Hi!
When Maxim posted his patch today, I had a look and found that we emit
these indirect constants into .data sections, even when they always contain
just some pointer that needs relocation, but shouldn't be changed
afterwards; after all, the indirect constants have TREE_READONLY set.
The problem is that we use DECL_INITIAL (decl) = decl; and that isn't
something the section selection code recognizes as valid or reasonable
initializer. This patch uses ADDR_EXPR of the decl instead, i.e. something
that needs a relocation if -fpic/-fpie, but can be relro, and in position
dependent binaries can be .rodata from the beginning. This will make it
harder from somebody trying to change these.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2018-03-19 Jakub Jelinek <ja...@redhat.com>
PR sanitizer/78651
* dwarf2asm.c: Include fold-const.c.
(dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
of decl rather than decl itself.
--- gcc/dwarf2asm.c.jj 2018-02-09 06:44:29.952809199 +0100
+++ gcc/dwarf2asm.c 2018-03-19 17:18:09.824000048 +0100
@@ -33,6 +33,7 @@ along with GCC; see the file COPYING3.
#include "dwarf2.h"
#include "function.h"
#include "emit-rtl.h"
+#include "fold-const.h"
#ifndef XCOFF_DEBUGGING_INFO
#define XCOFF_DEBUGGING_INFO 0
@@ -954,7 +955,7 @@ dw2_output_indirect_constant_1 (const ch
SET_DECL_ASSEMBLER_NAME (decl, id);
DECL_ARTIFICIAL (decl) = 1;
DECL_IGNORED_P (decl) = 1;
- DECL_INITIAL (decl) = decl;
+ DECL_INITIAL (decl) = build_fold_addr_expr (decl);
TREE_READONLY (decl) = 1;
TREE_STATIC (decl) = 1;
Jakub
