Re: gcc64 float bug