Segfaults in some tests about acid HOT 6 CLOSED

On my system, Tutorial 1-6 instantly segfault in the constructor of engine.
I run them on Ubuntu 20.04, compiled with g++ 9.

from acid.

On Windows when SPIRV is built with ENABLE_NV_EXTENSIONS=ON I am getting a segfault at the exact same point, this needs to be investigated further but is a cross platform issue. Resources would be distributed in a few data.X.zip files in the future, for now having Resources installed with cmake would be fine. Also on Ubuntu I am having issues compiling, it appears to be finding libSPIRV, but when a Tests is linked it cannot find it at all.

from acid.

I noticed Sources/Renderer/Pipelines/ShaderProgram.hpp uses a function from <glslang/Public/ShaderLang.h> but it's not included, and only SPIRV is linked against. Maybe that is causing some issue?

Also, SPIRV is privately linked, so that linkage won't be shared to any dependant tests/end-users.

Tests built just fine for me, it's just running that doesn't work.

from acid.

I have been trying out a lot of compile options with glslang and what I found is that AMD and NVIDIA options were causing issues in the reflections generated in Acid (some combinations just outright causing segfaults). I was able to find some workarounds with bitter-sweat results, I posted my findings here: KhronosGroup/glslang#1617

from acid.

Which version of the SPIRV-libraries are used here?

from acid.

The segfaults seen in the Test applications are due to glslang, the temporary fix was to just build it as a part of the Acid project to avoid the problem. Tutorials 1-6 have an unrelated issue causing them to not run correctly on all platforms, this will be fixed with a batch of graphics objects destruction fixes.

This ties into testing Acid on more platforms to catch dependencies not working as expected, driver issues, and use of a compiler sometimes resulting in runtime errors.

from acid.