Thomas Schwinge, le Tue 06 Sep 2011 01:56:59 +0200, a écrit : > On Thu, 01 Sep 2011 01:02:26 +0000, Samuel Thibault > <samuel.thiba...@ens-lyon.org> wrote: > > commit 47a835d22b1a608fac5ee241474f128f64cb6a0f > > Author: Samuel Thibault <samuel.thiba...@ens-lyon.org> > > Date: Thu Sep 1 03:00:26 2011 +0200 > > > > Close kernel stacks. > > > > * i386/i386/pcb.c (stack_attach): Initialize ebp to 0. > > Is this another frame pointer than the one that is cleaned every time in > i386/i386/cswitch.S:Thread_continue?
Mmm, good question. I hadn't seen that ebp was cleared there. The odd thing is that I did get a case where the backtrace was not bound by the 0 value, and it was apparently solved by initializing the value in stack_attach along the esp initialization, which to me was making more sense than setting it in some assembly code. > (If yes, can we drop the assembly, or does it have another purpose?) Thread_continue is only used in stack_attach, so it must be redundant now. Samuel
