Bug 715 – incorrect IEEE 754 handling of -0i and +0i

# # union F{ # float f; # uint i; # } # # union I{ # ifloat f; # uint i; # } # # void foo(){ # // success with "float" # F f; # f.f = 0.0f; # f.f *= -1.0f; # assert(f.i == 0x80000000); # # // failure with "ifloat" # I i; # i.f = 0.0fi; # i.f *= -1.0fi; # assert(i.i == 0x80000000); # } # test cases: http://dstress.kuehne.cn/run/ieee_754_zerosdq_01.d http://dstress.kuehne.cn/run/ieee_754_zerosdq_02.d http://dstress.kuehne.cn/run/ieee_754_zerosdq_03.d http://dstress.kuehne.cn/run/ieee_754_zerosdq_04.d

This creates a real, not an imaginary. It shouldn't compile at all. > # union I{ > # ifloat f; > # uint i; > # } > # > # void foo(){ > # I i; > # i.f = 0.0fi; > # i.f *= -1.0fi; << This should be i.f *= -1.0f; > # assert(i.i == 0x80000000); > # }

Thanks for catching this. I've fixed the test cases and refiled the issue as #730