Re: Catching errors from scm_run_hook

[Greg Badros <gjb@cs.washington.edu>]

Mikael Djurfeldt <mdj@mdj-pc.nada.kth.se> writes:

> Greg Badros <gjb@cs.washington.edu> writes:
> 
> > Mikael Djurfeldt <mdj@mdj-pc.nada.kth.se> writes:
> > 
> > > Greg Badros <gjb@cs.washington.edu> writes:
> > > 
> > > > What is the right, supported way to catch errors when invoking
> > > > scm_run_hook?
> > > 
> > > The right thing would be to use scm_internal_catch.  Is that too
> > > complicated?
> > 
> > Maybe not, if it were well documented.  But, I think it's almost always
> > the case that the user will want to run hooks catching errors, so that
> > common case should have a simpler facility.
> 
> I think you should include such a facility in SCWM.  If enough
> application writers require such a facility, we'll include it in a
> later release.

We do:

__inline__ SCM scwm_run_hook(SCM hook, SCM args)
{
  return scwm_safe_apply(run_hook_proc, gh_cons(hook,args));
}

which uses:

SCM
scwm_safe_apply (SCM proc, SCM args)
{
  SCM_STACKITEM stack_item;
  struct scwm_body_apply_data apply_data;

  apply_data.proc = proc;
  apply_data.args = args;

  return scm_internal_stack_cwdr(scwm_body_apply, &apply_data,
				 scwm_handle_error, "scwm",
				 &stack_item);
}


which in turn uses:

SCM
scm_internal_stack_cwdr (scm_catch_body_t body,
			 void *body_data,
			 scm_catch_handler_t handler,
			 void *handler_data,
			 SCM_STACKITEM *stack_item)
{
  struct cwssdr_data d;
  d.tag = SCM_BOOL_T;
  d.body = body;
  d.data = body_data;
  d.handler = handler;
  return scm_internal_cwdr_no_unwind (cwssdr_body, &d, handler, handler_data, 
			    stack_item);
}


which seems pretty complicated for this simple desire.  I believe Maciej 
wanted to use a new dynamic root to avoid misusing continuations.

Comments?  If all this machinery is really necessary, then it's way too
hard for the application programmer to get right.

Thanks,
Greg