[PATCH, RFC] Fix _bfd_elf_provide_symbols for weak imports

[PATCH, RFC] Fix _bfd_elf_provide_symbols for weak imports

[Julian Brown]

Hi,

This patch fixes running __libc_{init,fini}_array (for 
.{init,fini}_array sections) with Newlib. The symbols 
__init_array_{start,end} etc. are imported as weak in Newlib, which 
conflicts with (I think) this change to use 
gld${EMULATION_NAME}_provide_bound_symbols rather than a linker script, 
to provide such symbols:

   http://sources.redhat.com/ml/binutils/2005-03/msg00688.html

Without the patch, the symbols resolved to zero, which stops static 
object construction/destruction from working. This affects the current 
csl-arm-branch gcc (and also soon mainline for the ARM EABI target, with 
a patch that is currently blocked on this problem).

Tested on arm-none-eabi with cross and with all targets on 
i686-pc-linux-gnu.

OK to apply?

Julian

ChangeLog:

bfd/

   * elflink.c (_bfd_elf_provide_symbol): Provide symbol for weak import.

Re: [PATCH, RFC] Fix _bfd_elf_provide_symbols for weak imports

[Julian Brown]

[-- Attachment #1: Type: text/plain, Size: 468 bytes --]

Julian Brown wrote:
> Hi,
> 
> This patch fixes running __libc_{init,fini}_array (for 
> .{init,fini}_array sections) with Newlib. The symbols 
> __init_array_{start,end} etc. are imported as weak in Newlib, which 
> conflicts with (I think) this change to use 
> gld${EMULATION_NAME}_provide_bound_symbols rather than a linker script, 
> to provide such symbols:
> 
>   http://sources.redhat.com/ml/binutils/2005-03/msg00688.html
> 

And now, with the patch.

Julian

[-- Attachment #2: binutils-weak-boundary-syms --]
[-- Type: text/plain, Size: 1140 bytes --]

? bfd/doc/bfd.info
? bfd/doc/bfd.info-1
? gas/doc/as.info
? gas/doc/as.info-1
? gprof/gprof.info
? gprof/gprof.info-1
Index: bfd/elflink.c
===================================================================
RCS file: /cvs/src/src/bfd/elflink.c,v
retrieving revision 1.148
diff -c -p -r1.148 elflink.c
*** bfd/elflink.c	27 Apr 2005 10:09:43 -0000	1.148
--- bfd/elflink.c	27 Apr 2005 19:01:52 -0000
*************** _bfd_elf_provide_symbol (struct bfd_link
*** 9808,9814 ****
    struct elf_link_hash_entry *h;
    h = elf_link_hash_lookup (elf_hash_table (info), name, FALSE, FALSE,
  			    FALSE);
!   if (h != NULL && h->root.type == bfd_link_hash_undefined)
      {
        h->root.type = bfd_link_hash_defined;
        h->root.u.def.section = bfd_abs_section_ptr;
--- 9808,9815 ----
    struct elf_link_hash_entry *h;
    h = elf_link_hash_lookup (elf_hash_table (info), name, FALSE, FALSE,
  			    FALSE);
!   if (h != NULL && (h->root.type == bfd_link_hash_undefined
! 	            || h->root.type == bfd_link_hash_undefweak))
      {
        h->root.type = bfd_link_hash_defined;
        h->root.u.def.section = bfd_abs_section_ptr;

Re: [PATCH, RFC] Fix _bfd_elf_provide_symbols for weak imports

[H. J. Lu]

On Wed, Apr 27, 2005 at 08:31:34PM +0100, Julian Brown wrote:
> Hi,
> 
> This patch fixes running __libc_{init,fini}_array (for 
> .{init,fini}_array sections) with Newlib. The symbols 
> __init_array_{start,end} etc. are imported as weak in Newlib, which 
> conflicts with (I think) this change to use 
> gld${EMULATION_NAME}_provide_bound_symbols rather than a linker script, 
> to provide such symbols:
> 
>   http://sources.redhat.com/ml/binutils/2005-03/msg00688.html
> 
> Without the patch, the symbols resolved to zero, which stops static 
> object construction/destruction from working. This affects the current 
> csl-arm-branch gcc (and also soon mainline for the ARM EABI target, with 
> a patch that is currently blocked on this problem).
> 
> Tested on arm-none-eabi with cross and with all targets on 
> i686-pc-linux-gnu.
> 
> OK to apply?
> 
> Julian
> 
> ChangeLog:
> 
> bfd/
> 
>   * elflink.c (_bfd_elf_provide_symbol): Provide symbol for weak import.

Where is the patch?


H.J.

Re: [PATCH, RFC] Fix _bfd_elf_provide_symbols for weak imports

[H. J. Lu]

On Wed, Apr 27, 2005 at 08:35:39PM +0100, Julian Brown wrote:
> Julian Brown wrote:
> >Hi,
> >
> >This patch fixes running __libc_{init,fini}_array (for 
> >.{init,fini}_array sections) with Newlib. The symbols 
> >__init_array_{start,end} etc. are imported as weak in Newlib, which 
> >conflicts with (I think) this change to use 
> >gld${EMULATION_NAME}_provide_bound_symbols rather than a linker script, 
> >to provide such symbols:
> >
> >  http://sources.redhat.com/ml/binutils/2005-03/msg00688.html
> >
> 
> And now, with the patch.
> 
> Julian

> ? bfd/doc/bfd.info
> ? bfd/doc/bfd.info-1
> ? gas/doc/as.info
> ? gas/doc/as.info-1
> ? gprof/gprof.info
> ? gprof/gprof.info-1
> Index: bfd/elflink.c
> ===================================================================
> RCS file: /cvs/src/src/bfd/elflink.c,v
> retrieving revision 1.148
> diff -c -p -r1.148 elflink.c
> *** bfd/elflink.c	27 Apr 2005 10:09:43 -0000	1.148
> --- bfd/elflink.c	27 Apr 2005 19:01:52 -0000
> *************** _bfd_elf_provide_symbol (struct bfd_link
> *** 9808,9814 ****
>     struct elf_link_hash_entry *h;
>     h = elf_link_hash_lookup (elf_hash_table (info), name, FALSE, FALSE,
>   			    FALSE);
> !   if (h != NULL && h->root.type == bfd_link_hash_undefined)
>       {
>         h->root.type = bfd_link_hash_defined;
>         h->root.u.def.section = bfd_abs_section_ptr;
> --- 9808,9815 ----
>     struct elf_link_hash_entry *h;
>     h = elf_link_hash_lookup (elf_hash_table (info), name, FALSE, FALSE,
>   			    FALSE);
> !   if (h != NULL && (h->root.type == bfd_link_hash_undefined
> ! 	            || h->root.type == bfd_link_hash_undefweak))
>       {
>         h->root.type = bfd_link_hash_defined;
>         h->root.u.def.section = bfd_abs_section_ptr;

I think it is an obvious patch.


H.J.

Re: [PATCH, RFC] Fix _bfd_elf_provide_symbols for weak imports

[Nick Clifton]

Hi Julian, Hi H.J.


>>    struct elf_link_hash_entry *h;
>>    h = elf_link_hash_lookup (elf_hash_table (info), name, FALSE, FALSE,
>>  			    FALSE);
>>!   if (h != NULL && h->root.type == bfd_link_hash_undefined)
>>      {
>>        h->root.type = bfd_link_hash_defined;
>>        h->root.u.def.section = bfd_abs_section_ptr;
>>--- 9808,9815 ----
>>    struct elf_link_hash_entry *h;
>>    h = elf_link_hash_lookup (elf_hash_table (info), name, FALSE, FALSE,
>>  			    FALSE);
>>!   if (h != NULL && (h->root.type == bfd_link_hash_undefined
>>! 	            || h->root.type == bfd_link_hash_undefweak))
>>      {
>>        h->root.type = bfd_link_hash_defined;
>>        h->root.u.def.section = bfd_abs_section_ptr;

> I think it is an obvious patch.

I agree - Julian - I have applied your patch.

Cheers
   Nick