This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [2.91.34] Compiling linux-2.0.34 ?
- To: bryan at terran dot org (B. James Phillippe)
- Subject: Re: [2.91.34] Compiling linux-2.0.34 ?
- From: Joe Buck <jbuck at synopsys dot com>
- Date: Fri, 5 Jun 98 9:59:18 PDT
- Cc: egcs at cygnus dot com
> I hate to poke another building-linux-with-egcs question on the
> list, but I am counting on this one being unique in one way. :) Now that
> 2.0.34 is "officially" out, I'm wondering if anyone has appraised it's
> compatibility with egcs yet. I am aware that the reasons for miscompiles
> in the 2.0 series are due to code problems in the kernel. I'm wondering
> specifically if a.) these problems are fixed in 2.0.34, and b.) if not, is
> there a patch someplace that I can apply?
Alan Cox (who did most of the work on 2.0.34) said the following a while
back (on April 23):
> The situation is as follows
>
> o Linux 2.0.x builds correctly with gcc 2.7.2 (-1 and -3).
>
> o Linux 2.1.x builds correctly with gcc 2.7.2, 2.8.1 and egcs (a few
> errors keep popping up and people keep squashing them).
>
> I don't think there is anyone who doesn't intend 2.2.0 to work with 2.8.1
> and all the bugs people are finding now with gcc 2.8.1 are Linux not gcc
> bugs where the kernel forgot to stop gcc 2.8.1 optimising something we
> didnt want optimised, or from poor asm constraints on the linux code.
>
> The problem is _nobody_ knows how many of those are still hiding in 2.0.x
> and introducing subtle races, so for safety we say build 2.0.x with
> gcc-2.7.2-*.
>
> So its an argument for what you build kernels with but its not an argument
> to which is the better compiler. Kernel code is just very touchy about
> such things.
In the absence of another statement from him, I think we need to say
that building 2.0.34 with egcs or gcc 2.8.x is not supported (even if,
as I suspect, a couple of the known bugs were fixed).