Re: Bug on octave 3.1.51+ with pkg install

[Jaroslav Hajek]

On Wed, Aug 20, 2008 at 9:59 PM, John W. Eaton <address@hidden> wrote:
> On 20-Aug-2008, John W. Eaton wrote:
>
> | On 18-Aug-2008, Jaroslav Hajek wrote:
> |
> | | Since there is no reason to wrap the subfunction
> | | is_architecture_dependent in an identity anonymous function, the
> | | attached workaround should make the problem vanish (and it is not a
> | | temporary solution).
> |
> | I applied the changeset.
> |
> | | Still, the underlying problem persists, as the subfunction seem to be
> | | unreachable in an anonymous function. Not always, though: it seems to
> | | work when pkg.m is copied into local directory. I'll try to submit a
> | | new bug report for this.
> |
> | Does anyone have a simpler test case that shows the problem?  That
> | would help, but either way I'll try to look at this problem sometime
> | this week.
>
> I found a simpler case:
>
>  function foo ()
>    c = {-1, 0, 1};
>    cellfun (@(x) sub2 (x), c)
>    sub1 (c);
>  endfunction
>  function sub1 (c)
>    cellfun (@(x) sub2 (x), c)
>  endfunction
>  function y = sub2 (x)
>    y = x > 0;
>  endfunction
>
> I think the attached patch should solve the problem.
>
> This code will probably have to change again in order to properly
> handle nested functions (not planned until after 3.2).
>

Could there just be parentage relations between scopes? That seems
simpler. A subfunction (or anonymous function) would just set up a
proper parent scope for its local scope. Name lookup, when
unsuccessful in a given scope, would simply proceed recursively on
parent scope, until there is no parent. There would be no need for
inherit (), I think.
Or is there a catch that I don't see?


> jwe
>
>
> # HG changeset patch
> # User John W. Eaton <address@hidden>
> # Date 1219262234 14400
> # Node ID 24701aa75ecbe5f64da30b756890888a858ef029
> # Parent  faf0abc5fd515423c7b95083b03260bdfcc6f777
> scope fixes for anonymous and inline functions that appear inside subfunctions
>
> diff --git a/src/ov-fcn-inline.cc b/src/ov-fcn-inline.cc
> --- a/src/ov-fcn-inline.cc
> +++ b/src/ov-fcn-inline.cc
> @@ -91,7 +91,20 @@
>          octave_user_function *uf = fcn.user_function_value ();
>
>          if (uf)
> -           uf->stash_parent_fcn_scope (octave_call_stack::current_scope ());
> +           {
> +             octave_function *curr_fcn = octave_call_stack::current ();
> +
> +             if (curr_fcn)
> +               {
> +                 symbol_table::scope_id parent_scope
> +                   = curr_fcn->parent_fcn_scope ();
> +
> +                 if (parent_scope < 0)
> +                   parent_scope = curr_fcn->scope ();
> +
> +                 uf->stash_parent_fcn_scope (parent_scope);
> +               }
> +           }
>        }
>     }
>
> diff --git a/src/pt-fcn-handle.cc b/src/pt-fcn-handle.cc
> --- a/src/pt-fcn-handle.cc
> +++ b/src/pt-fcn-handle.cc
> @@ -114,7 +114,13 @@
>   if (curr_fcn)
>     {
>       uf->stash_parent_fcn_name (curr_fcn->name ());
> -      uf->stash_parent_fcn_scope (curr_fcn->scope ());
> +
> +      symbol_table::scope_id parent_scope = curr_fcn->parent_fcn_scope ();
> +
> +      if (parent_scope < 0)
> +       parent_scope = curr_fcn->scope ();
> +
> +      uf->stash_parent_fcn_scope (parent_scope);
>     }
>
>   uf->mark_as_inline_function ();
>
>



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz