Re: [PATCH] libpoke,etc: Re-write `ASTREF`/`ASTDEREF` as an inline function

[Jose E. Marchesi]

> Hi, Jose.
>
> On Sun, Nov 29, 2020 at 10:18:18AM +0100, Jose E. Marchesi wrote:
>> 
>> > @@ -1894,6 +1888,22 @@ union pkl_ast_node
>> >    struct pkl_ast_print_stmt_arg print_stmt_arg;
>> >  };
>> >  
>> > +static inline pkl_ast_node __attribute__ ((always_inline, 
>> > warn_unused_result))
>> > +ASTREF (pkl_ast_node ast)
>> > +{
>> > +  if (ast)
>> > +    ++ast->common.refcount;
>> > +  return ast;
>> > +}
>> > +
>> > +static inline pkl_ast_node __attribute__ ((always_inline, 
>> > warn_unused_result))
>> > +ASTDEREF (pkl_ast_node ast)
>> > +{
>> > +  if (ast)
>> > +    --ast->common.refcount;
>> > +  return ast;
>> > +}
>> > +
>> 
>> Hm, I'm unsure regarding warn_unused_result and the next thunks in your
>> patch.  Sometimes ASTREF/ASTDEREF are called only for their side effect,
>> as you have found. Forcing using an assignment in these cases is a bit
>> ugly...
>> 
>
> I admit the ugliness, but I didn't find another way to enforce the `ast`
> to be a l-value.
>
> These are a few cases and maybe we can accept this ugliness.
> Or finding another way to ensure the l-valueness of `ast`.

Yeah I know.
Ok, lets put this in.  It is certainly better than the previous macros.

Thanks.