On Sat, 9 Jun 2001, Herbert Xu wrote: > With gcc-3.0 -O2, it's a lot worse. The last loop where the fault was > observed is totally stuffed, e.g., the loop condition is now the first thing > that gets tested after (disk <= 0). > > -O looks alright. Argh! Can you file a bug and include the generated instructions? It would be great to get that info into gcc's GNATS so we can get resolution into 3.0.x. Thanks! C