This is the mail archive of the
mailing list for the GDB project.
Re: time to workaround libc/13097 in fsf gdb?
- From: Doug Evans <dje at google dot com>
- To: Pedro Alves <palves at redhat dot com>
- Cc: Jan Kratochvil <jan dot kratochvil at redhat dot com>, "gdb-patches at sourceware dot org" <gdb-patches at sourceware dot org>
- Date: Fri, 12 Sep 2014 18:06:12 -0700
- Subject: Re: time to workaround libc/13097 in fsf gdb?
- Authentication-results: sourceware.org; auth=none
- References: <CAP9bCMRko50FiANwa+h2FadG-k6Me69N04F+Le-bUnTVLQYKuQ at mail dot gmail dot com> <5411CFAE dot 7040805 at redhat dot com> <20140912115452 dot GA5626 at host2 dot jankratochvil dot net> <5412E3AC dot 80203 at redhat dot com>
Pedro Alves writes:
> On 09/12/2014 12:54 PM, Jan Kratochvil wrote:
> > On Thu, 11 Sep 2014 18:37:02 +0200, Pedro Alves wrote:
> >> Also, we know the address of the vDSO/gate (symfile-mem.c). Can't
> >> we use that to match instead of the name?
> > [...]
> >> ISTR having seen a patch that does that, but I can't seem to find it.
> > Re: [patch+7.5.1] Work around PR libc/13097 "linux-vdso.so.1" #3
> > https://sourceware.org/ml/gdb-patches/2012-11/msg00636.html
> > Message-ID: <20121125181505.GA26194@host2.jankratochvil.net>
> > It is pending/unreviewed/unapproved.
> Ah, yeah, I think that was it.
> I was more inclined to leave the vdso in the shared library list
> though, like ldd does, than filtering it out.
I like this too.
No point in hiding it from the user.
> Like, similar to
> your gdbarch_solib_file_not_found_is_ok patch, but look at the
> addresses rather than filenames in the hook. I'm not sure
> whether that'd complicate things too much.
I'm not sure either.
It *would* be nice to connect the processing of vdso when seen
by svr4_read_so_list (called during shared library loading)
with add_syscall_page (called as an observer when an inferior
is created). The same file (vdso) is processed in two disparate places
with no linkage for the human reader between them - a bit confusing.
> > Also I am not sure it is really still an issue on latest upstream glibc, it is
> > not an issue on Fedora 21 x86_64 glibc with the reproducer from:
> > https://sourceware.org/bugzilla/show_bug.cgi?id=13097
> > (That is old Fedoras workarounded it in glibc, then some Fedoras exposed the
> > issue in GDB but now it is not visible - so either Fedoras workaround it again
> > or just upstream glibc switched/workarounds it also.)
I downloaded glibc 2.20 and tested it with gdb 7.8.
Still an issue.
[I just did a simple build, if there's a configure option
that will fix things I didn't try one.]
I think FSF gdb 7.8[.1?] (the latest) should work with FSF glibc 2.20