unresolved reference to function `currentPhaseFunction' on amdgpu on Linux

[oldherl]

I am using Arch Linux on my AMD Radeon 680M (Ryzen 7 6800H) and calcmysky only produces the following error message.

I have reported to Arch Linux packager https://gitlab.archlinux.org/archlinux/packaging/packages/calcmysky/-/issues/1
but the feedback is that it works fine on NVIDIA machines. Is it a bug or that amdgpu is known to be not working?

$ calcmysky --opengl-debug-full /usr/share/stellarium/atmosphere/default/params.atmo --out-dir /tmp
CalcMySky 0.3.4
Compiled against Qt 6.9.0
Running on Arch Linux x86_64
OpenGL vendor  : AMD
OpenGL renderer: AMD Radeon Graphics (radeonsi, rembrandt, LLVM 19.1.7, DRM 3.61, 6.14.4-1-lily)
OpenGL version : 4.6 (Core Profile) Mesa 25.1.2-arch1.1
 GLSL  version : 4.60
GL_ARB_shading_language_420pack is supported
Writing parameters to output description file... done
debug callback called: severity Notification, source Shader compiler, type Other, msg: Shader Stats: SGPRS: 128 VGPRS: 8 Code Size: 116 LDS: 0 Scratch: 0 Max Waves: 16 Spilled SGPRs: 0 Spilled VGPRs: 0 PrivMem VGPRs: 0 LSOutputs: 0 HSOutputs: 0 HSPatchOuts: 0 ESOutputs: 0 GSOutputs: 0 VSOutputs: 0 PSOutputs: 0 InlineUniforms: 0 DivergentLoop: 0 (CS, W64)
debug callback called: severity Notification, source Shader compiler, type Other, msg: Shader Stats: SGPRS: 16 VGPRS: 12 Code Size: 64 LDS: 0 Scratch: 0 Max Waves: 16 Spilled SGPRs: 0 Spilled VGPRs: 0 PrivMem VGPRs: 0 LSOutputs: 0 HSOutputs: 0 HSPatchOuts: 0 ESOutputs: 0 GSOutputs: 0 VSOutputs: 0 PSOutputs: 1 InlineUniforms: 0 DivergentLoop: 0 (PS, W64)
debug callback called: severity Notification, source Shader compiler, type Other, msg: Shader Stats: SGPRS: 24 VGPRS: 16 Code Size: 256 LDS: 0 Scratch: 0 Max Waves: 16 Spilled SGPRs: 0 Spilled VGPRs: 0 PrivMem VGPRs: 0 LSOutputs: 0 HSOutputs: 0 HSPatchOuts: 0 ESOutputs: 0 GSOutputs: 0 VSOutputs: 1 PSOutputs: 0 InlineUniforms: 0 DivergentLoop: 0 (VS, W32)
Test texture true average: 0.634921, 0.47619, 0.31746, 0.15873
Test texture mipmap average: 1, 0, 0, 0
WARNING: Mipmap average is unusable, will resize textures to power-of-two size when average value is required.
Working on wavelengths 360, 391.333, 422.667, 454 nm (set 1 of 4):
debug callback called: severity Notification, source Shader compiler, type Other, msg: Shader Stats: SGPRS: 32 VGPRS: 16 Code Size: 224 LDS: 0 Scratch: 0 Max Waves: 16 Spilled SGPRs: 0 Spilled VGPRs: 0 PrivMem VGPRs: 0 LSOutputs: 0 HSOutputs: 0 HSPatchOuts: 0 ESOutputs: 0 GSOutputs: 0 VSOutputs: 1 PSOutputs: 0 InlineUniforms: 0 DivergentLoop: 0 (VS, W32)
debug callback called: severity Notification, source Shader compiler, type Other, msg: Shader Stats: SGPRS: 16 VGPRS: 8 Code Size: 76 LDS: 0 Scratch: 0 Max Waves: 16 Spilled SGPRs: 0 Spilled VGPRs: 0 PrivMem VGPRs: 0 LSOutputs: 0 HSOutputs: 0 HSPatchOuts: 0 ESOutputs: 0 GSOutputs: 0 VSOutputs: 0 PSOutputs: 1 InlineUniforms: 0 DivergentLoop: 0 (PS, W64)
[WARN] QOpenGLShader::link: error: unresolved reference to function `currentPhaseFunction'

Failed to link zero-order scattering rendering shader program

[10110111]

Yes, I also see it on my Radeon 890M (Ryzen AI 9 HX 370). This also reproduces with LIBGL_ALWAYS_SOFTWARE=1. So it's a regression in Mesa. It used to work in Mesa 21.2.6, but fails on Mesa 25.0.3-1ubuntu2. Works fine on the proprietary NVIDIA driver on the same machine.

This should be reported to Mesa.

[oldherl]

Yes, I also see it on my Radeon 890M (Ryzen AI 9 HX 370). This also reproduces with LIBGL_ALWAYS_SOFTWARE=1. So it's a regression in Mesa. It used to work in Mesa 21.2.6, but fails on Mesa 25.0.3-1ubuntu2. Works fine on the proprietary NVIDIA driver on the same machine.

This should be reported to Mesa.

Thanks for your timely response. However, I am not familiar to OpenGL or graphics in general, so I have no idea how should I report that to mesa. Is there any chance that you could report it for me?

[10110111]

Is there any chance that you could report it for me?

Well, I've been experiencing this for a week already (since purchasing a new laptop with the Ryzen), but haven't found time to report it yet. Hopefully I'll be able soon, but no guarantees.

[10110111]

The problem appears to be that some unused functions aren't eliminated by the Mesa's compiler, which by itself isn't a bug, just an inconvenience, so it's better to add a stub in CalcMySky.

I've fixed this in CalcMySky, will push an updated set of shaders to Stellarium shortly.

[10110111]

Done in Stellarium/stellarium@a3ef040.