Re: [RFA] Patch for cgen_rtx_error()

Re: [RFA] Patch for cgen_rtx_error()

[Ben Elliston]

fche wrote:

   On Sun, Oct 08, 2000 at 11:38:00AM +1100, Ben Elliston wrote:

   > +/* Emit an error message from CGEN RTL.  */
   > +
   > +void
   > +cgen_rtx_error (SIM_CPU *cpu, const char * fmt, ...)
   > +{
   > +  va_list ap;
   > +  va_start(ap, fmt);
   > +  sim_io_printf (CPU_STATE (cpu), fmt, ap);
   > +  va_end(ap);
   > +  sim_io_printf (CPU_STATE (cpu), "\n");
   > +}
   
   Rather than just print the given message (is it really a varargs
   function?), it would be good to do a sim_engine_halt() there too.
   
Upon further inspection, it seems that the CGEN `error' rtx accepts only a
single argument, so this function need not accept varargs.  Right, Doug?

Without making assumptions about the architecture of any given port, what
are you proposing that I pass for the `pc' argument to sim_engine_halt?  
It's wrong to assume that GET_H_PC() and friends will be present.

Ben

Re: [RFA] Patch for cgen_rtx_error()

[Doug Evans]

Ben Elliston writes:
 > fche wrote:
 > 
 >    On Sun, Oct 08, 2000 at 11:38:00AM +1100, Ben Elliston wrote:
 > 
 >    > +/* Emit an error message from CGEN RTL.  */
 >    > +
 >    > +void
 >    > +cgen_rtx_error (SIM_CPU *cpu, const char * fmt, ...)
 >    > +{
 >    > +  va_list ap;
 >    > +  va_start(ap, fmt);
 >    > +  sim_io_printf (CPU_STATE (cpu), fmt, ap);
 >    > +  va_end(ap);
 >    > +  sim_io_printf (CPU_STATE (cpu), "\n");
 >    > +}
 >    
 >    Rather than just print the given message (is it really a varargs
 >    function?), it would be good to do a sim_engine_halt() there too.
 >    
 > Upon further inspection, it seems that the CGEN `error' rtx accepts only a
 > single argument, so this function need not accept varargs.  Right, Doug?

Can't remember if I've replied to this.

Leaving it as a varargs fn seems ok to me.
[I don't feel like being anally correct on this one.
Maybe the `error' rtx will be varargs some day.]

 > Without making assumptions about the architecture of any given port, what
 > are you proposing that I pass for the `pc' argument to sim_engine_halt?  
 > It's wrong to assume that GET_H_PC() and friends will be present.

Maybe the `error' rtx should pass `pc' as an argument.

Re: [RFA] Patch for cgen_rtx_error()

[Ben Elliston]

   Leaving it as a varargs fn seems ok to me. [I don't feel like being
   anally correct on this one. Maybe the `error' rtx will be varargs some
   day.]

I believe I followed fche's suggestion and made the function take a single
char * argument.  If we discover this is insufficient, we can easily change
it down the track (when the `error' rtx is modified).

Ben