Some bug reports_(re)

Wed, 26 Feb 1997 21:06:34 +0100 (MET)

According to Frank Heckenbach:
> Yes, I'm using DJGPP - sorry I forgot to mention. It seems redirection is
> turned off by default even without the BPcompat package. This was confusing
> me. I don't use Crt in my program, so I didn't think of DirectVideo. Now when
> I copy the declaration of DirectVideo from Crt (I don't want to use the
> whole unit in this program), it works.

Sorry for that inconvenience; now `libgpc.a' will get "DirectVideo = false"
by default.

> I guess redirection is turned off by default because the usual output via
> Dos is quite slow!?

No, it was just a bug in the patched `libgpc.a'.  (The unpatched version
of `libgpc.a' from the gpc-2.0 distribution should work.)

> > (And redirection is not easy with DJGPP anyway.)
> 
> Why? The only problem I know of is that Command.Com can't redirect stdout.

You mean: `stderr'.

> But that's merely a problem of Dos' shell, and is solved by the REDIR
> utility. Otherwise it works fine for me. Or are there any other problems?

That's what I meant with "not easy": you need `REDIR' to redirect `stderr'.

> 3.
> Yes, I know. Actually I'm using now:
> CONST MaxVar=$FFF0;
> TYPE XArray=ARRAY[1..MaxVar {$IFDEF TP} DIV SizeOf(X) {$ENDIF}] OF X;

The bug is fixed for the next alpha release.  If you want the patch now,
here is the diff for gpc-parse.y:

        if (PEDANTIC (E_O_PASCAL)
            && (! TREE_CONSTANT ($3) || ! TREE_CONSTANT ($1)))
          pedwarn ("ISO 7185 Pascal does not allow non-constant range bounds");
-       if (TREE_TYPE ($1) != TREE_TYPE ($3))
+       if (TREE_TYPE ($1) == TREE_TYPE ($3)
+           || (TREE_CODE (TREE_TYPE ($1)) == INTEGER_TYPE
+               && TREE_CODE (TREE_TYPE ($3)) == INTEGER_TYPE))
+         $$ = build_range_type (TREE_TYPE ($1), $1, $3);
+       else
          {
            error ("subrange bounds are not of the same type");
            $$ = error_mark_node;
          }
-       else
-         /* @@@ Changed in 2.4.3.1 */
-         $$ = build_range_type (TREE_TYPE ($1), $1, $3);
      }
    ;


> 6.
> > > Is there anything like TP's Exit (return from current procedure/function)?
> > > I couldn't find it in the docs.

`Return'.  In the alpha versions, `Exit' works, too.

> 4.
> {$DEFINE Const PROTECTED}
> [...]
> PROCEDURE p(Const c:t);

That bug is fixed as well.  Here is the diff for gpc-util.c:

     }
   else
     {
+      if (varparm && type != error_mark_node)
+       type = build_reference_type (type);
+
       /* If this is a protected parameter, make it read-only */
       if (protected)
        type = build_type_variant (type, 1, 0);

-      if (varparm && type != error_mark_node)
-       type = build_reference_type (type);
-
       type = build_tree_list (NULL_TREE, type);

       for (link = idlist; link; link = TREE_CHAIN(link))

> 7.
> ...but this led me to another problem. It seems more involved (perhaps
> something about GPI files?).

Yes. :-(  It's a known bug in the GPI mechanism that pointer types
transported through a GPI file aren't compatible with themselves.

> gpc complains about
> line {}: incompatible type for argument 1 of `F'

To work around, patch your gpc-decl.c (or wait for the next alpha),
and use `Const' :-P instead fo `Protected'.  Then you will get some
warnings, but the code will work.

    Peter

  Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer
peter.gerwinski@uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201]
       maintainer GNU Pascal - http://home.pages.de/~gnu-pascal/ [970125]

Peter Gerwinski (peter@agnes.dida.physik.uni-essen.de)