
    Code Type: X86-64 (Native)
    Parent Process: launchd [186]
    Responsible: MIDIServer [67931]
    User ID: 501


    PlugIn Path: /Library/Audio/MIDI Drivers/Virus TI MIDI Driver.plugin/Contents/MacOS/Virus TI MIDI Driver
    PlugIn Identifier: de.access-music.virusti.driver.midi
    PlugIn Version: 1.7.0.0 (1.7.0.0)


    Date/Time: 2015-05-19 01:15:02.316 +1000
    OS Version: Mac OS X 10.9.5 (13F1077)
    Report Version: 11


    Crashed Thread: 0 Dispatch queue: com.apple.main-thread

    Close to what's happening I think.


    I need to take my daughter somwhwre this evening, but will try and check things out when I get back.


    The only thing I'm not 100% sure of, is if I had 5.1.0.0 on my MacBook. I *might* have gone from 5.0.8.0 straight to 5.1.1.0. The effect should be similar though, from what you've described.

    @ChrisCabbage , I expect so ... as I said both 5.0.8 and 5.1.0 are both fine to use, so if its working ... its likely to be the case.


    @willbanks looks like the issue, but can you just confirm the next lines include , the following (or similar) , as this is an important bit.

    Code
    Application Specific Information:
*** CFRelease() called with NULL ***
Thread 0 Crashed:: Dispatch queue: com.apple.main-thread
0   com.apple.CoreFoundation          0x00007fff87e77010 CFRelease + 912


    as I mentioned in the original post, Ive seen this in a few different cases.. some Ive not listed, its just the first 2, I knew were reproducible and so hoped would help Access isolate the issue.

    My MacBook Pro's Virus MIDI driver looks like:


    drwxr-xr-x 3 root admin 102 22 Jun 2011 Virus TI MIDI Driver.plugin


    That's a very old timestamp!


    Mac Pro to follow...


    Edit: hang on, that's a directory! Do you want the size of the contents of the directory?


    Edit2: Ah - shouldn't have used the terminal. Here.s what Finder tells me:


    Size: 227,922 bytes
    Date as per command line
    It's showing as V1.8.0.0

    The Mac Pro is showing:


    227,992 bytes


    Times: 21 October 2014 2:37pm


    Also showing V1.8.0.0


    ...and I only got my Mac Pro about 3 weeks ago, so the timestamp must be for the driver release.


    I'd say the time/date is a better indication of which release it is, if the versioning is wrong.


    Same sizes though, which is interesting.

    Here's your smoking gun from the Mac Pro. I have around 40 instances! Marc and the Access guys should have a whole load of these too. I gave them full diags dumps a couple of weeks ago. Marc also mentioned in one of the support emails that my MIDI driver was crashing.


    ...and I've just checked the MacBook Pro.


    There are a couple of similar crashes on there too (Null pointer passed to the same function)!


    Looks like it's a bug which has been lurking for some considerable time.


    As to why it's being tickled so easily on my new machine - timing on the faster machine with more cores?

    @technobear


    Thanks, yes:


    Application Specific Information:
    *** CFRelease() called with NULL ***


    Thread 0 Crashed:: Dispatch queue: com.apple.main-thread
    0 com.apple.CoreFoundation 0x00000001052fa4c4 CFRelease + 724
    1 de.access-music.virusti.driver.midi 0x000000010c02b45f 0x10c028000 + 13407

    @ChrisCabbage ... are you sure, on the MBP when you installed 5.1.1.0, you didn't at any point, do an uninstall first?
    (I ask, Access support (understandably) seem to always tell you to uninstall/install if you report any issue)


    anyway, I can't really speculate on this on how some 5.1.1.0 don't have the issue, as I have never seen a 5.1.1.0 install that doesn't exhibit the issue, so can't investigate this, but Access can as have said there 5.1.1.0 installs don't have midi server errors... hence why I asked for binary sizes, etc. ... and Ive not the time to start doing test installs of old version on new versions etc... besides, I have to leave some work for Access to do ;)


    btw: you can see what I mean about the incorrect versioning, the getinfo... which is what you see in Finder, is showing 1.8.0.0, but notice how the crash report details 1.7.0.0

    Code
    PlugIn Identifier:       de.access-music.virusti.driver.midi
PlugIn Version:          1.7.0.0 (1.7.0.0)


    @ChrisCabbage , @willbanks - If you want , I can send you a copy of the 5.1.0.0 midi driver, which you can use to replace the 5.1.1.0, and then you can see if these MIDI server crashes disappear.
    send me a PM if you want to do this, as I don't want to upset Access by appearing to be distributing modified installs.
    ( you can always reset your install, replacing with the driver currently there, or by doing an uninstall and then a reinstall)


    again, to be clear... Im not saying this fixes all issues, I can only vouch for the issues I have encountered, and re-tested with the old driver.

    Have you by any chance done a binary comparison between the driver versions?


    I've never had a problem with the MacBook, so I've never had to uninstall. Not unless one of the older installers does an uninstall first.


    Have dropped you a PM.