1 |
----- Original Message ----- |
2 |
From: "Mike Frysinger" <vapier@g.o> |
3 |
To: <gentoo-dev@l.g.o> |
4 |
Sent: Tuesday, April 20, 2004 8:19 PM |
5 |
Subject: Re: [gentoo-dev] some portage related issues |
6 |
|
7 |
|
8 |
> On Tuesday 20 April 2004 01:14 pm, Dan Podeanu wrote: |
9 |
> > 1. Kernel compilation and gcc |
10 |
> > |
11 |
> > The recommended gcc version for kernel compile is gcc 2.95.3. Using |
12 |
emerge |
13 |
> > to get any of the supported kernel packages should have a dependency set |
14 |
on |
15 |
> > 2.95.3 and a message informing the users to use gcc-config to change the |
16 |
> > gcc version while compiling the kernel (great tool gcc-config, btw.) |
17 |
> |
18 |
> theres no way this is ever going to happen, end of story. |
19 |
> you want gcc-2.x on your system for some reason, emerge it. |
20 |
|
21 |
The standard Gentoo installation requires one to recompile the linux kernel. |
22 |
A kernel WILL NOT be 100% stable if its compiled with a 3.3.x gcc, although |
23 |
this will be only really visible in certain conditions (higher traffic in |
24 |
terms of requests, cpu load, etc. generally server usage). |
25 |
|
26 |
# grep 2.95.3 /usr/src/linux-2.4.24-xfs-r5/README |
27 |
- Make sure you have gcc 2.95.3 available. gcc 2.91.66 (egcs-1.1.2) may |
28 |
|
29 |
Please note, not 2.95.3 and above (re: Greg KH). And again, compiled with |
30 |
3.3.2 the kernel behaves strange, as opposed to when its compiled with |
31 |
2.95.3. But thats just personal experience. |
32 |
|
33 |
> > 4. It would be really nice if we had a tool to automate the detection of |
34 |
> > the build architecture (CHOST, -march & -mcpu) based on the info from |
35 |
> > /proc/cpu to help the newbie with the installation procedure. |
36 |
> |
37 |
> considering the *large* number of /proc/cpuinfo and the fact that one set |
38 |
does |
39 |
> not map properly onto the other set, this is impossible to a degree |
40 |
> (-march/-mcpu) |
41 |
> |
42 |
> in terms of CHOST, simply picking the right stage tarball would be enough |
43 |
... |
44 |
> most people shouldnt have to change it |
45 |
|
46 |
Atleast for x86 there's a fairly small number of /proc/cpuinfo layouts. I |
47 |
cannot speak for other architectures though. And its all in the kernel |
48 |
sources, imho its fairly simple to write an automated detection system. |
49 |
|
50 |
CHOST will probably be changed by all those using stage1. |
51 |
|
52 |
|
53 |
-- |
54 |
gentoo-dev@g.o mailing list |