Change artifact folder structure

[almarklein]

Closes #410

This updates the package command in the makefile to introduce a folder structure instead of putting all files in a flat namespace. The idea is to prepare the contents of the archive in a directory and then zipping that dir as a whole.

Results can be seen here: https://github.com/almarklein/wgpu-native/actions/runs/10703101259 (can others access these?)

One example:

[almarklein]

cc @Korijn

I don't have a lot of experience with makefiles and shell scripts, so I would not be surprised if there are better ways to do things. E.g. I hate the cd ../.., but popd was not available on one of the builds (IIRC the android build).

[rajveermalviya]

LGTM! Feel free to merge.

Though just noticed that ios archives (and the action name) have prefix iOS instead of ios making it inconsistent with other builds, would be good to have that fixed.

[almarklein]

Though just noticed that ios archives (and the action name) have prefix iOS instead of ios making it inconsistent with other builds, would be good to have that fixed.

Will do that in a new pr

[almarklein]

Oh, I misunderstood your comment. Naming the archive to be lowercase is something that makes sense doing in this pr :)

[almarklein]

New builds on my fork: https://github.com/almarklein/wgpu-native/actions/runs/10717624628

[almarklein]

https://github.com/almarklein/wgpu-native/actions/runs/10722743457

[Sahnvour]

Hi, i'm sorry if I didn't make it clear in the issue, but an intermediate webgpu/ folder is missing in this case to be entirely useful.

In meta-build systems (cmake, meson, conan...) we'd want to do something like
add_include_path('path/to/wgpu-unzipped/include')
where include/ still contains a webgpu/ subfolder, so that other libraries depending on it can do #include <webgpu/webgpu.h>.

And if for any reason the project consuming wegpu wanted to use #include <webgpu.h> instead, they could do add_include_path('path/to/wgpu-unzipped/include/webgpu'): this way either option can be made to work.

[almarklein]

[...] where include/ still contains a webgpu/ subfolder, so that other libraries depending on it can do #include <webgpu/webgpu.h>.

For my understanding, why the need for the subfolder? The name "webgpu.h" seems unique enough by itself. I.e. can't these simply use #include <webgpu.h>?

[Sahnvour]

For my understanding, why the need for the subfolder? The name "webgpu.h" seems unique enough by itself. I.e. can't these simply use #include <webgpu.h>?

They can, but apparently the consensus is to use <webgpu/webgpu.h> in widely used libraries (that are often consumed via package managers and can't be manually edited by clients), see for example:
https://github.com/ocornut/imgui/blob/67cd4ead6505b3386d4ef9c713c98546e86e26ca/backends/imgui_impl_wgpu.h#L21
https://github.com/floooh/sokol/blob/1eb96dd0f96b9ea73065f9078244c2255c2b75d9/sokol_gfx.h#L4480