On Wed, Dec 28, 2011 at 12:34:25PM -0800, Richard Henderson wrote:
> On 12/28/2011 09:39 AM, Michael Meissner wrote:
> > && in_section != text_section
> > - && !unlikely_text_section_p (in_section)
> > + && (in_section && (in_section->common.flags & SECTION_CODE)) == 0
>
> You should be able to delete the text_section test as well,
> and in_section should *never* be null, when emitting data.
>
> Otherwise this looks much better to me.
In talking to David, I convinced myself that initialized pointers should never
be in the text section (switch statements use relative pointers, not absolute
addresses currently), so I committed this patch on top of the previous one:
2011-12-29 Michael Meissner <meiss...@linux.vnet.ibm.com>
PR target/51623
* config/rs6000/rs6000.c (rs6000_assemble_integer): Delete check
for an initialized pointer being in a text section with
-mrelocatable, since it should never happen.
Index: gcc/config/rs6000/rs6000.c
===================================================================
--- gcc/config/rs6000/rs6000.c (revision 182730)
+++ gcc/config/rs6000/rs6000.c (working copy)
@@ -15456,12 +15456,13 @@ rs6000_assemble_integer (rtx x, unsigned
{
static int recurse = 0;
- /* For -mrelocatable, we mark all addresses that need to be fixed up
- in the .fixup section. */
+ /* For -mrelocatable, we mark all addresses that need to be fixed up in
+ the .fixup section. Since the TOC section is already relocated, we
+ don't need to mark it here. We used to skip the text section, but it
+ should never be valid for relocated addresses to be placed in the text
+ section. */
if (TARGET_RELOCATABLE
&& in_section != toc_section
- && in_section != text_section
- && (in_section && (in_section->common.flags & SECTION_CODE)) == 0
&& !recurse
&& GET_CODE (x) != CONST_INT
&& GET_CODE (x) != CONST_DOUBLE
--
Michael Meissner, IBM
5 Technology Place Drive, M/S 2757, Westford, MA 01886-3141, USA
meiss...@linux.vnet.ibm.com fax +1 (978) 399-6899
