The error is likely trigger by these llvm commits.

------------------------------------------------------------------------
r122189 | lattner | 2010-12-19 12:19:20 -0800 (Sun, 19 Dec 2010) | 3 lines

now that generic vector types aren't selected onto MMX operations,
we don't need -disable-mmx anymore.

------------------------------------------------------------------------
r122188 | lattner | 2010-12-19 12:12:58 -0800 (Sun, 19 Dec 2010) | 3 lines

now that generic vector types aren't selected onto MMX registers, these
tests don't need -disable-mmx.

commit 492afbce18af21ce5bbc89675c797e45a0b72bf5
Author: Vinson Lee <vlee@vmware.com>
Date:   Wed Dec 22 19:44:00 2010 -0800

    gallivm: Disable MMX-disabling code on llvm-2.9.
    
    The disable-mmx option was removed in llvm-2.9svn by revisions 122188
    and 122189.
    
    Fixes FDO bug 32564.

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
index f56ddee..46dd00d 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
+++ b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
@@ -144,6 +144,7 @@ lp_set_target_options(void)
    llvm::UnsafeFPMath = true;
 #endif
 
+#if HAVE_LLVM < 0x0209
    /*
     * LLVM will generate MMX instructions for vectors <= 64 bits, leading to
     * innefficient code, and in 32bit systems, to the corruption of the FPU
@@ -162,6 +163,7 @@ lp_set_target_options(void)
       llvm::cl::ParseCommandLineOptions(2, const_cast<char**>(options));
       first = FALSE;
    }
+#endif
 
    /*
     * By default LLVM adds a signal handler to output a pretty stack trace.

mesa: 97195d04fd3d07c15dc8f22707c511f0798cfbc9 (master)

Verified fixed with llvm-2.9svn.

Thanks Vinson.

As llvm-2.9 has been released I propose to merge this into the 7.10-branch.

My X server crashed after my upgrade today and it took me quite a while to figure out what's going wrong. Gentoo already uses this patch in their 7.10.2-r1 ebuild.

I've cherry-picked the relevant commit into 7.10.

Thanks.