 | | | kernel build problem | kernel build problem 2002-05-14 - By D. Hugh Redelmeier
Back
| From: Michael Schwendt <msrh@(protected) >
| With regard to the comment by Thomas Molina, I don 't do the "make
| clean ", but that shouldn 't matter, because when I exit "make
| menuconfig " or "make xconfig ", the config file is correct already.
I agree. He may be right that there is a bug there, but it isn 't what
I 'm tripping over.
| Usually I do:
|
| make mrproper
| cp configs/kernel-2.4.18-athlon.config .config
| make menuconfig
| [just save & exit]
| make dep bzImage modules modules_install
|
| You 've found a bug. The real difference is that you load a config
| file from within "make xconfig ". Most people including me don 't do
| that, but create the .config before running either "make oldconfig ",
| "make menuconfig ", or "make xconfig ".
|
| I can reproduce your problem when I run:
|
| make mrproper
| make xconfig
| [load "configs/kernel-2.4.18-athlon.config "]
| [save & exit]
| make dep bzImage
|
| This gives both zlib inflate and deflate as module only,
| which breaks the kernel build.
Yes. Very interesting. I 've poked at this some more.
Just running xconfig a second time and only save-and-exiting solves
my problem!
Theory: if a config comes in through .config, the settings become
correct.
Evidence (from Michael, but replicate by me):
cp configs/kernel-2.4.18-athlon.config .config
make xconfig
save and exit
[.config will be good]
Evidence:
rm .config,
make xconfig
load configs/kernel-2.4.18-athlon.config
save and exit
[.config will be bad]
make xconfig
save and exit
[.config will be good]
Theory: even if a config gets loaded without .config, if you visit the
right menus, the settings become correct.
Evidence:
rm .config,
make xconfig
load configs/kernel-2.4.18-athlon.config
visit "Code maturity level options "; cancel out
save and exit
[.config will be good]
Evidence:
rm .config,
make xconfig
load configs/kernel-2.4.18-athlon.config
visit "Mulitimedia devices "; cancel out
save and exit
[.config will be half good: it differs from the good for by
CONFIG_RAMFS being y, not m and not having a
CONFIG_HIGHMEM_EMULATION comment]
Theory: if the initial settings come from .config, even loading a
config file doesn 't get you to a bad state:
Evidence:
cp configs/kernel-2.4.18-athlon.config .config
make xconfig
load configs/kernel-2.4.18-athlon.config
save and exit
[.config will be good]
This last observation surprises me: I infer that 'Load Configuration
from File " does not reset all settings. If you want to "load from
zero " you must erase .config before running xconfig. This feels like
a bug, but I haven 't seen documentation so it is hard to say
GLobal theory: the code that implements constraints isn 't run in all
the places that it should be.
Thanks for your help,
Hugh Redelmeier
hugh@(protected) voice: +1 416 482-8253
PS: I 'm trying to record this stuff in
<
http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=64814
>
|
|
|