Vilem Otte said:
Other option is praying for AMD to fix their drivers, and I guess everyone here knows that it's not going to happen anytime soon. Sadly.

I have reported driver bugs two times (VK and CL, think it was ‘devguru’ forums). They requested a repo and fixed the bugs within weeks.
But i have seen OpenGL guys there getting less help.

Sadly, giving out repo for this project is not even an option (it is huge, requires special hardware or specific data sources … and then there is licensing, agreements, etc.)

Simple application remaking the issue, possible - yes. I might give that a shot, it's not going to take more than few hours. And would give me more time to move over to newer tech stack.

Vilem Otte said:
Simple application remaking the issue,

Yeah that's what i did. Also makes it easier for them. Maybe you find a work around while doing so.

Oh, i just remember: For me it did not work on AMD to mix old school glVertex() with modern OpenGL. I had to remove each and every deprecated GL calls, only then it worked. Was no problem on NV to mix.

JoeJ said:
Oh, i just remember: For me it did not work on AMD to mix old school glVertex() with modern OpenGL. I had to remove each and every deprecated GL calls, only then it worked. Was no problem on NV to mix.

I originally thought that the problem might be this. So I literally uncommented whole drawing and kept only glClear. Even just simple clear is still broken though (although I do quite a lot of GL operations in initialization - creating buffers, loading textures, loading shaders, etc.).

Just FYI and follow up.

It really is also crashing Blender and pretty much every single other OpenGL based software (basically Blender crashes within 30 seconds of usage).

The lack of support or any kind of response from AMD is very concerning to me.

I have problems with blender too

@Vilem Otte yes, i have the same experience from 2002. i had a sis6326 only, so i had to go to the street markets and bought an s3 virge dx, bought a riva 128, then bought a 3dfx voodoo rush. then i bought a 3dlabs permedia2, an intel i740, and a matrox mga mistyque, later on an ati rage, and a sis 305, then an s3 savage4.

after i have gathered all the 3d accelerators of the era for 1- 3 dollars, i have fixed all the bugs on all of them before releasing most of my software to the public.

but half decade ago i had enough. it was fewer vendors for the pc, but i didnt wanted to make same thing again and again with all the arm crappery from samsung fimg to adreno, mali, via, vivante, etc chipsets, so i just switched to software rendering instead - and problem solved.

I have temporarily solved the problem.

So what was breaking it - Adrenalin software (which is normally installed together with the driver). Even disabling it doesn't really help. One has to uninstall it and install the graphics driver by hand. I honestly doubt every user would do this.

This being said, I'm still going to move this forward to GTK4 and Vulkan - as the above solution is far from being convenient for users in general.

Vilem Otte said:
So what was breaking it - Adrenalin software (which is normally installed together with the driver). Even disabling it doesn't really help.

Oh, then my bet is it is caused by annoying popups like ‘Press Alt-X to record or do other bullshit things, and we're sure you do not need that key shortcut for some of your applications where you'd do something useful usually’.

After each driver update i have to disable all those crap shortcuts one after another in dirver to can work. >:(

With that Adrenalin hint, AMD might be able fix it quickly, i think.

To the op, I understand your pain, but from dealing with AMD and driver bugs especially OpenGL driver bugs, I would exercise patience. I've encounter 2 drivers bugs in the past and having filed reports along with reproducible cases, it took months to get a response on their forums( along the lines of what you are experiencing), but they did eventually fix the bugs. The annoying part was that they did not mentioned it was fixed, I just decided to test my app with the workaround removed and that code path worked without issues.
My advice is that if you have narrowed down the GL calls responsible for the issue, then try to see if there is a vendor extension that capable of doing the same thing. The reason I say this, is that strange enough, the issues I encountered were broken in AMD core OpenGL implementation, but their implementation of the Nvidia vendor extensions worked fine…imagine that.