MediaLab freezes, running on Vista
MediaLab freezes, running on Vista
We are building a study that includes a large number of variables (multiple personality instruments), and are having a problem with MediaLab freezing up after about 150 items. Specifically, after an arbitrary item it will go to a blank purple screen and nothing new will come up. We have found a few clues as to what is going on:
- It does not happen on a computer we have that is still running MicroSoft XP. It only happens with Vista
- The ctrl-alt-right arrow will exit the program if done immediately, but if one continues to press keys after reaching the blank purple screen it will lock up the program completely.
Any thoughts?
Thanks
How reliable is this? e.g., It happens every time, or x out of n sessions? When it does stop, what item type is it that it freezes on?
How arbitrary is the item--You mentioned it's after 150 items or so but what kind of variation is there here? Are we talking 149-151 sort of thing or a 140-160 kind of thing?
Also when it does freeze up after hitting a variety of keys, does it help if you click on the black border at the bottom and then pressing ctrl+alt+right (or just ctrl-right to got to the next item)?
So far, it has frozen every time we try to run the full experiment (running the individual questionnaires, all are fine).
Because the questionnaires are presented in random order, it stops on different items each time --usually a Scale Response item, because those are the most commonly used -- but it is consistently at about item 145-155. If it has to do with the number of items written into the data files, then it is about item 160-170 (i.e., including time stamps, instructions, "ok" statements, etc.).
Once frozen, neither ctrl+right or ctrl+alt+right works; the only escape we've found (so far) is ctrl+alt+delete to get the Task Manager.
Further information based on our efforts to figure out what is going on:
a) The program locks up after a certain number of responses have been made, regardless of how many experiments are run. I.e., somethings (Vista?) is tracking the total number of responses. We can run through a whole experiment that is lengthy but not long enough to reach the lock up point, then start another experiment and the second experiment will lock up quickly as it reaches the lock up point (based on the total responses across both experiments).
b) If we run an experiment, close MediaLab, and then run a second experiment, the total number of responses factor resets to zero.
This is really useful information. I agree it certainly sounds like some kind of memory issue--especially if closing and restarting MediaLab allows for the "count" to reset. My only remaining question is whether this also depends on some specific content in the experiment (i.e., a 3-way OS x #Items x Content interaction) or a simple 2-way OS x #Items interaction. Without seeing the content of your questionnaires I can't say. But one way to tell would be to administer a 200 item questionnaire, with the most basic content possible--say 200 instruction items. I've attached such a sample que file. Would you mind running it on your system and see if the result is the same?
The 200 item questionnaire freezes right after instruction item number 156 (every time, for three consecutive trials).
OK, that's hugely useful info. I don't think I asked you this yet--is this happening on multiple machines or is it unique to this single Vista machine?
The same thing happens on every Vista-loaded machine in the lab (n=8)
We have the exact same thing. 6 Vista laptops. Medialab freezes after item 150 or so. The experiment runs fine on any XP PC. When the sequence of questionnaires is changed Medialab will freeze on a different trial. Splitting up the experiment in 2 parts and restarting Medialab between the 2 sections resolves the issue.
We are experiencing the same problem in our lab on our vista computer. We seem to have maybe gotten further than item 150 (though I am not sure exactly how many items were complete), however, we experience the same purple freezing screen. If we do the exit code immediately (ctrl,alt,right arrow) we can save the data we have accumulated so far, but if other buttons are pressed, we have to ctrl, alt, delete out and lose all data. The funny thing is that we are experiencing the same problem with a machine that came with vista, but was later switched to XP??? Is it something about the vista computers and not necessarily the OS itself???
Did you do a clean install of XP on this machine that used to have Vista? e.g., first formatting the hard drive? If so, that is super odd and, yes, I would expect that would be a hardware issue. Would be curious to see the dxdiag.txt files from all these machines to compare their hardware.
For info on how to generate a dxdiag.txt file:
www.empirisoft.com/support/showthread.php?t=75
When I try to get the dxdiag.txt files from the vista machines, the computer says it cannot find the files. I can get the dxdiag.txt files off of xp machines, however, the one that gave us the medialab problem is currently being rebuilt so I cannot get that file either. =/ I am sorry I cannot be of much help.
That's odd. Hmm. Are you typing "dxdiag.exe" from the Run command on the Windows Start Menu? If so, what exactly is the feedback message when you try that?
I got the dxdiag.txt file. I am not sure if it will provide any info
The dxdiag.txt file indicates you are still running Vista on this machine. Is this the one that was supposed to have been converted back to XP? If not, do you have that one for comparison?
Actually, that dxdiag.txt file is from a Vista desktop machine. The laptop computer that had vista and was rebuilt with XP and was giving us problems, has since been rebuilt again back with Vista. Is everyone who is running medialab on Vista computers having this same problem? or is it just a few people?
Well at least that makes sense--I had never heard of anything like this on XP. We are working on a fix for this for Vista. In the mean time, all I can suggest is to either run it on XP or split the experiment as described above. As of this time I do not know the extent to which this is happening across Vista machines. I can at least say we have had many more requests for Vista authorization codes that we have had reports of this problem. But then that may simply be due to the fact that most experiments maybe be shorter than 150 items, or that the problem is not be reported by most users.
Another count for the Vista problem.
As described above. It seems to be happening in longer experiments (100+ trials). And yes, dividing the experiment into two or more smaller ones (with a restart of medialab) seems to help. Also, running on an XP machine avoids the whole problem. However, we've purchased a dozen laptops with vista to run subjects, and we can't reliably use them.
Any updates or news on this issue would be greatly appreciated
Bonus information:
Running all the medialab software (and Excel) in Windows XP (SP2) compatibility mode does not solve the problem.
I am having the exact same problem. Please let me know when there is a salution to this problem.
Thanks for the problem solving tip KAS, we can use lots of info like that to help rule stuff out. It almost certainly has to be memory related but it can't be as simple as a quantity issue. My guess is that it is related somehow to memory management in terms of how XP and Vista differ in this regard. Will let you know if we discover anything more useful. A prize for sure for anyone who can figure this one out--or at least find a contingency that leads us to a solution. Hmm.
Note also that we are working on a version of MediaLab that is fully Vista compliant. We did DirectRT first (which is in its 10th Vista beta as of Aug 19, 2008) because it simply would not run on Vista at all. Now that it seems to be running ok we are in the alpha stages of a Vista MediaLab which should be available for user beta testing within a few months (or earlier if all goes well). Fortunately, this issue and a couple problems with playing videos seems to be all the trouble we have seen so far with MediaLab on Vista.
See also...
www.empirisoft.com/support/showthread.php?t=704
another count for this problem... is there any update on this issue?
try this latest revision (as of October 31, 2008):
www.empirisoft.com/download/ml2008/ml2008.1.33.exe
[edited from user follow up email]
It worked on v2008.1.33. Thank you very much.
Looks like we maybe have a fix for Vista and Windows 7. Please feel free to post feedback here if you find the new install package resolves the issue for you.
See: http://www.empirisoft.com/Support/sh...ad.php?p=13484
