On Mon, Aug 28, 2017 at 11:42:53AM -0600, Jeff Law wrote:
> On 07/17/2017 08:35 AM, Joerg Sonnenberger wrote:
> > Hello,
> > attached patch fixes inconsistent handling of section flags when using
> > the section attribute, i.e.:
> >
> > __attribute__((section("writable1"))) int foo1;
> > __attribute__((section("readonly1"))) const int foo1c;
> > __attribute__((section("writable2"))) int foo2 = 42;
> > __attribute__((section("readonly2"))) const int foo2c = 42;
> >
> > should give section attributes of "aw", "a", "aw", "a" in that order.
> > Currently, "foo1c" is classified as BSS though and therefore put into a
> > writable section.
> ISTM the test we need here is whether or not the underlying DECL is
> readonly. If it READONLY, then it shouldn't go into .bss, but should
> instead to into a readable section.
>
> Testing based on names seems wrong.
>
> Does the attached (untested) patch work for you?
The intention should work, will test it. The attached patch will likely
also be needed on top.
Joerg
Index: varasm.c
===================================================================
RCS file: /home/joerg/repo/netbsd/src/external/gpl3/gcc/dist/gcc/varasm.c,v
retrieving revision 1.2
retrieving revision 1.3
diff -u -p -r1.2 -r1.3
--- varasm.c 17 Jul 2017 19:53:10 -0000 1.2
+++ varasm.c 22 Jul 2017 20:52:52 -0000 1.3
@@ -6428,7 +6428,8 @@ categorize_decl_for_section (const_tree
location. -fmerge-all-constants allows even that (at the
expense of not conforming). */
ret = SECCAT_RODATA;
- else if (TREE_CODE (DECL_INITIAL (decl)) == STRING_CST)
+ else if (DECL_INITIAL (decl) != NULL
+ && TREE_CODE (DECL_INITIAL (decl)) == STRING_CST)
ret = SECCAT_RODATA_MERGE_STR_INIT;
else
ret = SECCAT_RODATA_MERGE_CONST;
