Arch Linux User Repository

@kescherAUR good news, that is now the case.

@abouvier Unfortunately, the prior package maintainer chose to replace the patch version with the number of Git revisions.

I highly disagree with that decision, but I also don't want to bump the package's epoch. I also don't want to break expectations set by versions either.

Therefore, until a release tag after 2.0 is made (2.1, 3.0, ...), I will keep this version as-is. I've made preparations in the PKGBUILD already for an easy switch once this is the case.

Where does this version string come from? It should just be 2.0.94, exactly like the git tag that is fetched.

@HurricanePootis Yes, I agree with this. This is why the former maintainer of this package did this in the first place. Me having to change the prepare function due to git changes was what ended up causing the issue you described.

Currently, I specifically remove the submodules of vcpkg and Vulkan-Headers: https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=cemu&id=e13613b9ae0d98e089b87aff73b9ebdb15f2ad88#n69

I have also validated that the build, after fetching only the source array and nothing else, works completely offline. So yes, after your patch suggestion, which I partially applied (still with fixed commit hashes, however), this is working fine again.

@kescherAUR the point of having all the submodules in the source array is not that we are going to use them. It's that you are going to download them anyways, so it's better to have them in your source() so, for whatever reason, the src/ folder gets nuked, you don't have to spend bandwidth redownloading this.

2.0-91 is just a fix for CI builds - so not relevant to end-users.

@HurricanePootis I see now what a prior comment by @FabioLolix meant - your comment patch is indeed pointing this issue out well. Thanks for that.

I am, however, still not adding vcpkg and Vulkan-Headers to the source array. The main reason for that is: - We genuinely do not need vcpkg here (it's disabled in build()) - We use the vulkan-headers package as a dependency instead of the submodule, and this is intentional.

Additionally, I am keeping the direct references to specific commits, as it allows checksums to be used still.

Still, this is better now. Thanks again.

Hey, you are doing your submodules wrong. I've noticed that although you are cloning them in your source() array, due to how you are trying to do the submodules in your prepare() function, you just end up cloning them again from the internet. I have made the process better with this patch, and also added some missing submodules which should always be cloned:

diff --git a/PKGBUILD b/PKGBUILD
index 0cce902..37c6f5c 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -33,19 +33,23 @@ install=cemu.install
 source=(
    git+https://github.com/cemu-project/Cemu#tag=v2.0-82
    # submodules
-   git+https://github.com/mozilla/cubeb#commit=dc511c6b3597b6384d28949285b9289e009830ea
-   git+https://github.com/ocornut/imgui#commit=8a44c31c95c8e0217f6e1fc814cbbbcca4981f14
-   git+https://github.com/Exzap/ZArchive#commit=d2c717730092c7bf8cbb033b12fd4001b7c4d932
+   git+https://github.com/mozilla/cubeb.git
+   git+https://github.com/microsoft/vcpkg.git
+   git+https://github.com/KhronosGroup/Vulkan-Headers.git
+   git+https://github.com/ocornut/imgui.git
+   git+https://github.com/Exzap/ZArchive
    # cubeb submodules
-   git+https://github.com/arsenm/sanitizers-cmake#commit=aab6948fa863bc1cbe5d0850bc46b9ef02ed4c1a
-   git+https://github.com/google/googletest#commit=800f5422ac9d9e0ad59cd860a2ef3a679588acb4
+   git+https://github.com/arsenm/sanitizers-cmake.git
+   git+https://github.com/google/googletest
 )
 sha256sums=('8302fb4de0bbba1e9f79da2c8badeb6e1f3b146fca35d40f7bbe306d910c3f22'
-            'ad74dba4d6bd055e7862ef6916654e7572cb3601a7882715b0617207b159fc94'
-            'b2f59d28eca6c4d702c845ad82818b5378e9edafff12394f9321bb69e9f80642'
-            '778720b0b08b122af1a958faf5bb1340d44fa33057124373c81822f1e1918e19'
-            'a24b426203153fe25c9678a0ed2335293eec40e4a922a81d61fe0b60987573bc'
-            '2353b7e1877666d242fd45083ae730667fbcdbb418dfaa43b39d33a828ce6e08')
+            'SKIP'
+            'SKIP'
+            'SKIP'
+            'SKIP'
+            'SKIP'
+            'SKIP'
+            'SKIP')

 pkgver() {
    cd Cemu
@@ -60,15 +64,14 @@ prepare() {
    cd Cemu

    # cemu submodules
-   for submodule in dependencies/{cubeb,imgui,ZArchive}; do
-       git config submodule.$submodule.url "file://$srcdir/${submodule##*/}"
-       git submodule--helper update --init $submodule
+   for submodule in {cubeb,imgui,ZArchive,vcpkg,Vulkan-Headers}; do
+       git config --file=.gitmodules submodule.dependencies/$submodule.url "$srcdir/$submodule"
    done
+   git -c protocol.file.allow=always submodule update --init
    pushd dependencies/cubeb > /dev/null
-   for submodule in {cmake/sanitizers-cmake,googletest}; do
-       git config submodule.$submodule.url "file://$srcdir/${submodule##*/}"
-       git submodule--helper update --init $submodule
-   done
+   git config --file=.gitmodules submodule.googletest.url "$srcdir/googletest"
+   git config --file=.gitmodules submodule.cmake/sanitizers-cmake.url "$srcdir/sanitizers-cmake"
+   git -c protocol.file.allow=always submodule update --init
    popd > /dev/null

    # unbundled fmt

@That1Calculator if you mean bumping the version dependency of llvm-libs, yes, that should definitely be at 17 by now

@That1Calculator llvm-libs is already a dep, and the build works without llvm17-libs entirely, so, I don't think so