On Sat, Jan 31, 2009 at 10:58 AM, Diego Novillo <email@example.com> wrote: > HJ, > > I had to revert the patch you committed for PR 39010. It had > caused 478 regressions on x86_64 and 625 on i686. Presumably, > there are other places where we are off by one. > > I noticed that also. However, I believe my patch just exposed some latent LTO bugs. -- H.J.