This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [v3] libstdc++/28125
- From: "Kaveh R. GHAZI" <ghazi at caip dot rutgers dot edu>
- To: Benjamin Kosnik <bkoz at redhat dot com>
- Cc: gcc-patches at gcc dot gnu dot org, libstdc++ at gcc dot gnu dot org
- Date: Fri, 22 Dec 2006 18:03:36 -0500 (EST)
- Subject: Re: [v3] libstdc++/28125
- References: <20061211230942.ac73b022.bkoz@redhat.com>
On Mon, 11 Dec 2006, Benjamin Kosnik wrote:
>
> Removes the remaining AC_CHECK_LIB in libstdc++ config.
>
> tested x86/linux
>
> gcc
> gcc-4_2-branch
> gcc-4_1-branch
Ben,
I'm getting tons of libstdc++ regressions now on my solaris10 box (so many
that the results get rejected from the testsuite mailing list for being
too large.)
I believe it's due to your patch. My box has a 32-bit copy of gnu
libiconv in /usr/local/lib. (Not my choice, I don't control it.) I run
the testsuite additionally with 64-bit multilibs (among other things) and
all these tests die because:
ld: fatal: file /usr/local/lib/libiconv.so: wrong ELF class: ELFCLASS32
which makes sense.
For the 32-bit tests, I'm also getting:
ld: warning: file /usr/local/lib/libiconv.so: attempted multiple inclusion of file
which makes less sense to me.
Anyway, solaris10 has the iconv functions in libc, and should be able to
resolve 32 or 64 bit automatically without linking anything else. So
IMHO, your configury magic should prefer the system's iconv functions
(i.e. try linking with no -liconv first) over any libiconv installed in
/usr/local/lib.
Thanks,
--Kaveh
--
Kaveh R. Ghazi ghazi@caip.rutgers.edu