[PATCH] gas/symbols: do not panic upon resolving O_md

[Alan Modra]

On Fri, Jul 15, 2022 at 12:26:52AM +0300, Dmitry Selyutin wrote:
> Assuming GMSD is a special operand, marked as O_md1, the code:
> 
>     .set VREG, GMSD
>     .set REG, VREG
>     extsw REG, 2
> 
> ...fails upon attempts to resolve the value of the symbol. This happens
> since machine-dependent values are not handled in the giant op switch.
> 
> Machine-dependent expressions don't really need to be resolved, since
> the resolving process is really machine-dependent. We could have marked
> such symbols as resolving in port. However, we don't want to access
> the field which seems to be internal, and we especially don't want
> to perform this for each and every port.
> ---
>  gas/symbols.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/gas/symbols.c b/gas/symbols.c
> index 6904a3102c..fd2117f7da 100644
> --- a/gas/symbols.c
> +++ b/gas/symbols.c
> @@ -1408,6 +1408,16 @@ resolve_symbol_value (symbolS *symp)
>  	  BAD_CASE (op);
>  	  break;
>  
> +	case O_md1...O_md32:

Please don't use gcc extensions.

> +	  /* Machine-dependent expressions don't really need
> +	   * to be resolved, since the resolving process is
> +	   * really machine-dependent. We could have marked
> +	   * such symbols as resolving in port. However, we
> +	   * don't want to access the field which seems to be
> +	   * internal, and we especially don't want to perform
> +	   * this for each and every port. */
> +	  break;

This might avoid the fatal error, but will still hit an error later
due to resolved not being set.  I think this calls for an
md_resolve_symbol.

> +
>  	case O_absent:
>  	  final_val = 0;
>  	  /* Fall through.  */
> -- 
> 2.37.0

-- 
Alan Modra
Australia Development Lab, IBM