Joe Ciccone wrote:
>> 5.8.2. Internationalization:
>>
>> a. localedef doesn't seem to work no matter what options used. I tested
>> it with the --prefix option without success.
>>
>> b. Because I wasn't sure if anything downstream in the build process
>> required it, I just used the first instruction to install all locales at the
>> expense of time.
>>
>> c. Perhaps the localedef command could be included later during a chroot
>> stage in the process, or after first boot into the new system.
>>
>>
>I'm guessing you did an x86 build on x86 then. If you're running make
>localedata/install-locales when cross-compiling it should fail because
>the localedata Makefile will run the localedef in locale/. Hopefully
>I'll be able to finish and get my new patch into the book tonight, if
>not tonight it will definitely be tomorrow, it fixes basically all of
>the cross-compiling nightmares with locales/timezoneinfo/sunrpc.
True. It was a cross-compile within the same architecture. The machine is a
pentium-m based laptop (i686-pc-linux-gnu), with the distribution being the LFS
livecd (i486-pc-linux-gnu). I've been wanting to learn how to do a sysroot
cross-compile, this has been a great guide. The other guides and howtos I
studied all stopped at the toolchain.
>> 6.16.1. Installation of Perl:
>>
>> a. I had to unset CFLAGS at this point. The CFLAGS I had set caused perl
>> to go into an endless compile loop during >compilation of miniperl.
>>
>>
>I havn't done a build with CFLAGS set yet. I can imagine that that would
>happen because miniperl is built against the host then the whole package
>is cross-compiled. A big warning box seems in order.
The CFLAGS used were "-O2 -g -pipe -march=pentium-m". Perl was the only
application that complained.
>> 6.42.2. Configuring Shadow:
>>
>> a. The pwconv and grpconv commands won't work unless you either chroot
>> into the new system, or until after first >boot. To opted for the latter
>> this time, but recommend these be placed in a chroot stage.
>>
>> 6.42.3. Setting the root password:
>>
>> a. Same as the comment for 6.42.2 above.
>>
>Chroot stage would only work if you're actually building on the
>architecture you're building for. Does building another set of tools,
>say $CLFS_TARGET-{passwd,pwconv,grpconv} seem like a decent solution? I
>don't know how hard it would be to do, but I think it would be possible.
One of my future projects is to do a cross-architecture cross-compile x86 -->
powerpc. With that said, maybe those three commands could just be relegated to
a post-install-instructions-after-first-boot-stage? That would be the simple
approach.
>> Missing from the text:
>>
>> a. Grub instructions. I just reused my boot partition and added an entry
>> in menu.lst.
>>
>The only reason the book doesn't have a bootloader is because after the
>bootloader is built, how the heck do you install it? maybe with dd.
dd seems the simple way, otherwise your idea above adding it as a tool also?
this is a definite piece to the puzzle of building for a different architecture.
------------------------------
_______________________________________________
Clfs-dev mailing list
[email protected]
http://lists.cross-lfs.org/cgi-bin/mailman/listinfo/clfs-dev
End of Clfs-dev Digest, Vol 5, Issue 9
**************************************
____________________________________________________________________________________
Sponsored Link
Rates near 39yr lows. $510,000 Loan for $1698/mo.
Calcuate new payment. www.LowerMyBills.com/lre
_______________________________________________
Clfs-dev mailing list
[email protected]
http://lists.cross-lfs.org/cgi-bin/mailman/listinfo/clfs-dev
