MIPS %half in gas

[Richard Sandiford]

Daniel Jacobowitz <drow@false.org> writes:
> +    case BFD_RELOC_16:
> +      /* If we are deleting this reloc entry, we must fill in the
> +	 value now.  This can happen if we have a .word which is not
> +	 resolved when it appears but is later defined.  */

Minor nit, but this comment is for BFD_RELOC_16 in isolation, which can
never come from a .word directive.  I think you should just generalise
".word" to "instruction", or something like that.

> +      if (fixP->fx_done)
> +	{
> +	  if (*valP & ~0xffffull)
> +	    {
> +	      char value [32];
> +
> +	      sprintf_vma (value, *valP);
> +	      as_bad_where (fixP->fx_file, fixP->fx_line,
> +			    _("Number (0x%s) larger than 16 bits"), value);
> +	    }
> +
> +	  if (target_big_endian)
> +	    buf += 2;
> +	  md_number_to_chars ((char *) buf, *valP, 2);
> +	}
> +      break;
> +

Regarding the discussion downthread, the ABI specifically says that an
R_MIPS_16 addend is signed rather than an unsigned, so (a) it really should
be OK to use %half for ADDIU and LW, and (b), the *valP check above probably
needs to check a signed range rather than an unsigned one, or at least
allow both.

Richard