Subject: Re: Error in ip_nat.c when building -current
To: Tobias Nygren <tnn@NetBSD.org>
From: Raymond Meyer <raymond.meyer@rambler.ru>
List: current-users
Date: 05/22/2007 10:35:54
On Tue, 22 May 2007 11:11:31 +0200
Tobias Nygren <tnn@NetBSD.org> wrote:

> Beware of building kernel code with -O3.
> GCC's frivolous inlining has been known to cause problems in the past.
> But if you're determined to try, explicitly initialising the variables
> is a valid fix. This is just GCC guessing wrong due to it's loop
> unswitching.

How do those problems manifest themselves? Do you get compilation errors, or is
it possible with -O3 optimisation for kernel to compile OK, but then cause
panics during run time?