On 12/15/2015 5:23 PM, Jean Delvare wrote: > Looks like something different from what we were discussing here. In which way? That patch fixes pointer aliasing in the same way yours breaks it, AFAICT? - Derek