fix: missing math.h for NAN

[neko-para]

According to cppreference, NAN macro in defined in math.h.

quickjs/quickjs.h

Line 93 in 8ac7c0d

#define JS_FLOAT64_NAN NAN

https://en.cppreference.com/w/cpp/numeric/math/NAN
https://en.cppreference.com/w/c/numeric/math/NAN

[saghul]

Is there a way to test these? Would compiling with g++ fail without this?

[neko-para]

Is there a way to test these? Would compiling with g++ fail without this?

Yes, our CI failed on ubuntu building with gcc.
https://github.com/MaaXYZ/MaaFramework/actions/runs/9386698680/job/25847908600

[neko-para]

godbolt reproduce

[saghul]

Cheers!

[chqrlie]

Simple fix! I'm surprised this problem did not surface before.

Note however that as of c2x, NAN is now also defined in <float.h> as specified in 5.2.4.2.2 Characteristics of floating types <float.h> and the standard says:

Use of the macros INFINITY, DEC_INFINITY, NAN, and DEC_NAN in <math.h> is an obsolescent
feature. Instead, use the same macros in <float.h>.

This suggests including <float.h> instead of <math.h> or at least in addition to <math.h>, but gcc needs the option -std=c2x to allow using NAN with just <float.h> included so just including <math.h> is portable and should remain valid for a few more decades.