CI: macos ninja setup failure

[sezero]

We use turtlesec-no/get-ninja for installing ninja to CI host and it is failing on current macOS 14 hosts. Upstream issue (where someone made a fork with a workaround): turtlesec-no/get-ninja#4

This is so with SDL_image, SDL_mixer, SDL_net and sdl2-compat. On the other hand, we use ashutoshvarma/setup-ninja in SDL_ttf and SDL_rtf, and it doesn't seem to be failing.

Should we switch to using ashutoshvarma/setup-ninja for SDL_image, SDL_mixer, SDL_net and sdl2-compat? Or should we switch to using the fork of turtlesec-no/get-ninja? OR: Why aren't we just doing a simple brew install for it?

CC: @madebr

[madebr]

The fork is much simpler and has less dependencies.
Let's make the switch.

[sezero]

OK then. We should do the same for all satellite libs and sdl2-compat, as well.

[sezero]

Same issue on SDL2 and release branches of satellite libs, btw.

[madebr]

I went around all satellite libraries and switched the action everywhere, meanwhile also fixing the dylib current/compatibility version <=> soversion thing.

There are 2 macos/arm64 issues left:

• the build-time SDL_image test fails (because of the relative recent libjxl)
• vendored libmpg123 fails to build. Perhaps we need to bump the vendored version.

[sezero]

vendored libmpg123 fails to build. Perhaps we need to bump the vendored version.

Before blindly bumping it, I looked at optimize.h line 198, it errors if REAL_IS_FIXED is defined, because cmake_host_system_information isn't setting HAVE_FPU? Why?

[madebr]

Looks like a CMake bug to me.
https://github.com/madebr/throwaway/actions/runs/8915872718/job/24486263066#step:3:11

I created an upstream bug report.

The mpg123 cmake script already contains an exception for MSVC, so it looks like the test has also failed there in the past.

[sezero]

OK, so we should add an APPLE exception then?

[madebr]

Yeah, I'm looking into creating a patch for current main atm.
It should be easy to backport.

[madebr]

Done, and also applied to mpg123 of SDL2.

[sezero]

macos/arm64 issues left:

the build-time SDL_image test fails (because of the relative recent libjxl)

The CI log says brew is installing version 0.10.2 of libjxl. Do we have any other
runners that use system libjxl and at least with version >=0.10.0 ?

P.S.: The libjxl + macos issue should be noted in a new ticket?

[madebr]

The CI log says brew is installing version 0.10.2 of libjxl. Do we have any other runners that use system libjxl and at least with version >=0.10.0 ?

P.S.: The libjxl + macos issue should be noted in a new ticket?

Yes, a new ticket should be created.

Only toolchain that has libjxl>=0.10 is msys2.

edit: their website mentions 0.10.2, but our runners only use 0.9.1.

[sezero]

Created #450

Closing this one.