New function emscripten_has_asyncify

[Beuc]

Cf. emscripten-ports/SDL2#70 (comment)

[gabrielcuvillier]

Good idea.

Actually, I can think of a way to already get the info:

emscripten_get_compiler_setting("ASYNCIFY")

It needs -s RETAIN_COMPILER_SETTINGS=1 though

[sbc100]

True, but I wonder what the space cost of RETAIN_COMPILER_SETTINGS is?

[Beuc]

Also emscripten_get_compiler_setting hits a JavaScript error if not compiled-in.
Since we want to use the new function in the SDL2 port by default, that's annoying, and requires an update to existing build systems that use SDL2 outside of the port system (e.g. upstream SDL2 tarballs) so they add -s RETAIN_COMPILER_SETTINGS at link time. Not to mention people might not want to disclose their compiler settings :)
EDIT: size: out.js += 6kB

To a lesser extent the new functions also currently encapsulates ASYNCIFY||EMTERPRETER_ASYNC.

[kripken]

Thanks! Looks good with some minor comments.

And agreed that while RETAIN_COMPILER_SETTINGS can do the same, it's nice to have a direct API for this that doesn't require a new build flag for users. So this makes Asyncify easier for people to use.

[Beuc]

Done!

[kripken]

Perfect, thanks!

[kripken]

Oh wait, tests are failing, that's odd.

Do you see the failures locally as well?

[Beuc]

test_emscripten_has_asyncify works locally for me.
Not sure what's happening :/

[Beuc]

I can reproduce with fastcomp which complains about ASYNCIFY being removed, while we need to set it for the testcase - I disabled the test in fastcomp.

[kripken]

Makes sense to disable it for fastcomp, yeah.