Arch Linux User Repository

This update reworks the PKGBUILD to better match Sunshine's current upstream build system and Arch packaging expectations.

Main changes:

• updated maintainer/contributor metadata
• expanded runtime and build dependencies to cover the current upstream feature set
• added configurable build toggles for tests, headless testing, CUDA, and pinned commits
• simplified source handling by using the main git source with optional commit pinning
• switched submodule handling to a recursive init/update flow
• replaced the old Boost workaround with a smaller patch so the packaged Boost version is preferred instead of forcing an exact upstream version
• cleaned up the build logic and exported upstream version metadata from the actual checked out source state
• added build-time validation for AppStream metadata and desktop files
• made the test step optional but more complete when enabled, including XML results and gcovr coverage output
• kept CUDA support flexible, with autodetection and fallback behavior when CUDA is unavailable
• aligned install/build variables and package metadata with current upstream paths and AUR package naming

Overall, the goal of this patch is to make the package more reproducible, easier to maintain, and closer to upstream while still fitting Arch/AUR conventions.

'qt6-base' 'qt6-svg'

Are missing from makedepends

Unfortunately, building the package fails for me since yesterday:

==> Making package: sunshine-git 2026.403.124357.r0.gba4db46-2 (Tue May 26 10:48:02 2026)

[...]
-- Could NOT find Qt6Svg (missing: Qt6Svg_DIR)
CMake Warning at cmake/macros/common.cmake:27 (_find_package):
  Found package configuration file:

    /usr/lib/cmake/Qt6/Qt6Config.cmake

  but it set Qt6_FOUND to FALSE so package "Qt6" is considered to be NOT
  FOUND.  Reason given by package:

  Failed to find required Qt component "Svg".

  Expected Config file at "/usr/lib/cmake/Qt6Svg/Qt6SvgConfig.cmake" does NOT
  exist



  Configuring with --debug-find-pkg=Qt6Svg might reveal details why the
  package was not found.

  Configuring with -DQT_DEBUG_FIND_PACKAGE=ON will print the values of some
  of the path variables that find_package uses to try and find the package.

Call Stack (most recent call first):
  third-party/tray/CMakeLists.txt:79 (find_package)


CMake Error at cmake/macros/common.cmake:27 (_find_package):
  By not providing "FindQt5.cmake" in CMAKE_MODULE_PATH this project has
  asked CMake to find a package configuration file provided by "Qt5", but
  CMake did not find one.

  Could not find a package configuration file provided by "Qt5" with any of
  the following names:

    Qt5.cps
    qt5.cps
    Qt5Config.cmake
    qt5-config.cmake

  Add the installation prefix of "Qt5" to CMAKE_PREFIX_PATH or set "Qt5_DIR"
  to a directory containing one of the above files.  If "Qt5" provides a
  separate development package or SDK, be sure it has been installed.
Call Stack (most recent call first):
  third-party/tray/CMakeLists.txt:83 (find_package)


-- Configuring incomplete, errors occurred!
==> ERROR: A failure occurred in build().
    Aborting...
error: failed to build 'sunshine-git-2026.403.124357.r0.gba4db46-2': 

Seems like it needs a dependency on qt6-svg

I'm getting this since this morning:

/tmp/makepkg.builddir/sunshine-git/src/sunshine/src/platform/linux/wayland.h:37:9: warning: type ‘struct dmabuf_t’ violates the C++ One Definition Rule [-Wodr]
   37 |   class dmabuf_t {
      |         ^
/tmp/makepkg.builddir/sunshine-git/src/sunshine/src/platform/linux/wayland.h:37:9: note: a different type is defined in another translation unit
   37 |   class dmabuf_t {
      |         ^
/tmp/makepkg.builddir/sunshine-git/src/sunshine/src/platform/linux/wayland.h:95:24: note: the first difference of corresponding definitions is field ‘gbm_device’
   95 |     struct gbm_device *gbm_device {nullptr};
      |                        ^
/tmp/makepkg.builddir/sunshine-git/src/sunshine/src/platform/linux/wayland.h:95:24: note: a field of same name but different type is defined in another translation unit
   95 |     struct gbm_device *gbm_device {nullptr};
      |                        ^
/tmp/makepkg.builddir/sunshine-git/src/sunshine/src/platform/linux/wayland.h:95:12: note: type name ‘wl::gbm_device’ should match type name ‘gbm_device’
   95 |     struct gbm_device *gbm_device {nullptr};
      |            ^
/usr/include/gbm.h:46:8: note: the incompatible type is defined here
   46 | struct gbm_device;
      |        ^
/usr/bin/ld: /tmp/ccmjG09Z.ltrans0.ltrans.o: undefined reference to symbol '_ZNSt8__detail13__notify_implEPKvbRKNS_16__wait_args_baseE@@GLIBCXX_3.4.35'
/usr/bin/ld: /usr/lib/libstdc++.so.6: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
ninja: build stopped: subcommand failed.
==> ERROR: A failure occurred in build().
    Aborting...

Update: I managed to get sunshine-git built by setting _versioned_gcc=true in the PKGBUILD.

It's no longer needed to set cap_sys_admin on the binary, if you do, it breaks the vulkan encoder detection. Also see: https://github.com/LizardByte/Sunshine/pull/5075

This update reworks the PKGBUILD to better match Sunshine's current upstream build system and Arch packaging expectations.

Main changes:

• updated maintainer/contributor metadata
• expanded runtime and build dependencies to cover the current upstream feature set
• added configurable build toggles for tests, headless testing, CUDA, and pinned commits
• simplified source handling by using the main git source with optional commit pinning
• switched submodule handling to a recursive init/update flow
• replaced the old Boost workaround with a smaller patch so the packaged Boost version is preferred instead of forcing an exact upstream version
• cleaned up the build logic and exported upstream version metadata from the actual checked out source state
• added build-time validation for AppStream metadata and desktop files
• made the test step optional but more complete when enabled, including XML results and gcovr coverage output
• kept CUDA support flexible, with autodetection and fallback behavior when CUDA is unavailable
• aligned install/build variables and package metadata with current upstream paths and AUR package naming

Overall, the goal of this patch is to make the package more reproducible, easier to maintain, and closer to upstream while still fitting Arch/AUR conventions.

Maintenance notice (2026-04-05)

A package update is planned for today around 24:00 (midnight, Warsaw time).

The update will include fixes and general improvements. Please expect a brief period where the package may be temporarily out-of-date or rebuilding.

Thank you for your patience.

@mambuco please, write to me on oskar@gerlicz.space if you want to be added to maintainers, today i will send the updated package :)

Hi, I'll update the package as soon as I can, probably in the next 24 hours. I haven't been able to do so in the last days, I'm sorry for ignoring previous comments.

Latest libboost update broke even existing installations:

sunshine: error while loading shared libraries: libboost_locale.so.1.89.0: cannot open shared object file: No such file or directory

Package definitely needs an update.

Initializing the submodules in third-party/build-deps fixes the build:

diff --git a/PKGBUILD b/PKGBUILD
index baed8f2..bf3dc21 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -8,7 +8,7 @@

 _pkgname="sunshine"
 pkgname="$_pkgname-git"
-pkgver=2025.1021.185115.r0.g8df1003
+pkgver=2026.331.130344.r0.gb172a98
 pkgrel=1
 pkgdesc="A self-hosted game stream host for Moonlight"
 url="https://github.com/LizardByte/Sunshine"
@@ -69,6 +69,7 @@ prepare() {

   git submodule update --init --depth 1
   git -C third-party/moonlight-common-c submodule update --init --depth 1
+  git -C third-party/build-deps submodule update --init --depth 1

   ## disable unwanted macros
   sed 's&macro(find_package)&macro(_disable_find_package)&' -i cmake/macros/common.cmake