Followup to: Chapter 6 binutils 'make check' failures

Anthony Borla ajborla at bigpond.com
Sun Feb 12 04:28:09 PST 2006


----- Original Message -----
From: "Ken Moffat" <ken at linuxfromscratch.org>
To: "Anthony Borla" <ajborla at bigpond.com>;
"LFS Support List" <lfs-support at linuxfromscratch.org>
Sent: Sunday, February 12, 2006 10:01 PM
Subject: Re: Followup to: Chapter 6 binutils 'make check' failures

Ken,

>
> On Sun, 12 Feb 2006, Anthony Borla wrote:
>
> > I've had a quick look at the 'ld.log' file and the tests
> > which have failed consist entirely of 'visibility' tests.
> > Some of these pass and some of these fail - a quick
> > example:
> >
> >    ...
> >    tmpdir/vnp >tmpdir/vnp.out
> >    diff tmpdir/vnp.out
> > /binutils-2.16.1/ld/testsuite/ld-elfvsb/elfvsb.dat
> >    9c9
> >    < shlib_shlibcall2 () == 7
> >    ---
> >    > shlib_shlibcall2 () == 8
> >    child process exited abnormally
> >    FAIL: visibility (protected) (non PIC)
> >    ....
> >
> > Anyway, I'm not after any detailed diagnosis, merely a little quick
> > reassurance [given that the failing tests involve 'ld', a key system
> > component] before proceeding any further and possibly creating
> > a 'broken' build.
>
> I wondered about the visibility tests when I saw the number of
> failures.  Yes, I've seen this, but I think it was on ppc64 multilib
> clfs!  The good news is that the basic system worked ok, as if the
> failure might be in the test code.
>

Thanks, Ken - that's the sort of reassurance I was definitely after :) !

>
> I haven't completed a regular LFS with tests for some weeks,
> but I wouldn't expect these failures on x86.
>
> Which kernel version are you using, and what host system
> and cpu ?
>

A 'uname -a' dump of the host:

    Linux linux001 2.6.5-1.358custom #5 Mon Jan 23 10:54:05 EST 2006
    i686 i686 i386 GNU/Linux

In plainer terms, a Fedora Core 2 system with a recompiled kernel, and the
latest updates applied, running on an IBM NetVista PIII 1 Ghz. Original
'gcc' used was:

    gcc (GCC) 3.3.3 20040412 (Red Hat Linux 3.3.3-7)

My build is actually targetting an i486 [my aim is to build a very lean ADSL
router / gateway / firewall], so am using a variation of the settings
recommended in the 'cross-compiling' hints:

http://www.linuxfromscratch.org/hints/downloads/files/crosscompiling-x86.txt

I have used the following environment variables throughout:

    CHOST="i486-pc-linux-gnu"
    CFLAGS="-s -Os -march=i486"
    CXXFLAGS="-s -Os -march=i486"

A typical 'configure' script would be something like:

    ../binutils-2.16.1/configure  --build="$CHOST" --host="$CHOST" \
    --target="$CHOST" --prefix=/usr --enable-shared

Like I said earlier, so far, so good - all the package building has been
error free [except for a hitch with 'expect' early in the piece (because I
used the wrong package version)] - and I thank my lucky stars for that. I
expected a few errors in the 'glibc' tests, so am not too concerned about
those, but did become somewhat concerned about the 'binutil' failures
because it seems that *no* failures are expected since the tests are run
via:

    make check

rather than:

    make -k check

However, from what you've said it seems safe enough to proceed with the
building process, and I hope to do so as soon as I can. Thanks again.

Cheers,

Anthony Borla




More information about the lfs-support mailing list