[Bug ld/21532] AArch64: Symbol address inconsistency across compilation units

[jiwang at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=21532

Jiong Wang <jiwang at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at sourceware dot org   |jiwang at gcc dot 
gnu.org

--- Comment #4 from Jiong Wang <jiwang at gcc dot gnu.org> ---
(In reply to Nick Clifton from comment #3)
> Hi Jiong,
> 
>   Thanks for taking a look at this.
> 
> >    Looks like AArch64 could avoid emiting COPY relocation if it's a 
> > relocation
> > against writable section. We could adopt similar code from
> > elf_x86_64_adjust_dynamic_symbol or ppc64_elf_adjust_dynamic_symbol.
> 
> Ah - maybe not.  The original testcase (in the BZ) shows the bug happening
> when
> constant data is used rather than writeable data.  You can also extend the
> second
> testcase I posted to use constant data and the same problem happens.
> 
> 
> >> It looks like the linker creates a copy of IEEEsingle in the executable but
> >> then the GOT entry for IEEEsingle is not updated in libLLVM.so, causing any
> >> accesses to IEEEsingle in libLLVM.so to refer to the original definition
> > 
> >    Does this means this is a glibc bug as well? dynamic linker needs to 
> > redirect
> > the reference in the .so to the copy as well?
> 
> This is a possibility, although I am not an expert on the dynamic linker.  I
> think
> however that we need to verify that the static linker is producing the
> correct
> dynamic relocations first.

Thanks, I will prepare a patch to fix the BFD static linker.

-- 
You are receiving this mail because:
You are on the CC list for the bug.