Jatos experiments that work perfect on my mac freeze for other (windows?) users
I have a puzzling problem and am hoping to find help here in finding a solution as I have time pressure to get this experiment to work for students. My issue might be related to yesterdays' about black screen freezing, but it in my case it seems specific to some users (so far it seems to be specific to windows users).
I programmed two OS web experiments on my mac, and the two of them run perfectly both in Open Sesame, in the external browser as launched from Open Sesame. For myself as well as the other mac user in my house the experiments on JATOS (via a general multiple link) work perfect as well. I tried Chrome, firefox, etc.
However, for my colleague and a couple of students who are windows users the two JATOS experiments, accessed via the same link, crash (seemingly every time around a similar point in time). There simply suddenly is a black screen (and no error message, also not in the browser console) after around 6 blocks of trials in one experiment (around trial 90-100) and around trial 50 of the first block in the other. For both experiments the freezing occurs not only when using an up-to-date Chrome browser and also when using an Edge browser.
In trying to solve the issue I already tried:
- Changing my OS version from the latest release 3.3.5 back to 3.3.0 (which I successfully used for JATOS experiments before).
- Changing the back-end from Legacy to PsychoPy.
- To log only a selected, small number of variables.
- Asking my colleague with a windows pc to export the experiment as a JATOS study from his computer, and upload it on JATOS.
Neither of these things changed the freezing behaviour.
Issues with the server seem highly unlikely, non of my colleagues have issues and when my colleague and I run an experiment simultaneously from a mac vs. windows pc it only crashes for the user on windows.
I would be happy to share the link to the experiment and code in PM if that is helpful.
Thanks,
Louisa
Comments
Update: I am presenting 30 different .png images over the course of a block (2 images on each trial). After reading posts describing similar issues (here and here) I tested if the black screen freezing for some users could be a memory issue rather than some incompatibility between mac/windows. And yes, when reducing the quality of those images it is now working fine.
This left me pretty confused however... @kri @Daniel What is being stored in memory that builds up in a block loop and that depends on the size of the .png images presented?
It seems that this is quite limiting for what type of experiments can be run online. Any tips on how to deal with memory use issues other than reducing the number of images used and their size/quality?
Hi @LousiaB.
This sounds very peculiar, especially since it is so platform dependent. I have no clue as to what may be the problem here. As you stated, no error messages are printed in the console, so there is little to go on. If the experiment hits a memory limit due to the space taken by images, I would expect an error message to appear in the console as well.
What happens if you check the resources used by the browser in Window's performance monitor (press ctrl+alt+del, select task manager, and click on the tab performance). Do you see a pattern of increasing memory usage by chrome there?
Does the problem occur on every windows computer, or just a few? It may be that these laptops are simply too old or contain too little memory for the demands of the experiment?
To test the memory issue further, could you try converting all images to .jpg and see if the freeze occurs then? png is a very resource hungry format, so if you don't need it (for instance, if your images do not contain any transparent areas) you might as well use jpg, which is much more economical.
Sorry for not being able to offer more help. As I said, there is little to go on, and there are no general tips to be given to tackle memory issues, except for the ones you have already done yourself (e.g. decreasing image size/quality).
I agree with what Daniel said but just to get an idea: what is the file size of your images?
Thanks! I don't have a windows pc available to test (happy to do so next time I get the opportunity) but can confirm that indeed for both of the experiments we no longer have issues with smaller images (originally +- 20 KB per image, in the new version 5-10 KB per image).