[patch] Fix handling of common symbols with plugins

[Alan Modra]

On Mon, Sep 29, 2014 at 10:21:44AM -0700, Cary Coutant wrote:
> Is this still failing even after my patch last Thursday for PR
> gold/17432? Or did it only start failing after that patch?
> 
> I see 8 in the value field in the .syms file, and an alignment of 8 on
> .bss in the executable.
> 
> > Fails on x86_64 when using mainline gcc.  c1 is aligned (value = 8) in
> > the relocatable object file
> >     20: 0000000000000008     4 OBJECT  GLOBAL DEFAULT  COM c1
> 
> That looks like a dump of plugin_common_test_2.o.

Yes.

> > plugin_common_test_1.syms:
> >     23: 0000000000000004     4 OBJECT  GLOBAL DEFAULT  COM c1
> 
> But this is from plugin_common_test_1.o. I expect
> plugin_common_test_1.syms to have an alignment of 4.

Right.

> > Final object file:
> >     39: 0000000000401b7c     4 OBJECT  GLOBAL DEFAULT   25 c1
> > and .bss only has an alignment of 4.
> 
> Is this from plugin_test_10?

Yes.  The breakage appears to be due to 5efeedf6.  Prior to that the
test passes.

-- 
Alan Modra
Australia Development Lab, IBM