Oh, btw:

On Tue, May 14, 2024 at 11:00:38AM +0800, Jiufu Guo wrote:
> >> --- a/gcc/config/rs6000/rs6000.cc
> >> +++ b/gcc/config/rs6000/rs6000.cc
> >> @@ -14659,6 +14659,12 @@ print_operand_address (FILE *file, rtx x)
> >>    else if (SYMBOL_REF_P (x) || GET_CODE (x) == CONST
> >>       || GET_CODE (x) == LABEL_REF)
> >>      {
> >> +      if (this_is_asm_operands && !address_operand (x, VOIDmode))
> >> +  {
> >> +    output_operand_lossage ("invalid expression as operand");
> >> +    return;
> >> +  }

That error message is not so good.  Firstly, it typically *is* a valid
expression here, just not a correct expression to have for an address.
But, more generally and usefully, the error message should say *what* is
wrong about the expression (namely, it is not an address).

Most of the time you can use the same error message for asm and other
expressions, and you get a great message in all contexts.
operand_lossage already takes care of telling the user "you did
something foolish" for inline asm, or "ICE" if it is a compiler problem
instead.

In error messages you do not often know what caused the problem, so
just report on the facts you *do* know (and moreso with warnings, there
you typically only know something looks unusual).


Segher

Reply via email to