elf32-mips, objcopy, reginfo

elf32-mips, objcopy, reginfo

[Don Bowman]

If I do a simple objcopy infile outfile,
the ri_gp_value register in the .reginfo section is set to 0.
This occurs since:

1) The input section is copied to the output section
2) mips_elf_section_processing subsequently copies elf_gp there,
   but elf_gp was never set for an objcopy.

Does anyone have any suggestions on the best place to change this?

It seems that all of the processing except mips_elf_section_processing
is architecture independent. The bfd_map_over_sections on both
setup_section and copy_section doesn't set the output elf_gp
to the input_gp ever.

At the time I do the mips_elf_section_processing, I no longer
have the input bfd.

Advice?

--
 Don Bowman        **      bowman@waterloo.hp.com     **           519-883-3019 

Re: elf32-mips, objcopy, reginfo

[Ian Lance Taylor]

   From: Don Bowman <bowman@waterloo.hp.com>
   Date: Mon, 29 Apr 96 17:24:33 EDT

   I've found a problem in objcopy/bfd 960425.

   If I do a simple objcopy infile outfile,
   the ri_gp_value register in the .reginfo section is set to 0.
   This occurs since:

   1) The input section is copied to the output section
   2) mips_elf_section_processing subsequently copies elf_gp there,
      but elf_gp was never set for an objcopy.

   Does anyone have any suggestions on the best place to change this?

   It seems that all of the processing except mips_elf_section_processing
   is architecture independent. The bfd_map_over_sections on both
   setup_section and copy_section doesn't set the output elf_gp
   to the input_gp ever.

   At the time I do the mips_elf_section_processing, I no longer
   have the input bfd.

   Advice?

This type of thing should normally be handled in the copy_private_data
routine: mips_elf_copy_private_bfd_data.

Ian