On 7/25/12 6:58 AM, Andrew Bradford wrote:
On Tue, 24 Jul 2012 22:12:03 -0500
Ron McDowell<[email protected]> wrote:
Starting with an empty $CLFS, and with this env:
[clfs@d430(Linux)]> env
CLFS_HOST=i486-cross-linux-gnu
TERM=xterm-color
CLFS_TARGET=armv4t-unknown-linux-uclibceabi
CLFS_FPU=
CLFS_ARM_ARCH=armv4t
CLFS_ARM_MODE=arm
CLFS=/mnt/clfs
PATH=/mnt/clfs/cross-tools/bin:/bin:/usr/bin
PWD=/mnt/clfs/sources/Embed
CLFS_ARCH=arm
EDITOR=/usr/bin/vi
OTYP=Linux
PS1=\w\n[\u@\h(Linux)]>
HOME=/home/clfs
SHLVL=1
CLFS_ABI=aapcs-linux
CLFS_ENDIAN=little
CLFS_FLOAT=soft
HISTFILE=/home/clfs/.sh_history
_=/usr/bin/env
Ok, that looks good.
I strongly recommend you start by rebuilding binutils and gcc again
now that you've changed your env. Else I think you'll run into more
troubles, even with what I say later.
I did. $CLFS was empty to start with.
It still craps out in the first pass at gcc-4.6.1...
I think you mean uClibc, right?
Oops, yeah.
+ sed -i s/CONFIG_ARM_OABI/CONFIG_ARM_EABI/g .config
+ make oldconfig
MKDIR include/config
HOSTCC-o extra/config/conf.o
GEN extra/config/zconf.tab.c
GEN extra/config/lex.zconf.c
GEN extra/config/zconf.hash.c
HOSTCC-o extra/config/zconf.tab.o
HOSTCC extra/config/conf
./.config:39:warning: override: reassigning to symbol CONFIG_ARM_EABI
*
* Restart config...
*
*
* Target Architecture Features and Options
*
Target ABI
1. OABI (CONFIG_ARM_OABI) (NEW)
> 2. EABI (CONFIG_ARM_EABI)
choice[1-2?]: Target Processor Type
> 1. Generic Arm (CONFIG_GENERIC_ARM)
2. Arm 610 (CONFIG_ARM610)
3. Arm 710 (CONFIG_ARM710)
4. Arm 7TDMI (CONFIG_ARM7TDMI)
5. Arm 720T (CONFIG_ARM720T)
6. Arm 920T (CONFIG_ARM920T)
7. Arm 922T (CONFIG_ARM922T)
8. Arm 926T (CONFIG_ARM926T)
9. Arm 10T (CONFIG_ARM10T)
10. Arm 1136JF-S (CONFIG_ARM1136JF_S)
11. Arm 1176JZ-S (CONFIG_ARM1176JZ_S)
12. Arm 1176JZF-S (CONFIG_ARM1176JZF_S)
13. Arm Cortex-M3 (CONFIG_ARM_CORTEX_M3)
14. Arm Cortex-M1 (CONFIG_ARM_CORTEX_M1)
15. Intel StrongArm SA-110 (CONFIG_ARM_SA110)
16. Intel StrongArm SA-1100 (CONFIG_ARM_SA1100)
17. Intel Xscale (CONFIG_ARM_XSCALE)
18. Intel Xscale With WMMX PXA27x (CONFIG_ARM_IWMMXT)
choice[1-18?]: 1
*
* Using ELF file format
*
Target Processor Endianness
1. Big Endian (ARCH_WANTS_BIG_ENDIAN)
> 2. Little Endian (ARCH_WANTS_LITTLE_ENDIAN)
choice[1-2?]: 2
Target CPU has a memory management unit (MMU) (ARCH_HAS_MMU) [Y/n/?] y
Do you want to utilize the MMU? (ARCH_USE_MMU) [Y/n/?] y
Enable floating point number support (UCLIBC_HAS_FLOATS) [Y/n/?] y
Target CPU has a floating point unit (FPU) (UCLIBC_HAS_FPU) [Y/n/?] y
Make sure you have soft float ability in your kernel because of this
answer.
Have tried y and n in there.
Enable full C99 math library support (DO_C99_MATH) [N/y/?] n
Enable XSI math extensions to the ISO C standard (bessel)
(DO_XSI_MATH) [N/y/?] n
Enable C99 Floating-point environment (UCLIBC_HAS_FENV) [N/y/?] n
Linux kernel header location (KERNEL_HEADERS) [/usr/include] /usr/include
#
# configuration written to ./.config
#
+ make
MKDIR include/bits
GEN include/bits/uClibc_config.h
LN include/fpu_control.h
LN include/dl-osinfo.h
LN include/hp-timing.h
LN include/bits/byteswap-common.h
...
LN include/sys/elf.h
LN include/sys/io.h
LN include/sys/procfs.h
LN include/sys/ucontext.h
LN include/sys/user.h
MKDIR libm/arm
GEN include/bits/sysnum.h
cc1: error: unrecognized command line option "-mlittle-endian"
cc1: error: unrecognized command line option "-mlittle-endian"
ERROR: Could not generate syscalls.
Make sure that you have proper kernel headers.
Your .config in KERNEL_HEADERS="" was set to:
/usr/include/
make: *** [include/bits/sysnum.h] Error 1
+ exit 1
Sorry about that... It's my fault. The uClibc-0.9.32 configs patch I
made forgot 2 rather important things, the KERNEL_HEADERS directory and
the CROSS_COMPILER_PREFIX. Get a (slightly) better uClibc-0.9.32
configs patch from [1].
[1]:http://git.cross-lfs.org/?p=abradford/clfs-embedded.git;a=blob_plain;f=patches/uClibc-0.9.32-configs-3.patch;h=e4671aa7af78b76b1764168e6628513c922a7a22;hb=HEAD
> From a quick look through that uClibc configs patch, though, I think
there's still quite a bit that's left in states that aren't ideal.
For example, multithreading and the multilib changes with the RUNTIME
and DEVEL_PREFIX settings. I strongly suggest you use menuconfig to set
up uClibc how you want and don't blindly trust my configs patch, yet, it
still needs some work, or go back to 0.9.31.
You may end up hitting [2] with 0.9.32. I'm now looking into that. It
looks like building gcc with --disable-shared caused libgcc_eh.a to not
be built, but uClibc wants it.
I did run into that.
[2]: http://permalink.gmane.org/gmane.comp.gcc.cross-compiling/14074
I think I'm headed down a rabbit hole with this one...
I suggest, if you want to build uClibc, use 0.9.31 and apply the patch
to get around the .size directive issue [3]. Use the configs patch
from the current stable book. GCC 4.6.1 should be fine, though, and
avoid needing any patches other than the softfloat one.
I run into the libgcc_eh.a roadblock this way, too.
[3]:http://git.uclibc.org/uClibc/commit/?id=3e68c52b941636714d2599ea8adda9520ec2de23
uClibc 0.9.33 is out now, maybe they've fixed up some of these issues
there or have some documentation on getting around this stuff. I just
haven't had time to really pay attention to everything clfs related.
Sorry. Maybe I'll find a day soon...
-Andrew
Yay!! Success!!! At least this far... Using linux-2.6.38.2,
binutils-2.22, and gcc-4.6.1... uClibc-0.9.33 built exactly as described
in CLFS-EMBEDDED-GIT-0.0.1-20110514/arm/cross-tools/uclibc.html, using
the existing uClibc-0.9.31-configs-2.patch. Onward through the fog! :)
--
Ron McDowell
San Antonio TX
_______________________________________________
Clfs-support mailing list
[email protected]
http://lists.cross-lfs.org/listinfo.cgi/clfs-support-cross-lfs.org
