New bug/crash from 15 days with new ES release

Hex

@pjft : @ChauncyBlack 's error comes from themes while @darknior 's error stems from Joysticks. Seems independent hence my request

pjft

@hex Oh, sorry. I'm not sure what suggests that it is a themes related issue in any way, but I might have missed something in this thread then.

We'll follow up separately then.

Hex

Oh my bad its not themes but it definitely not related to the Joystick bug.

@darknior 's bug is at a specific assert. that is the only point of failure. I dont know about the other bug so I asked for it to be posted in another thread. Dont want to confuse logs of both.

darknior

This night i have try ES with Black screensaver,no video and Power Saving to none and i have no crash ... strange ?

Hex

ok tomorrow night try video screensaver and PS disabled.

msheehan79

Hi guys, I thought I would add some input to this issue as I have experienced this rather frequently as well.

I am running ES 2.3.1 so this issue is not related to PS. From what I have been able to gather it is definitely related to a joystick disconnecting while ES is loading and it is not specific to Bluetooth as I do not use any BT devices in my setup.

My system is using 2 x XBox 360 Wireless controllers connected to an HDE USB Xbox 360 Wireless Receiver for Windows (https://www.amazon.com/gp/product/B0096PLB9O/ref=oh_aui_detailpage_o07_s00?ie=UTF8&th=1). I also have 2 x iBuffalo USB gamepads but I think the issue occurs mostly with the XBox controllers.

In my system I am using xboxdrv and the runcommand scripts to launch and kill xboxdrv with different controller mappings for specific systems as shown in the Wiki (https://github.com/RetroPie/RetroPie-Setup/wiki/Universal-Controller-Calibration-&-Mapping-Using-xboxdrv). So I suspect that in many cases the split seconds that xboxdrv is being killed/restarted after exiting an emulator happens to be occurring at the same time that ES is checking for joysticks and therefore crashes.

However I am sure it is not always tied to xboxdrv - to do some testing, I disabled xboxdrv to simplify my setup, and I was able to get ES to crash consistently by turning off the XBox 360 controller immediately after exiting an emulator.

There is nothing unusual in the es_log.txt file, so the only error message is the assert failure that is reported on the screen.

I guess the question comes up, is there a better way to handle the error in the scenario where a joystick disconnects rather than coming to a complete halt? Maybe some sort of timeout loop that attempts to reconnect a few times before failing a bit more gracefully, or some process to allow the user to reconnect a controller?

Just thinking out loud a bit. Happy to help in whatever way I can to troubleshoot further if needed.

Hex

@darknior Is this resolved for you? You can try different screensavers with PS disabled to check if the error happens

pjft

@msheehan79 thank you for the detailed account. Indeed that angle makes a lot more sense.

We could perhaps try to look into better handling of that particular cycle in that edge case, though given that the problem seems to be elusive enough I'm not sure how to confirm whether that change would effectively be affecting and fixing it in any way.

@Hex any thoughts?

Hex

I think the assert fails when ES encounters a false JS disconnect. What should be done in such case is the joystick should be re-added. I shall look into submitting a PR for you to test. Since this bug is not reproduceable, testing will be a nightmare.

msheehan79

Hi guys,

Thanks for any help or ideas on what the issue is.

I was working on some Amiga game settings which has involved frequent starting and exiting the emulators, and just in the past hour or so I've crashed it twice. So at least for my setup the error is reasonably reproduced probably once every 10 times I exit an emulator.

One detail that I did notice however is when it crashed, the assert that is failing is actually the one at line 71 and not line 67. ( assert(joy); )

I might try adding a short delay to my runcommand-onend.sh script to see if adding a second or two to allow xboxdrv to finish whatever it may be doing before it goes back to ES makes a difference. Just grasping at straws probably, but I figure it can't hurt to try.

darknior

@hex Hi sorry for the delay ... i don't know what was update but i don't have any crash from more than 15 days now !
I think it is ok.
I have update to Retropie 4.3 yesterday, i will try it too.