Home → Problems →

load_stimulus issue

• Edit

llglick

20 Dec, 2021 04:11 PM

Hi Chris,

We are having an issue with an experiment in which we are presenting images from a local directory. The experiment loads fine, but when load_stimulus is called the program crashes. The strange part is that this experiment worked fine a few months ago, and as far as we know nothing has changed about MWorks, the OS, etc.

I've attached a copy of the crash error and the xml- the error occurs on when these lines are executed:
<action tag="Load Stimulus 1" type="load_stimulus" stimulus="images[tstimOne]"></action>
<action tag="Load Stimulus 2" type="load_stimulus" stimulus="images[tstimTwo]"></action>

Do you have any sense of what the issue might be?

Thanks,
Lindsey.

• loadStimulusCrashReport.rtf 135 KB
• bunnies6.xml 25 KB

1. Support Staff 1 Posted by Christopher Sta... on 20 Dec, 2021 05:06 PM

   

   Hi Lindsey,

   It looks like the crash occurs not in load_stimulus but rather during update_stimulus_display (possibly immediately after the load_stimulus, although I can't tell from the crash log). It also looks like part of the load is happening asynchronously and hasn't completed yet. This may be the cause of the crash, which occurs inside the graphics driver.

   Would it be possible to add a short wait (maybe 100ms) right after the load_stimulus actions? That should give the asynchronous part of the load (which MWorks unfortunately has no knowledge of or control over) time to finish.

   Alternatively, I suspect this wouldn't be an issue with a recent nightly build. MWorks' stimulus rendering code has been modernized and basically completely rewritten since the build you're using, and it wouldn't even use the driver that's triggering this crash.

   Cheers,
   Chris

   • Edit

2. 2 Posted by llglick on 21 Dec, 2021 05:13 PM

   

   Hi Chris,
   We updated to the nightly and added the wait statement. Now there is a
   crash after the first trial.
   I've attached the Apple and Console crash reports.
   Thanks
   Lindsey.

   • load_stimulus_ConsoleCrash.rtf 11.7 KB
   • load_stimulus_AppleCrash.rtf 231 KB
   • Edit

3. 3 Posted by llglick on 21 Dec, 2021 05:18 PM

   

   Also- this seems to be a more general issue as we get the same crash when
   running other experiments that do not load images.
   Lindsey.

   • Edit

4. 4 Posted by llglick on 21 Dec, 2021 05:57 PM

   

   Hi Chris,
   Looks like there is an issue with Matlab communication. We ensured that
   the matlab setup is correct (using these
   <https://mworks.tenderapp.com/kb/data-analysis/setup-for-using-matlab>
   instructions), but it did not solve the issue.
   This is the error report that we got when using the terminal query:

   hullglick5:/ hullglick$ /Applications/MWClient.app/Contents/MacOS/MWClient
   2021-12-21 12:41:44.287 MWClient[2955:9428162] log file path:
   /Users/hullglick/Library/Logs/MWorks/20211221_124144287.log
   dyld: lazy symbol binding failed: Symbol not found: _mxCreateStructArray_800
     Referenced from: /Library/Application
   Support/MWorks/Plugins/Client/MWorksMATLABWindow.bundle/Contents/MacOS/MWorksMATLABWindow
     Expected in: /Applications/MATLAB_R2014b.app/bin/maci64/libmx.dylib

   dyld: Symbol not found: _mxCreateStructArray_800
     Referenced from: /Library/Application
   Support/MWorks/Plugins/Client/MWorksMATLABWindow.bundle/Contents/MacOS/MWorksMATLABWindow
     Expected in: /Applications/MATLAB_R2014b.app/bin/maci64/libmx.dylib

   Abort trap: 6

   Lindsey.

   • Edit

5. Support Staff 5 Posted by Christopher Sta... on 21 Dec, 2021 07:01 PM

   

   Did you recompile your LabJack plugin against the nightly build?

   • Edit

6. 6 Posted by llglick on 21 Dec, 2021 07:09 PM

   

   Yes

   • Edit

7. Support Staff 7 Posted by Christopher Sta... on 21 Dec, 2021 07:14 PM

   

   OK. Is it just MWClient crashing now, or does MWServer crash, too?

   Regarding MATLAB, MWorks' MATLAB tools and MWClient's MATLAB window are currently built against MATLAB R2018b. With the older MWorks build you were using, the MATLAB stuff was built against R2017b. It looks like you're using MATLAB R2014b, so I'm surprised that things worked even with the older build. Switching to R2018b or later should fix things.

   Chris

   • Edit

8. 8 Posted by llglick on 21 Dec, 2021 07:18 PM

   

   Yes- it is mwServer that crashes and mwClient stays open (though it is
   "frozen" and needs to be force quit).
   We're updating Matlab now...
   Lindsey.

   • Edit

9. Support Staff 9 Posted by Christopher Sta... on 21 Dec, 2021 07:59 PM

   

   OK. I asked because the latest crash report you sent was for MWClient. If there's a crash report for MWServer, can you send that, too?

   Also, if it's possible for you to provide me with the image files you're using, I can try running the experiment myself and see if I can reproduce the crash.

   Thanks,
   Chris

   • Edit

10. 10 Posted by llglick on 21 Dec, 2021 08:04 PM

    

    Sorry- I had that backwards- it's the Client that crashes not the Server.
    Let's see if the new Matlab install fixes everything. If not, I'll send
    you the images for testing.
    Thanks,
    Lindsey.

    • Edit

11. 11 Posted by llglick on 22 Dec, 2021 01:37 PM

    

    Updating MWorks and Matlab solved it- thanks!
    Lindsey.

    • Edit

12. Support Staff 12 Posted by Christopher Sta... on 22 Dec, 2021 02:10 PM

    

    Great! Thanks for letting me know.

    Chris

    • Edit

13. Christopher Stawarz closed this discussion on 22 Dec, 2021 02:10 PM.

14. llglick re-opened this discussion on 22 Dec, 2021 05:35 PM

15. 13 Posted by llglick on 22 Dec, 2021 05:35 PM

    

    Hi Chris,
    There is still one small issue: MWServer does not quit normally. You have
    to "close experiment" rather than just command-Q, or it will crash and need
    to be force-quit.
    Thanks,
    Lindsey.

    • Edit

16. Support Staff 14 Posted by Christopher Sta... on 23 Dec, 2021 03:02 PM

    

    Hi Lindsey,

    If there's a crash report, you can send it to me, and I'll take a look. However, you really should explicitly close the experiment whenever possible. Some things can't shut down properly otherwise.

    FYI, I'm on vacation starting today through the end of the month. If needed, I can follow up on this in January.

    Happy holidays!

    Chris

    • Edit

17. 15 Posted by llglick on 24 Dec, 2021 06:03 PM

    

    Hi Chris,
    Ok- we'll close the experiment going forward- but I'll grab an error report
    next time and send it.
    Also- we have a resurgence of the ehFeedback error with the new nightly.
    We merged your pull request from July that should fix this- so maybe the
    nightly introduces some new issue? Would be great if you could check on
    this when you're back in town.
    Have a good holiday,
    Lindsey.

    • Edit