alvr-org
reorganise flatpak stuff - and update wiki
do we really need to keep both?...
I am using it to do my test builds... Arch has problem with nvcc and gcc - I prefer to use containerised builds.
Flatpak is a pretty bad option for development. Which is fair considering its mainly intended for distribution anyway. The way better option for dependency issues when developing is nix anyway. Its ridiculously easy to just add a tool to your build environment with it. We could think about shipping additional support for developing with nix, but thats a different issue.
In your case this should be enough to have a shell with a working nvcc: nix-shell -p cudaPackages.cuda_nvcc
Tho I'm not sure whether some other cuda stuff in alvr might not cause issues regardless. I guess I'll try it out when I can
In your case this should be enough to have a shell with a working nvcc:
nix-shell -p cudaPackages.cuda_nvccTho I'm not sure whether some other cuda stuff in alvr might not cause issues regardless. I guess I'll try it out when I can
There is no impact - it's a seperate flatpak for developers only. It is not related to the launcher flatpak that users run.
Someone else created this originally, it was there when I came along. All I did was add cuda to it so nvidia build would work. Why did they choose to do the build in flatpak, rather than copying the binary from outside? I don't know - but it's already in the repo. Flatpak provides convenient mechanisms for building OSS - you can see this being put to use. The caching mechanism etc is quite good.
It turns out that this is not a bad option in this case - because the build is done in exactly the correct runtime environment. And because it's already sharing the same flatpak, it's easy to copy the built binary to the launcher folder for testing.
Building and development are not the same thing. Having a flatpak build does not change anyones development workflow.
So - do not treat this like some big new thing - it has been there for years but only to build the flatpak version of alvr binary (which hardly anyone used).
There is no impact - it's a seperate flatpak for developers only. It is not related to the launcher flatpak that users run.
It's a tool for developers and codebase bloat has an impact for developers too. Plus I think the nix solution solves this exact problem way more elegantly (way less complex, easier to use, takes up less storage).
I don't know - but it's already in the repo.
It was in the repo, it got removed and I think removing it is the right choice and it should be kept that way.
Building and development are not the same thing. Having a flatpak build does not change anyones development workflow.
Building is a part of development. Regardless, I think a solution to improve the issues with cuda on linux for building alvr should have an impact on build workflow for everyone that wants to build with nvidia by making it way easier for them. And in my personal experience flatpak does not fulfill that.