[open-cobol-list] Re: OC issues (more)

[Roger While]

No, we cannot comment it out.
The test has to stay for SECTION names.
We need different tests depending on section/paragraph.
Problem is that the same routine is called from parser.y.
Checking now. (probably need to duplicate the routine)

The level patch I will take a look at, but I can say now
that we would need to make it dependent on a config option
(ie allowed in -std=mf/ibm, not allowed in cobol85/2002 etc).

I recall there was a discussion on this issue some
time ago. The problem is trying to second-guess what is
meant. It is very prone to incorrect interpretation.
Depending on group/elementary definition, consider :
  02
     05
    03
     05

Roger

> Roger,
>
> Yes, it looks not necessary.
> For the test I've commented out return and replaced redefinition_error  by 
> redefinition_warning.
> Test went ok.
>
> And here is the patch for the first issue (which is about 
> misaligned/missed levels)
> G:\open-cobol\cobc>diff -u /open-cobol-0.33/cobc/field.c field.c
> --- /open-cobol-0.33/cobc/field.c       Sun Jan 08 08:00:00 2006
> +++ field.c     Tue Jan 17 18:12:36 2006
> @@ -154,14 +154,35 @@
>     {
>       /* upper level */
>       struct cb_field *p;
> -      for (p = last_field->parent; p; p = p->parent)
> -       if (p->level == f->level)
> -         {
> -           last_field = p;
> -           goto same_level;
> +         for (p = last_field->parent; p; p = p->parent) {
> +               if (p->level == f->level)
> +               {
> +                       last_field = p;
> +                       goto same_level;
> +               }
> +               if (p->level < f->level)
> +                       break;
> +         }
> +         if(p == NULL) { /* actually it can't be NULL so we can skip that */
> +                 cb_error_x (name, _("no previous data item of level 
> %02d"), f->level);
> +                 return cb_error_node;
> +         } else {
> +                 cb_tree fil = cb_build_filler();
> +                 struct cb_reference *rfil = CB_REFERENCE (fil);
> +                 struct cb_field *ffil = CB_FIELD (cb_build_field (fil));
> +                 cb_warning_x (name, _("no previous data item of level 
> %02d"),f->level);
> +                 ffil->level = f->level;
> +                 ffil->storage = storage;
> +                 ffil->children = p->children;
> +                 ffil->parent = p;
> +                 for(p = p->children; p != NULL; p = p->sister) {
> +                         p->parent = ffil;
> +                 }
> +                 ffil->parent->children = ffil;
> +                 ffil->sister = f;
> +                 f->parent = ffil->parent;
> +                 last_field = ffil;
>          }
> -      cb_error_x (name, _("no previous data item of level %02d"), f->level);
> -      return cb_error_node;
>     }
>
>   /* inherit parent's properties */
>
>
> ----- Original Message ----- From: "Roger While" <address@hidden>
> To: <address@hidden>
> Cc: <address@hidden>
> Sent: Wednesday, January 18, 2006 8:56 AM
> Subject: Re: OC issues (more)
>
>
> > Well the problem is caused by this code in typeck.c line 302 :
> >         if (CB_REFERENCE (name)->word->count > 0) {
> >                 cb_tree x = CB_VALUE (CB_REFERENCE (name)->word->items);
> >                 /* used as a non-label name or used as a section name or
> >                    used as the same paragraph name in the same section */
> >                 if (!CB_LABEL_P (x) || CB_LABEL (x)->section == NULL
> >                     || CB_LABEL (x)->section == current_section) {
> >                         redefinition_error (name);
> >                         return cb_error_node;
> >                 }
> > The first test is OK (i think) but I wonder why we do the
> > subsequent tests ?
> > Hmm.
> > Uniqueness is checked later.
> > Roger