Re: Bug#179278: lintian: False positive for obsolete-c++-abi
- To: Brian Nelson <nelson@bignachos.com>, debian-gcc@lists.debian.org
- Subject: Re: Bug#179278: lintian: False positive for obsolete-c++-abi
- From: Josip Rodin <joy@srce.hr>
- Date: Sat, 1 Feb 2003 21:15:16 +0100
- Message-id: <[🔎] 20030201201516.GA5755@regoc.srce.hr>
- In-reply-to: <87isw3yht8.fsf@scabbers.bignachos.com>
- References: <87u1fo7txr.fsf@scabbers.bignachos.com> <20030201003445.GA17808@regoc.srce.hr> <87r8as7rwa.fsf@scabbers.bignachos.com> <20030201121134.GA30131@regoc.srce.hr> <87isw3yht8.fsf@scabbers.bignachos.com>
On Sat, Feb 01, 2003 at 10:30:43AM -0800, Brian Nelson wrote:
> >> > What does objdump -p say for those two files?
> >>
> >> They are missing the CXXABI_* mark as the lintian info suggests, but I
> >> don't know why that would be the case.
> >>
> >> $ objdump -p /usr/lib/libpspell.so.15.0.2
> >>
> >> /usr/lib/libpspell.so.15.0.2: file format elf32-i386
> >>
> >> Dynamic Section:
> >> NEEDED libaspell.so.15
> >> NEEDED libstdc++.so.5
> >> NEEDED libm.so.6
> >> NEEDED libgcc_s.so.1
> >> NEEDED libc.so.6
> >
> >> Version References:
> >> required from libc.so.6:
> >> 0x09691f73 0x00 02 GLIBC_2.1.3
> >
> > This definitely doesn't look like it was compiled against the right
> > libstdc++.so.5, since there's no stuff like:
> >
> > required from libstdc++.so.5:
> > 0x056bafd2 0x00 05 CXXABI_1.2
> > 0x081a2972 0x00 04 GLIBCPP_3.2
> >
> > Which I'm definitely seeing in other libraries. Are you sure you're really
> > checking the same file that was recompiled with g++ 3.2, and not some older
> > one?
>
> As far as I can tell. It's being linked with:
>
> g++ -shared dummy.lo -Wl,--rpath
> -Wl,/home/nelson/debian/aspell/aspell-0.50.3/lib/.libs
> ./.libs/libaspell.so -Wl,-soname -Wl,libpspell.so.15 -o
> .libs/libpspell.so.15.0.2
>
> where g++ is g++-3.2 as noted previously. The other binaries built by
> the package, including libaspell.so.15.0.2 and aspell do have the CXXABI
> specified.
>
> Hmm, I thought only g++-3.2 could build and link against libstdc++.so.5
> anyway. Otherwise, the binaries would be broken. These work fine,
> though.
>
> Perhaps a compiler or linker (whichever generates those headers) bug is
> at fault?
I'm sending this to debian-gcc so that they can comment.
--
2. That which causes joy or happiness.
Reply to: