On Mon, Apr 23, 2012 at 03:34:01PM -0500, Bruce Dubbs wrote:
>
> I'm still in the initial build, but the toochain seems to have done OK. One
> problem with gcc-4.7 is that the tests take a *very* long time. On my system
> which is not really slow, it took over an hour and a half to run the tests.
>
> === libmudflap Summary ===
> # of unexpected failures 1
>
> FAIL: libmudflap.c++/pass55-frag.cxx ( -O) execution test
>
> It's a known problem.
> http://old.nabble.com/-Bug-libmudflap-49843--New%3A-64-bit-libmudflap.c%2B%2B-pass55-frag.cxx-FAILs-at--O-td32132826.html
>
> Only one failure is the best I've ever seen.
>
On svn-20040419 x86_64, with static libraries mostly suppressed, I
saw failures in (alphabetical order, then gcc test summary at the
end) :
automake - instdir, instdir2, insthook, instmany, instspc - might
be my supprression of static libs.
binutils - static .* array .* and failed to link a static
executable, definitely my divergence from the book
glibc [ with -O2, as noted ] - only annex.c : I don't remember such
a good result here for years.
sed - utf8-{1,2,3,4} : again, maybe my lack of static libs
gcc -
17 unexpected failures in gcc, of which 16 were variants of
gcc.c-torture/compile/limits-exprparen.c and 1 was in
gcc.dg/cpp/_Pragma.c
3 unexpected failures in libitm, but none were specifically listed
1 in libmudflap, matching yours (the rest might be lack of static
libs).
People will know that I now have a low regard for the testsuites.
Those who build with static libs can draw their own conclusions.
Summary: glibc looks good (perhaps because I had to build it with
-O2), the rest are par for the course.
Will compare the jh branch somewhen, but Bruce's analysis shows it
certainly looks better for x86_64.
ĸen
--
das eine Mal als Tragödie, das andere Mal als Farce
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
