This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] PR ld/19579: [Regression] link error linking fortran code with PIE
- From: Alan Modra <amodra at gmail dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Binutils <binutils at sourceware dot org>
- Date: Tue, 8 Mar 2016 11:28:40 +1030
- Subject: Re: [PATCH] PR ld/19579: [Regression] link error linking fortran code with PIE
- Authentication-results: sourceware.org; auth=none
- References: <20160304134833 dot GA11350 at gmail dot com> <20160305015242 dot GE9617 at bubble dot grove dot modra dot org> <CAMe9rOq=mgfjaFgAdw=u3dzvtt3sG519Pipy4VyyPS1T3+GJSA at mail dot gmail dot com> <20160307152818 dot GH9617 at bubble dot grove dot modra dot org> <CAMe9rOpZh-Qem3Vt3vwsQG9K4T+9=Me3V_7QtW7_brc_pPo+Cg at mail dot gmail dot com>
On Mon, Mar 07, 2016 at 10:44:29AM -0800, H.J. Lu wrote:
> On Mon, Mar 7, 2016 at 7:28 AM, Alan Modra <amodra@gmail.com> wrote:
> > On Fri, Mar 04, 2016 at 07:41:42PM -0800, H.J. Lu wrote:
> >> On Fri, Mar 4, 2016 at 5:52 PM, Alan Modra <amodra@gmail.com> wrote:
> >> > On Fri, Mar 04, 2016 at 05:48:33AM -0800, H.J. Lu wrote:
> >> >> --- a/bfd/elflink.c
> >> >> +++ b/bfd/elflink.c
> >> >> @@ -1172,9 +1172,12 @@ _bfd_elf_merge_symbol (bfd *abfd,
> >> >>
> >> >> newdef = !bfd_is_und_section (sec) && !bfd_is_com_section (sec);
> >> >>
> >> >> + /* The old common symbol in executable is a definition if the new
> >> >> + definition comes from a shared library. */
> >> >> olddef = (h->root.type != bfd_link_hash_undefined
> >> >> && h->root.type != bfd_link_hash_undefweak
> >> >> - && h->root.type != bfd_link_hash_common);
> >> >> + && (h->root.type != bfd_link_hash_common
> >> >> + || (!olddyn && newdyn && bfd_link_executable (info))));
> >> >>
> >> >> /* NEWFUNC and OLDFUNC indicate whether the new or old symbol,
> >> >> respectively, appear to be a function. */
> >> >
> >> > Why is this the correct place to change, and not code after the
> >> > comment "We treat a common symbol as a definition"?
> >>
> >> olddef has been checked well before that.
> >
> > And do any of those matter?
[snip TLS example]
I don't think an error message difference is particularly relevant.
> >> We need to get it right.
> >
> > That's why I asked. You haven't yet replied with anything more than a
> > superficial reason for not moving the change to where it ought to go,
> > I think.
>
> The old common symbol is a definition in this case. Why shouldn't
> olddef set to yes?
A common is *not* a definition. We treat a common *as if* it were a
definition in some cases.
The distinction might be subtle, but I'm concerned that in some future
change to _bfd_elf_merge_symbol someone might think that olddef being
true really means we have a definition.
Also, we now have two places in _bfd_elf_merge_symbol where we modify
the code to treat a common as a definition. I'd rather have just one
place, again for maintainability reasons.
--
Alan Modra
Australia Development Lab, IBM