[uClibc] buildroot and AMD processors...

Charles Palmer charles.palmer at acutetechnology.com
Tue May 24 17:32:38 UTC 2005


OK, I think I get it:

If I enable AMD Alchemy DBAu1500 Development Board Support then after 
building cross compliers and uclibc libraries etc, and after building the 
root file system for the target board, the buildroot process then downloads 
a linux 2.4.29 kernel, patches it, builds it etc. This kernel-building 
exercise is being started by this line in the top-level Makefile: "include 
target/*/*.mk"

Is that right?

Questions (please excuse dumb newbie questions):

1(a) If the root file system is built with one set of tools and libraries, 
then must the kernel also be built using the same tools?
  (b)  And if my kernel is built with one set of tools then that explains 
why a root filesystem and hello world programs built with the shiny new 
tools built by buildroot don't work - is that right? (No that can't be 
right, since your pre-built mipsel rootfs works with my kernel. Hmmm.)
  (c)  And is that why you are building the kernel as part of the same 
massive make operation?

2 (a) My kernel source tree (for the DBAu1200 board) differs from what I see 
in the buildroot/build_mipsel/linux-2.4.29-erik (for the DB1500 board), and 
my kernel Makefile differs as well. While it might be nice to bring my 
kernel source and makefile in line with yours, I wonder if it is possible to 
halt buildroot's make operation before it starts on the kernel, and then use 
the new tools built by buildroot to compile my kernel?
  (b) If so, what is the cleanest way of doing this? Maybe deleting this: 
buildroot/target/device/AMD/DBAu1500/linux.mk?
  (c) Or will this idea come unstuck due to the need to rely on files in the 
buildroot directory?

3(a) The buildroot make process still stops for me when processing 
linux-2.4.29-erik/drivers/net/au1000_eth.c with this error: 
"au1000_eth.c:97: sorry, unimplemented: inlining failed in call to 
'str2eaddr': function body not availableau1000_eth.c:". I have tried this 
with Fedora (gcc 3.4.2) and RedHat 9 (gcc 3.2.2). Any clues? (I don't 
actually need to compile this file for the Au1200, but it feels like a 
symptom of some other problem).
(b) So I don't know what the buildroot process will try after it builds the 
kernel. If I build my own kernel, is there anything else I have to do 
afterwards?








More information about the uClibc mailing list