@bilelmoussaoui could you please take a look?

Isn't it the same one?

Isn't it the same one?

@BigmenPixel0 No. By default only the host architecture (x86_64 = amd64) is built. With these changes, this will also create an image for ARM (arm64) architecture.

This does not work since the binfmt tooling needs to be run at runtime, so the setup-qemu-action still needs to be used to load/register binfmt.

This does not work since the binfmt tooling needs to be run at runtime, so the setup-qemu-action still needs to be used to load/register binfmt.

What do you mean by "it will not work"? I built the images in my branch and am running them on an arm64 runner without problems.

This builds an arm64 images with all binaries using the arm64 version, so there is no need for binfmt on an arm64 runner.

Sorry I misunderstood the PR, so it concerns a non-"wildly available" type of GitHub runner.

Sorry I misunderstood the PR, so it concerns a non-"wildly available" type of GitHub runner.

Yes but it makes a huge difference. On https://github.com/nanu-c/axolotl this brings down the Flatpak build for arm64 from ~180min to ~20min (~14x speedup).

@bilelmoussaoui @barthalion could you please take a look? This is working for me since quite some time, would be great to make this available for others as well.

Applied it at flathub-infra@79d4438.

@barthalion The build in flathub-infra failed as it uses a different base image than this repository which is not available for arm64 yet.
I created flathub-infra/flatpak-builder-lint#298 to add this.