[patch][rfc] Complex Relocations gas and bfd

[Dave Brolley]

Alan Modra wrote:

>gas should use a single BFD_RELOC_RELC.  (The BFD_RELOC_* numbering is
>internal to BFD.)   This number could then be translated by each
>backend bfd_reloc_type_lookup function to the appropriate howto.
>  
>
Thanks again. Using this approach I now have an implementation for which 
R_RELC is defined on a target by target basis, as desired. As you 
suggested, I added R_RELC to the howto table for the target and it is 
now obtained by each backend using bfd_reloc_type_lookup.

The other key change was that perform_complex_relocations in 
bfd/elflink.c, which previously examined all relocs looking for complex 
ones (and hence needed to know the value of R_RELC) has been renamed to 
perform_complex_relocation and changed so that it only examines one 
reloc, provided by the caller. Thus it can assume that the reloc is 
complex and need not check for r_type == R_RELC. This function is now 
called by the target specific code from <target>_elf_relocate_section as 
each complex reloc is encountered.

Attached is a patch showing the changes. It must be applied after my 
previous patch. Is that ok, or would you like a new overall patch?

Dave

>Alternatively, you could do something clever in the generic code.
>Store the target reloc in elf_backend_data, and stash that value into
>a howto.
>
>  
>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: relc-1.ChangeLog
URL: <https://sourceware.org/pipermail/binutils/attachments/20070125/ba195aaf/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: relc-1.patch.txt
URL: <https://sourceware.org/pipermail/binutils/attachments/20070125/ba195aaf/attachment.txt>