This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix -fsection-anchors alias handling (PR middle-end/65003)


On Wed, 11 Feb 2015, Jakub Jelinek wrote:

> Hi!
> 
> This patch fixes a thinko in -fsection-anchors alias handling code
> Honza added in r211357.  Usually for aliases we don't create
> SYMBOL_REF_HAS_BLOCK_INFO_P, but in the following testcase we do,
> because the rtl is created through notice_global_symbol already
> when the symbol wasn't an alias, only later optimizations made it into
> an alias.  The thinko is that DECL_RTL of the alias target would be
> SYMBOL_REF, I think that can't ever happen, usually the DECL_RTL would
> be a MEM with SYMBOL_REF operand.

Yeah, indeed.

> I've bootstrapped/regtested this on
> {x86_64,i686,ppc64,ppc64le,aarch64,armv7hl,s390,s390x}-linux without
> regressions, so I hope the assertion this way is safe, if not, we'd
> need to avoid turning a symbol into an alias somewhere much earlier.
> 
> Ok for trunk?

Ok.

Thanks,
Richard.

> 2015-02-11  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR middle-end/65003
> 	* varasm.c (place_block_symbol): Assert that DECL_RTL of the
> 	ultimate alias is MEM with SYMBOL_REF satisfying
> 	SYMBOL_REF_HAS_BLOCK_INFO_P as its operand.  Don't pass the MEM
> 	to place_block_symbol, but instead pass the SYMBOL_REF operand of it.
> 
> 	* g++.dg/opt/pr65003.C: New test.
> 
> --- gcc/varasm.c.jj	2015-02-04 15:24:20.000000000 +0100
> +++ gcc/varasm.c	2015-02-10 15:37:55.872609142 +0100
> @@ -7180,6 +7180,10 @@ place_block_symbol (rtx symbol)
>  	{
>  	  rtx target = DECL_RTL (snode->ultimate_alias_target ()->decl);
>  
> +	  gcc_assert (MEM_P (target)
> +		      && GET_CODE (XEXP (target, 0)) == SYMBOL_REF
> +		      && SYMBOL_REF_HAS_BLOCK_INFO_P (XEXP (target, 0)));
> +	  target = XEXP (target, 0);
>  	  place_block_symbol (target);
>  	  SYMBOL_REF_BLOCK_OFFSET (symbol) = SYMBOL_REF_BLOCK_OFFSET (target);
>  	  return;
> --- gcc/testsuite/g++.dg/opt/pr65003.C.jj	2015-02-10 15:50:44.240734029 +0100
> +++ gcc/testsuite/g++.dg/opt/pr65003.C	2015-02-10 15:50:39.468814055 +0100
> @@ -0,0 +1,31 @@
> +// PR middle-end/65003
> +// { dg-do compile }
> +// { dg-options "-O2" }
> +// { dg-additional-options "-fpic" { target fpic } }
> +
> +struct A
> +{
> +  void operator= (A &);
> +  A ();
> +};
> +struct B
> +{
> +  A b;
> +};
> +struct C
> +{
> +  virtual bool foo (int &, bool) const;
> +};
> +struct D : virtual C
> +{
> +  bool foo (int &, bool) const;
> +  B e;
> +};
> +struct F : D
> +{
> +  F (int &, const int &, const A &);
> +  bool foo (int &, bool) const;
> +};
> +bool D::foo (int &, bool) const {}
> +F::F (int &, const int &, const A &) {}
> +bool F::foo (int &, bool) const {}
> 
> 	Jakub
> 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Jennifer Guild,
Dilip Upmanyu, Graham Norton HRB 21284 (AG Nuernberg)


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]