[HXCPP][XCODE16.0] thx.bigint.Bigs.canAdd(2000000000, 400000000); returns true wrongly.

[mousecute]

Source code is:

  public static function canAdd(a : Int, b : Int) {
    var v = a + b;
    if (a > 0 && b > 0 && v < 0)
      return false;
    return isPrecise(v);
  }

The generated cpp code is:

HXCPP_OPTIMIZATION_HINT_METHOD bool Bigs_obj::canAdd(int a,int b){
                HX_STACKFRAME(&_hx_pos_0c9c58070269a9c2_49_canAdd)
HXLINE(  50)		int v = (a + b);
HXLINE(  51)		bool _hx_tmp;
HXDLIN(  51)		bool _hx_tmp1;
HXDLIN(  51)		if ((a > 0)) {
HXLINE(  51)			_hx_tmp1 = (b > 0);
                    }
                    else {
HXLINE(  51)			_hx_tmp1 = false;
                    }
HXDLIN(  51)		if (_hx_tmp1) {
HXLINE(  51)			_hx_tmp = (v < 0);
                    }
                    else {
HXLINE(  51)			_hx_tmp = false;
                    }
HXDLIN(  51)		if (_hx_tmp) {
HXLINE(  52)			return false;
                    }
HXLINE(  53)		return ::thx::bigint::Bigs_obj::isPrecise(v);
                }

where HXCPP_OPTIMIZATION_HINT_METHOD is __attribute__((minsize))

assembly code:

thx::bigint::Bigs_obj::canAdd:
    0x10872ac04 <+0>:  sub    sp, sp, #0x40
    0x10872ac08 <+4>:  stp    x20, x19, [sp, #0x20]
    0x10872ac0c <+8>:  stp    x29, x30, [sp, #0x30]
    0x10872ac10 <+12>: add    x29, sp, #0x30
    0x10872ac14 <+16>: mov    x19, x1
    0x10872ac18 <+20>: mov    x20, x0
    0x10872ac1c <+24>: adrp   x1, 7446
    0x10872ac20 <+28>: add    x1, x1, #0x8b0            ; FBLink_UIPasteboard_FBSDKPasteboard + 3572664
    0x10872ac24 <+32>: bl     0x107dcaacc               ; _OUTLINED_FUNCTION_5
    0x10872ac28 <+36>: add    w0, w19, w20
->  0x10872ac2c <+40>: mov    w8, #0x35                 ; =53 
    0x10872ac30 <+44>: str    w8, [sp, #0x18]
    0x10872ac34 <+48>: bl     0x10872a78c               ; thx::bigint::Bigs_obj::isPrecise at Bigs.cpp:136
    0x10872ac38 <+52>: mov    x19, x0
    0x10872ac3c <+56>: bl     0x107dcaa84               ; _OUTLINED_FUNCTION_1
    0x10872ac40 <+60>: mov    x0, x19
    0x10872ac44 <+64>: ldp    x29, x30, [sp, #0x30]
    0x10872ac48 <+68>: b      0x107dcaadc               ; _OUTLINED_FUNCTION_7
    0x10872ac4c <+72>: mov    x19, x0
    0x10872ac50 <+76>: bl     0x107dcaa84               ; _OUTLINED_FUNCTION_1
    0x10872ac54 <+80>: bl     0x107dcaad4               ; _OUTLINED_FUNCTION_6

My guess is that Xcode compiler considers (a > 0 && b > 0 && v < 0) without overflow and optimizes it away. My mitigation is to use Int64 check overflow:

  public static function canAdd(a : Int, b : Int) {
    if (a > 0 && b > 0 && haxe.Int64.ofInt(a) + haxe.Int64.ofInt(b) > haxe.Int64.ofInt(2147483647))
      return false;
    return isPrecise(a + b);
  }

[fponticelli]

Please provide a PR and I will gladly merge ;)