fansh> a:=-1.0f
-1.0f
fansh> a*=0.0f
-0.0f
fansh> (a*=0.0f).bits
-9223372036854775808
fansh> Float.makeBits( (a*=0.0f).bits )
-0.0f
fansh> Float.makeBits(-9223372036854775808)
ERROR(36) : Invalid numeric literal '9223372036854775808'

So why is this happening?! It connects to that old can't reproduce the negative zero bug (I can't remember which ticket that was).

It has nothing to do with the negative zero bug.

9223372036854775808 is simply not a valid numerical literal in Fantom.

a := -9223372036854775808

produces the same error.

Yes, int64 included in the closed range [-9223372036854775808~9223372036854775807]. So -9223372036854775808 and 9223372036854775807 should be valid integer.

And -9223372036854775808 bit is 10000000000000000000000000000000000000000000000... same as -0f (the complement of float 0)

@helium: I know it isn't a valid literal, but as go4 said, it is the same as Int.minVal. So why isn't Int.minVal a valid Int literal (while Int.maxVal is)?

The problem is that -xxxx isn't a literal itself, but is parsed as -(xxxx), and the negative is handled later during constant folding. And in this case the actual positive value isn't a valid integer literal because it is one too big. So this happens to be one case where that prevents you from using a Fantom literal within the valid 64-bit range. Sort of annoying, but seems like such a small issue, that I didn't want to complicate the compiler with a special case.

I see. Is this problem Java-specific or will it crop up on other platforms as well?