[APP BUG]: Build broken on gcc due to issues in some of the amdgpu files

[C0rn3j]

Checklist

• I have searched for a similar issue in this repository and did not find one.
• I am using an official build obtained from releases or updated one of those builds using its in-app updater.

Describe the Bug

I am guessing this was introduced by #1869

[ 35%] Building CXX object CMakeFiles/shadps4.dir/src/qt_gui/elf_viewer.cpp.o
In file included from /home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/liverpool.h:25,
                 from /home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/renderer_vulkan/vk_presenter.h:8,
                 from /home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/imgui/renderer/imgui_core.cpp:17:
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/resource.h: In function ‘AmdGpu::CompMapping AmdGpu::RemapComponents(DataFormat, CompMapping)’:
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/resource.h:88:9: error: could not convert ‘{components.AmdGpu::CompMapping::b, components.AmdGpu::CompMapping::g, components.AmdGpu::CompMapping::r, components.AmdGpu::CompMapping::a}’ from ‘<brace-enclosed initializer list>’ to ‘AmdGpu::CompMapping’
   88 |         };
      |         ^
      |         |
      |         <brace-enclosed initializer list>
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/resource.h:96:9: error: could not convert ‘{components.AmdGpu::CompMapping::a, components.AmdGpu::CompMapping::b, components.AmdGpu::CompMapping::g, components.AmdGpu::CompMapping::r}’ from ‘<brace-enclosed initializer list>’ to ‘AmdGpu::CompMapping’
   96 |         };
      |         ^
      |         |
      |         <brace-enclosed initializer list>
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/liverpool.h: At global scope:
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/liverpool.h:892:34: error: declaration of ‘AmdGpu::DataFormat AmdGpu::Liverpool::ColorBuffer::DataFormat() const’ changes meaning of ‘DataFormat’ [-Wchanges-meaning]
  892 |         [[nodiscard]] DataFormat DataFormat() const {
      |                                  ^~~~~~~~~~
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/liverpool.h:892:23: note: used here to mean ‘enum class AmdGpu::DataFormat’
  892 |         [[nodiscard]] DataFormat DataFormat() const {
      |                       ^~~~~~~~~~
In file included from /home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/liverpool.h:23:
/home/c0rn3j/.cache/paru/clone/shadps4-git/src/shadps4/src/video_core/amdgpu/types.h:117:12: note: declared here
  117 | enum class DataFormat : u32 {
      |            ^~~~~~~~~~
[ 35%] Building CXX object CMakeFiles/shadps4.dir/src/qt_gui/main_window_themes.cpp.o

Reproduction Steps

Attempt to build AUR/shadps4-git -> this error

Judging by the Ubuntu CI passing on this, I am guessing something here on Arch is more strict. Seems to be clang vs gcc

I could not spot anything off about the recent changes vs the PKGBUILD other than having an extra cubeb dependency, which should be completely irrelevant to this issue.

Expected Behavior

No response

Specify OS Version

Arch Linux

[rharish101]

I modified the PKGBUILD to build shadPS4 with Clang instead of GCC, and it builds (and runs well). Here is the diff for the PKGBUILD:

diff --git a/PKGBUILD b/PKGBUILD
index c4687d7..f12e71a 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -19,7 +19,7 @@ depends=(
 makedepends=(
 	'boost>=1.84'
 	'cmake>=3.16.3'
-	'cubeb'
+	'clang'
 	'ffmpeg>=2:5.1.2'
 	'fmt>=10.2'
 	'git'
@@ -87,6 +87,8 @@ build() {
 	cmake -B build -S $_pkgname \
 		-DCMAKE_BUILD_TYPE=Release \
 		-DCMAKE_C_FLAGS_RELEASE="-DNDEBUG" \
+		-DCMAKE_CXX_COMPILER=clang++ \
+		-DCMAKE_C_COMPILER=clang \
 		-DCMAKE_CXX_FLAGS_RELEASE="-DNDEBUG" \
 		-DCMAKE_INSTALL_PREFIX=/usr \
 		-DCMAKE_SKIP_INSTALL_RPATH=ON \
@@ -102,7 +104,6 @@ package() {
 		'libavcodec.so'
 		'libavformat.so'
 		'libavutil.so'
-		'libcubeb.so'
 		'libfmt.so'
 		'libpng16.so'
 		'libswresample.so'

Therefore, I think the title should be renamed to reference that it is a GCC vs Clang problem.

[tomboylover93]

@abouvier You maintain the shadps4-git PKGBUILD, so you should look into replacing the default compiler with Clang as outlined above. It would be good practice as well as it is recommended over GCC.

[C0rn3j]

I think an actual solution would be to have a CI job that compiles with GCC, to catch these before they're merged.

This isn't the first time this has happened - #1729

Unless shad wants to explicitly drop gcc support, it's good that the environments vary, more testing happens that way.

[tomboylover93]

Since PR #2000 got merged this should be fixed. I've taken the idea of a GCC CI job into consideration and will open a PR for it later.

[Hermiten]

Can you try @C0rn3j ? Thanks you :)

[C0rn3j]

Compiles, thanks!