This is the mail archive of the
gcc-regression@gcc.gnu.org
mailing list for the GCC project.
Re: 31 GCC regressions, 25 new, with your patch on 2003-06-24T18:06:05Z.
- From: Jan Hubicka <jh at suse dot cz>
- To: Andrew Pinski <pinskia at physics dot uc dot edu>
- Cc: gcc-regression at gcc dot gnu dot org,falk dot hueffner at student dot uni-tuebingen dot de, jakub at redhat dot com, jh at suse dot cz,joern dot rennecke at superh dot com, mark at codesourcery dot com,nathan at codesourcery dot com, pcarlini at unitus dot it
- Date: Wed, 25 Jun 2003 10:06:47 +0200
- Subject: Re: 31 GCC regressions, 25 new, with your patch on 2003-06-24T18:06:05Z.
- References: <200306242257.h5OMvp3b028476@gcc-regress.apple.com> <01C9F0A4-A6B2-11D7-865B-000393A6D2F2@physics.uc.edu>
>
> On Tuesday, Jun 24, 2003, at 18:57 US/Eastern, GCC regression checker
> wrote:
> >
>
> This one is caused by the patch:
> >native gcc.sum gcc.dg/pch/global-1.c
> +Tue Jun 24 18:49:33 CEST 2003 Jan Hubicka <jh@suse.cz>
> +
> + * Makefile.in (cgraph.o): Depend on output.h, not depend on
> + tree-inline.h
> + * cgraph.c: Do not include tree-inline.h; include output.h
> + (known_fns): Rename to ...
> + (known_decls): ... this one; update all uses.
> + (cgraph_varpool_hash): New static variable.
> + (cgraph_varpool_n_nodes, cgraph_varpool_nodes_queue): New global
> + variables.
> + (cgraph_varpool_hash_node, eq_cgraph_varpool_node,
> cgraph_varpool_node,
> + cgraph_varpool_node_for_identifier, cgraph_varpool_mark_needed_node,
> + cgraph_varpool_finalize_decl, cgraph_varpool_assemble_pending_decls):
> + New functions.
> + * cgraph.h (cgraph_varpool_node): New structure.
> + (cgraph_varpool_n_nodes, cgraph_varpool_nodes_queue): Declare.
> + (cgraph_varpool_node, cgraph_varpool_node_for_identifier,
> + cgraph_varpool_finalize_decl, cgraph_varpool_mark_needed_node,
> + cgraph_varpool_asemble_pending_decls): Declare.
> + * cgraphunit.c (record_call_1): Notice variable references.
> + (cgraph_finalize_compilation_unit): Assemble pending variables.
> + * toplev.c (wrapup_global_declarations): Use varpool.
> + (compile_file): Assemble pending declarations.
> + (rest_of_decl_compilation): Use varpool in unit-at-a-time mode.
> + * varasm.c (assemble_name): Notice varpool references.
> +
>
> These might be related to this patch also:
> >native objc.sum objc/compile/20011211-1.m,
> >native objc.sum objc/execute/accessing_ivars.m
> >native objc.sum objc/execute/bycopy-1.m
> >native objc.sum objc/execute/bycopy-2.m
> >native objc.sum objc/execute/compatibility_alias.m
> >native objc.sum objc/execute/enumeration-1.m
> >native objc.sum objc/execute/enumeration-2.m
> >native objc.sum objc/execute/formal_protocol-1.m
> >native objc.sum objc/execute/formal_protocol-2.m
> >native objc.sum objc/execute/formal_protocol-3.m
> >native objc.sum objc/execute/formal_protocol-4.m
> >native objc.sum objc/execute/formal_protocol-5.m
> >native objc.sum objc/execute/formal_protocol-6.m
> >native objc.sum objc/execute/formal_protocol-7.m
> >native objc.sum objc/execute/initialize.m
> >native objc.sum objc/execute/load-2.m
> >native objc.sum objc/execute/load-3.m
> >native objc.sum objc/execute/load.m
> >native objc.sum objc/execute/nil_method-1.m
> >native objc.sum objc/execute/no_clash.m
> >native objc.sum objc/execute/np-1.m
> >native objc.sum objc/execute/object_is_class.m
> >native objc.sum objc/execute/object_is_meta_class.m
> >native objc.sum objc/execute/selector-1.m
>
> All like error messages:
> Like:
> /tmp/cco5RejY.s:unknown:Undefined local symbol L_OBJC_CLASS_NAME_0^M
> /tmp/cco5RejY.s:unknown:Undefined local symbol L_OBJC_METH_VAR_NAME_12^M
>
> Remember Objective-C differences on Darwin so I think this patch should
> be removed or fixed soon.
OK, the fix is easy - objective C emits some extra data at the end of
compilation and does not mark it as needed.
Honza
>
>
> Thanks,
> Andrew Pinski
>