Amerio.Stephane Posted January 15 Posted January 15 Hi, I randomly have this crash during the init of the engine (null pointer exception) A simple relaunch of the app (absolutely same conditions) fix it most of the time. This random crash is always with the same call stack. Happens both in debug and release builds. We set a custom loading screen, but no custom code is associated with it beyond this: //In AppSystemLogic::init LoadingScreen::setEnabled(true); LoadingScreen::setTransform(vec4{ 1,1,0,0 }); LoadingScreen::setTexturePath("boot_screen_AUM.png"); LoadingScreen::setText("<xy x=\"%50\" dx=\"0\" y=\"%50\" dy=\"0\"/>LOADING_PROGRESS"); Any way to help here?
silent Posted January 16 Posted January 16 Hi Stephane, We have a single bug report with a similar call stack, where the only affected GPU is the RTX 2070 Super. If you could send us parallel stack screenshot from Visual Studio by navigating to Debug → Windows → Parallel Stacks after the crash, it would help a us determine whether this is same case or not. So far, we have not identified any workarounds for this issue, and we have consistently reproduced it with relative stability. Thanks! How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN
silent Posted January 16 Posted January 16 Also it would be great to know additional system information, such as: CPU Model GPU Model and driver version OS version (including patch) Thanks! How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN
Amerio.Stephane Posted January 17 Author Posted January 17 New crash RTX3090 nvidia 566.36, PC is Dell5820 Tower, i9-10900X, Win10 22H2 (enterprise) multistack attached. 1
Amerio.Stephane Posted January 17 Author Posted January 17 (edited) Here I attach also the font file we use for the LoadingScreen. Who knows... bitcrusher_condensed_bd.ttf Edited January 17 by Amerio.Stephane
silent Posted January 20 Posted January 20 Could you please try to launch debug engine with -video_debug 2 as well to gather an additional information from the GPU side? We need to see Output messages and console. Also it would be nice to see the actual engine log from bin directory, maybe there are some errors or warnings that can help in debugging. So far we tried to reproduce this behavior on multiple PCs, but without any luck (using your attached font and code from the first post). Can you also confirm that when you remove custom loading screen there is no crash anymore? Also, please try to delete dx12 shader cache files from data directory, maybe that also would result in a different behavior. Thanks! How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN
Amerio.Stephane Posted January 21 Author Posted January 21 -video_debug 2 => absolutely no warning/errors before the crash -log file: no warning/errors before the crashlog_crash_video_debug_2.txt -restored default loading screen and font: same behavior (so custom code/font are not the cause) -removed cache: same behavior (so previous cache corruption is not the cause) But the most interesting of all: the crash occurs 100% after the boot up of the PC (so 1st launch of Unigine app). After that, I couldn't reproduce it, until next boot. This is very strange... To be noted: the launch can happen immediately after boot up or 10min later, the 1st launch will crash.
silent Posted January 21 Posted January 21 Thanks for the update! We will continue our attempts in reproduction. Could this be an issue with SteamVR initialization? When you reboot the PC and start the engine (without Steam running), SteamVR loads and initializes during the engine startup, which can lead to random crashes. However, if you start the engine after SteamVR has already finished loading, there are no further crashes. If you could also share the startup parameters of the engine, we could test those as well. If you can reproduce the issue with an empty project, that would be even better. However, if there is no crash in the empty project, then there must be something in your application causing the problem. It could be related to either the code or the content. The challenge here is to identify which part is the root cause of this strange behavior. How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN
silent Posted January 28 Posted January 28 One final approach we can take is to capture the error using the NVIDIA debugger. To do so, please follow these steps: Copy attached GFSDK_Aftermath_Lib.x64.dll to bin directory Launch debug engine with -video_debug_crash_dump 1 In Visual Studio in Output window you should see the detailed information on crash Please copy this info and attach latest engine log.txt to your next response And regarding your previous message: Quote -video_debug 2 => absolutely no warning/errors before the crash Please don't forget to check the Output window in Visual Studio. In case of device removed errors and -video_debug 2 there should be some valuable information. Thanks! How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN
Amerio.Stephane Posted January 29 Author Posted January 29 Hi, I created ticket 17716 to follow this issue more closely and to be able to give more traces to you. Thanks 1
Recommended Posts