Comment 2 for bug 500524

I've looked into the history of the underlying problem a bit more and found out:
- in 4.3 and earlier, the problem didn't yet exist
- in 4.5 and later, the problem is fixed
- however, in 4.4, the problem exists as latent wrong-code generation bug

I've suggested on the gcc-patches list that it would be a good idea to backport the fix to 4.4. If they agree, we'd get the fix automatically when moving forward to 4.4.5. Otherwise, we have to decide whether to take the backport into Linaro GCC.