Floating point woes; Different behavior on 32- vs 64-bit x86

I've the following snippet:

x, y, r  :: Double
x = -4.4
y = 2.4999999999999956
r = x * y

Using GHC 7.8.3, on a Mac, I get the following response from ghci:

*Main> decodeFloat r
(-6192449487634421,-49)

Using GHC 7.8.3, on a Linux machine, I get the following response:

*Main> decodeFloat r
(-6192449487634422,-49)

Note that off-by-one difference in the first component of the output.

I'm not 100% sure as to which one is actually correct; but the point is that these are IEEE floating point numbers running on the same architecture (Intel X86), and thus should decode in precisely the same way.

While I observed this with 7.8.3; I don't think this is a new regression; I suspect it'll show up in older versions as well.

Also, for full disclosure: I ran the Mac version naively; but the Linux version on top of a VirtualBox image. I'd very much doubt that would make a difference, but there might be 32/64-bit concerns. So, if someone can validate the Linux output on a true 64-bit Linux machine, that would really help track down the issue further.

Trac metadata

Trac field	Value
Version	7.8.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	high
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture