Re: PR42917 (make lambda-code VTA aware)

[Aldy Hernandez <aldyh at redhat dot com>]

> > +	    {
> > +	      if (is_gimple_debug (stmt))
> > +		{
> > +		  gimple_stmt_iterator gsi = gsi_for_stmt (stmt);
> > +		  gsi_remove (&gsi, true);
> 
> Shouldn't this be instead a debug reset?

Beats me.  I don't quite understand why you can't just remove DEBUG
statements, as I've asked elsewhere.  But I'm happy to reset the value
if that's the preferred method.

> > +	  if (ivdebug)
> > +	    {
> > +	      inner_gsi = gsi_after_labels (temp->inner->header);
> > +	      for (; !gsi_end_p (inner_gsi); gsi_next (&inner_gsi))
> > +		if (gimple_code (gsi_stmt (inner_gsi)) != GIMPLE_PHI)
> > +		  break;
> 
> GIMPLE_PHIs IMHO can't appear in gsi_after_labels sequence.  There is a
> separate sequence for PHIs and separate for statements.

Cool, so I can just do

	inner_gsi = gsi_after_labels (temp->inner->header);

and be done with it?

I love when people suggest I do less than what I'm doing.
Aldy