pic patch for arm-elf

[Daniel Jacobowitz]

On Sun, Oct 26, 2003 at 11:52:48PM +0000, Phil Blundell wrote:
> This patch implements GC for GOT and PLT relocs in the elf32-arm
> backend.  One effect of this is to avoid a spurious R_ARM_NONE reloc
> which was blighting glibc by causing libc.so to be marked as
> DT_TEXTREL.  
> 
> It also replaces the code sequence used for PLT thunks with a more
> efficient version, which can make a significant difference on some
> processors.
> 
> May I check this in?

Found a probem :(

> @@ -1134,16 +1307,21 @@ elf32_arm_final_link_relocate (howto, in
>  #ifndef OLD_ARM_ABI
>      case R_ARM_XPC25:
>  #endif
> +      /* r_symndx will be zero only for relocs against symbols
> +	 from removed linkonce sections, or sections discarded by
> +	 a linker script.  */
> +      if (r_symndx == 0
> +	  || (input_section->flags & SEC_ALLOC) == 0)
> +	return bfd_reloc_ok;
> +
>        /* When generating a shared object, these relocations are copied
>  	 into the output file to be resolved at run time.  */
> -      if (info->shared
> -	  && r_symndx != 0
> -	  && (r_type != R_ARM_PC24
> -	      || (h != NULL
> -	          && h->dynindx != -1
> -		  && (! info->symbolic
> -		      || (h->elf_link_hash_flags
> -			  & ELF_LINK_HASH_DEF_REGULAR) == 0))))
> +      if ((info->shared
> +	   && (h == NULL
> +	       || ELF_ST_VISIBILITY (h->other) == STV_DEFAULT
> +	       || h->root.type != bfd_link_hash_undefweak)
> +	   && (r_type != R_ARM_PC24
> +	       || !SYMBOL_CALLS_LOCAL (info, h))))
>  	{
>  	  Elf_Internal_Rela outrel;
>  	  bfd_byte *loc;

What's the SEC_ALLOC check for?  It breaks relocation of .debug_info.
I'm testing without it.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer