x86_64 multilib system

Casey Marshall cmarshall at pacificbiosciences.com
Sat Feb 23 01:42:59 UTC 2008


On Fri, 2008-02-22 at 00:42 +0200, Paul Sokolovsky wrote:
> Hello,
> 
> On Thu, 21 Feb 2008 22:23:20 +0100
> "Kees Jongenburger" <kees.jongenburger at gmail.com> wrote:
> 
> > Hello Casey,
> > 
> > I played with getting mamona to compile "native" on my amd64 machine
> > (I was quite far but gave up at one point). Here a some notes that
> > might or might not be helpfull
> > they where certainly not checked against somebody with positive OE
> > karma. but who knows
> 
> Thanks for the patches below, but for full use from them they'd rather
> be submitted to bugtracker where they have more chance to improve
> situation.
> 

I've made a little progress.

Getting a multilib build looks extremely difficult -- GCC does helpful
things like not looking in lib64 for 64-bit libraries, but *also* does
helpful things like installing stuff into lib64, even with
--disable-multilib. Compiling GCC would be funny if it wasn't so tragic.

I have managed to get up to a full install of gcc-cross, with the
attached patches (these aren't yet for wider consumption, they're just
my hacks as of today -- if this works for building more I'll clean them
up and post them properly). The key change here is to patch GCC to fake
"../lib" to be both the 64-bit and 32-bit library locations -- that
seems to be the ultimate way to convince GCC to avoid using the lib64
abomination.

For this build I'm forcing GCC version 4.2.2, glibc version 2.5, and
kernel 2.6.23.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: oe-for-x86_64.patch
Type: text/x-patch
Size: 7769 bytes
Desc: not available
URL: <http://lists.openembedded.org/pipermail/openembedded-users/attachments/20080222/edbecb8b/attachment-0002.bin>


More information about the Openembedded-users mailing list