So i’m still having issues with the current build LWJGL3 and OSX. I am getting a hard crash that is crashing OSX to the login screen. As a result, I don’t have any log to look at. I replicated this using the basic game project from the sdk with all the libraries built from master. I was also using the “-XstartOnFirstThread” flag. I’ve stepped through the code a little bit to try to narrow it down and it appears that it is crashing when it polls events at some point after the window is shown. Has anyone else had an issue on OSX with LWGL3? I can’t even get a crash log because of how badly it is crashing on osx.
So I went through my logs again and the only reference I can remotely find is a crash log in the system reports for the “WindowServer” which shows a divide by zero error in the apple geforce driver. This error repeats itself every time I try to run the basic application. I assume this error is what is kicking me back to the login screen, but I don’t know what the root cause is with lwjgl3.
Yes, that is the worst crash I’ve seen in awhile. I have tried both older / newer version of the osx driver as well as nvidia’s driver and updated to the latest version of osx. Changing between the drivers still crashes in exactly the same way. I’ve also tested with version 3.1.2 of lwjgl as well as 3.1.5.
Quick question: would it be possible to release beta-1 as SDK binaries? The last version I get offered as binary download is v3.2-prealpha-sdk1. My guess is it would increase beta-testing participation if an SDK download were present. Would definitely work for me
I’ve also looked into it deeper and have found a workaround for travis, but that’s linux-only then firstly.
I wonder what happens if you take the gradle template and just specify jme3-jbullet as dependency, you should also experience that problem?
We have jbullet as jar in our libs directory, however it’s not known to maven as jbullet:jbullet, it’s just there.
This means, I’d do an mvn install:install-file -Dfile=jbullet.jar -DgroupId=jbullet -DartifactId=jbullet -Dversion=0.0.1 -Dpackaging=jar
@Momoko_Fan Any Idea what a better solution would be?
The SDK itself isn’t really broken but I’ve decided against dropping jbullet because of that, so that SDK users can debug their physics problems by simply swapping out jbullet and bullet-native.
So removing jbullet is not a good workaround
Edit: Further Investigation has shown that the actual problem seems to be that jbullet did not mention a version and hence it couldn’t even be parsed.