Re: about Reloc Type: BFD_RELOC_8/16/32

[Nick Clifton <nickc at redhat dot com>]

Hi Daniel,

> So I wonder how to deal with the BFD_RELOC_32.

Treat it in the same way as a BFD_RELOC_16, except that the top 16 bits 
are zero.

So for example in your assembler source you have:

  [...]
  .section	.debug_abbrev,"",@progbits
  .Ldebug_abbrev0:
  [...]
  .section	.debug_info
  .4byte	0xb7
  .2byte	0x2
  .4byte	.Ldebug_abbrev0
  [...]

Which is generating a 32-bit reloc for the reference to Ldebug_abbrev0:

  Relocation section '.rela.debug_info' at offset 0x940 contains 23 
entries:
   Offset     Info    Type            Sym.Value  Sym. Name + Addend
  00000006  00000603 R_RICE_32         00000000   .debug_abbrev + 0

This relocation is here because the assembler does not know where the 
.Ldebug_abbrev0 symbol will be located at run time.  The relocation has 
to be a R_RICE_32 because the DWARF2 standard specifies that this field 
in the .debug_info header is a 4-byte value.

So in your code in bfd/elf16-rice.c (or whatever filename you have given 
to the RICE specific code in the BFD library) you need to handle 
R_RICE_32 relocations and insert a 32-bit value into the indicated place 
in the output binary.

For this particular example all that the code will have to do is to 
lookup the address of the Ldebug_abbrev0 symbol, (it is probably 0x0), 
zero-extend this from 16-bits to 32-bits, (you are using a 16-bit 
address space aren't you ?) and then store this 32-bit value into bytes 
6,7,8 and 9 of the .debug_info section.

Cheers
  Nick