From: Larry Martin <Larry@GlueLogix.com> - 2008-10-21 13:09:31
The problems with 1631 were mostly me, not seeing how much had changed
in the buildroot since I last worked with it. For the record, this is
what I had to do to 1631 on FC9:
1. Shorten all paths by running svn co in my home directory. I did not
explore to see what the limit was here, but the compile had issues in
/home/larry/Gumstix/br/Verdex1508/gumstix-buildroot. Worked great in
2. In the buildroot base, run 'chown -R me:me *'. I found that some
files in 1508 have the wrong owner. Don't know if this was a problem in
1631, or what the implications might have been.
3. In netkit telnetd utility.c, provide a #define for ISSUE_FILE
4. In toolchain/uClibc, run 'ln -s uClibc-verdex.config uClibc.config'
5. In 'make menuconfig', Select Verdex (IWMMXT) as the Arm Variant. It
looks like my 1508 tree was running with the generic ARM variant. I
think this was probably the source of the header file issues I reported
with 1631 yesterday.
6. As described in the buildroot wiki, select IWMMXT in the uClibc
config menu. I also answered Yes to CONFIG_ARM_EABI and USE_BX, and
have my fingers crossed.
The buildroot howto does not mention any of the above except the first
part of #6.
On the upside, Buildroot 1631 accepted my makeinfo 4.11 without any hacking.
If anyone is still reading, is it worth updating the buildroot wiki at
> What was the last known-good buildroot? I have been fussing with 1631
> most of the day. I don't know if it's me or what, but there are many,
> many missing header files, like asm/unistd.h and linux/version.h. I
> fixed up 4 or 5 of them from an earlier tree, but am starting to think
> there is something seriously wrong with 1631.
> For the record, I am following the instructions on:
> My system is Fedora 9 and previously built 1508 with no problems.
Get latest updates about Open Source Projects, Conferences and News.