From: Alan Cox <[EMAIL PROTECTED]> > Julie wrote: > > I have to agree with Andi. I've not encountered anything in any of the > > specs which require that the effective UID cannot change during a > > You misunderstood the original comments. Linux kernels assume that > current->euid, current->uid, current->fsuid etc are thread private. That > avoids massive amounts of locking and the use of atomic types for these > variables.
No, I understood it perfectly. There's a difference between "Oops, we made a mistake" and "No, we really are following the standard." The Linux threading model violates the standard. So the correct answer is "Oops, we made a mistake." -- Julie.
