Re: GCC 4.6 warnings, and dependency on nonexisting wrapset: standard

[Andy Wingo]

On Wed 07 Dec 2011 21:00, Mark H Weaver <address@hidden> writes:

> I recently compiled guile-gnome-platform from recent git, on top of
> Guile 2.0.3.50-d88f5, guile-cairo-1.9.91, and g-wrap 1.9.14.  This is on
> GCC 4.6.2 on MIPS little-endian with the N32 ABI.

I just pushed some fixes to a branch called `platform'.  I think we
should hack on that branch.  WDYT?

> First off, the guile-gnome build system uses -Werror by default, and
> this combined with some (apparently) new warnings in GCC 4.6 causes the
> build to fail.

I have fixed some of these, and disabled -Werror by default.

> Here's an excerpt of the generated code from the first warning above:
>
>   {
>     GList*gw__tmp7829_cursor = gw__c_arg0;
>     while(gw__tmp7829_cursor)
>     {
>       GdkPixbuf* gw__tmp7828_c_item;
>       gw__tmp7828_c_item = ( GdkPixbuf*) gw__tmp7829_cursor->data;
>   gw__tmp7829_cursor = gw__tmp7829_cursor->next;
>     }
>   }

I think I fixed this one.

> There was one "set but not used" warning in hand-written code, and
> here's a fix:
>
> diff --git a/gnome/gobject/gsignal.c b/gnome/gobject/gsignal.c
> index 59bccc4..1fc493b 100644
> --- a/gnome/gobject/gsignal.c
> +++ b/gnome/gobject/gsignal.c
> @@ -271,13 +271,13 @@ SCM_DEFINE (scm_gtype_instance_signal_connect_closure,
>      gtype = G_TYPE_FROM_INSTANCE (instance);
>      gclosure = g_value_get_boxed (gvalue);
>  
> -    g_signal_query (scm_to_ulong (id), &query);
> +    g_signal_query (signal_id, &query);
>      SCM_ASSERT (g_type_is_a (gtype, query.itype), object, SCM_ARG1, 
> FUNC_NAME);
>  
>  #ifdef DEBUG_PRINT
>      old_ref_count = gclosure->ref_count;
>  #endif
> -    handler_id = g_signal_connect_closure_by_id (instance, scm_to_ulong (id),
> +    handler_id = g_signal_connect_closure_by_id (instance, signal_id,
>                                                   detail_quark, gclosure,
>                                                   SCM_NFALSEP (after));
>      DEBUG_ALLOC ("GClosure %p connecting: %u->%u",

Ah, we should fix these.  Want to push to `platform' ?

> Another GCC 4.6 warning bites us in corba/gnome/corba/guile-gnome-corba.c:
>
>   guile-gnome-corba.c: In function 'scm_corba_servant_to_reference':
>   guile-gnome-corba.c:55:5: error: the comparison will always evaluate as 
> 'true' for the address of 'ev' will never be NULL [-Werror=address]
>   guile-gnome-corba.c: In function 'scm_bonobo_get_object':
>   guile-gnome-corba.c:82:9: error: the comparison will always evaluate as 
> 'true' for the address of 'ev' will never be NULL [-Werror=address]
>   guile-gnome-corba.c: In function 'scm_bonobo_object_query_interface':
>   guile-gnome-corba.c:109:9: error: the comparison will always evaluate as 
> 'true' for the address of 'ev' will never be NULL [-Werror=address]
>   cc1: all warnings being treated as errors

I "fixed" these.

> However, I run into problems at runtime with gtk/examples/calc.scm.
> The most relevant errors are probably these:
>
>   ;;; WARNING: compilation of /usr/share/guile-gnome-2/gnome/gtk.scm failed:
>   ;;; ERROR: In procedure %gw:wrapset-new: dependency on nonexisting wrapset: 
> standard
> [...]
>   ;;; WARNING: compilation of /home/mhw/bin/calc failed:
>   ;;; ERROR: In procedure %gw:wrapset-new: dependency on nonexisting wrapset: 
> standard
>
> There are many other warnings as well; the full transcript follows.
> Any ideas?

Hum, I'll give it a try.

Andy
-- 
http://wingolog.org/