Re: c++/1864

Re: c++/1864

[nathan]

The following reply was made to PR c++/1864; it has been noted by GNATS.

From: nathan@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org, nathan@gcc.gnu.org, snyder@fnal.gov
Cc:  
Subject: Re: c++/1864
Date: 5 Feb 2001 11:59:52 -0000

 Synopsis: V3 ABI bug
 
 State-Changed-From-To: analyzed->closed
 State-Changed-By: nathan
 State-Changed-When: Mon Feb  5 03:59:52 2001
 State-Changed-Why:
     2001-02-05  Nathan Sidwell  <nathan@codesourcery.com>
     
             * invoke.texi (-fdump-class-layout): Document.
             (-fdump-translation-unit): Use `=' as filename separator.
     
     2001-02-05  Nathan Sidwell  <nathan@codesourcery.com>
     
             Fix specification and implementation bugs in V3 ABI
             construction vtables.
             * cp-tree.h (flag_dump_class_layout): New flag.
             (BINFO_OVERRIDE_ALONG_VIRTUAL_PATH_P): Remove.
             (BINFO_LOST_PRIMARY_P): New flag.
             (SET_BINFO_NEW_VTABLE_MARKED): Adjust asserts.
             (BINFO_PRIMARY_MARKED_P): Rename to ...
             (BINFO_PRIMARY_P): ... here.
             (binfo_via_virtual): New prototype.
             * decl2.c (flag_dump_class_layout): New flag.
             (cxx_decode_option): Set it. Adjust -fdump-translation-unit to
             use `=' as a file name separator.
             * init.c (dfs_initialize_vtbl_ptrs): Walk into virtual primary
             bases.
             (build_vtbl_address): If this is a virtual primary base, then
             get the vtbl of what it is ultimately primary for.
             * search.c (dfs_skip_nonprimary_vbases_unmarkedp): Adjust
             for BINFO_PRIMARY_P.
             (dfs_skip_nonprimary_vbases_markedp): Likewise.
             (get_shared_vbase_if_not_primary): Likewise.
             (dfs_get_pure_virtuals): Likewise.
             (expand_upcast_fixups): Likewise.
             (fixup_virtual_upcast_offsets): Likewise.
             (dfs_find_vbase_instance): Likewise.
             (find_vbase_instance): Likewise.
             (binfo_from_vbase): Adjust comment to reflect reality.
             (binfo_via_virtual): New function.
             * class.c (VTT_TOP_LEVEL_P, VTT_MARKED_BINFO_P): New macros
             for binfo walking during VTT construction.
             (dfs_mark_primary_bases): Remove.
             (force_canonical_binfo_r): New function.
             (force_canonical_binfo): New function.
             (mark_primary_virtual_base): New function.
             (mark_primary_bases): Walk in inheritance graph order, use
             mark_primary_virtual_base.
             (determine_primary_base): Use some more intermediate variables.
             (dfs_find_final_overrider): Don't check for overriding along a
             virtual path.
             (dfs_modify_vtables): Walk into primary virtual bases too.
             (walk_subobject_offsets): Adjust for BINFO_PRIMARY_P.
             (build_base_fields): Likewise.
             (dfs_set_offset_for_unshared_vbases): Likewise.
             (layout_virtual_bases): Likewise.
             (end_of_class): Likewise.
             (finish_struct_1): Call dump_class_hierarchy, if requested.
             (dfs_get_primary_binfo): Use BINFO_TYPE for binfos.
             (dump_class_hierarchy_r): Add stream parameter. Emit more information.
             (dump_class_hierarchy): Add file parameter. Append to file, if
             required.
             (finish_vtbls): Adjust accumulate_vtbl_inits call.
             Use canonical base for virtual bases.
             (build_vtt): Add more comments. Adjust build_vtt_inits call.
             (build_vtt_inits): Remove VIRTUAL_VTTS_P parm.
             Only set BINFO_VPTR_INDEX on top level. Use VTT_TOP_LEVEL_P,
             VTT_MARKED_BINFO_P for binfo walking. Use canonical vbase for
             virtual VTTs.
             (dfs_build_secondary_vptr_vtt_inits): Extract VTT_TOP_LEVEL_P
             from DATA.  We want virtual primary bases and all bases via virtual.
             Only set BINFO_VPTR_INDEX for top level. Look up from a primary
             virtual base when not a construction vtable.
             (dfs_ctor_vtable_bases_queue_p): New DFS predicate.
             (build_ctor_vtbl_group): Adjust accumulate_vtbl_inits call.
             Use canonical bases when processing virtual bases.
             (accumulate_vtbl_inits): We're interested in any base via a
             virtual path.
             (dfs_accumulate_vtbl_inits): If this is a primary virtual base
             within a construction vtable, determine what is being overridden.
             (build_vtbl_initializer): Add more comments
             (add_vcall_offset_vtbl_entries_1): Adjust comment.
             (build_rtti_vtbl_entries): Check if the base has lost its
             primary.
     
     2001-02-05  Nathan Sidwell  <nathan@codesourcery.com>
     
             * g++.old-deja/g++.abi/primary2.C: New test.
             * g++.old-deja/g++.abi/primary3.C: New test.
             * g++.old-deja/g++.abi/primary4.C: New test.
             * g++.old-deja/g++.abi/primary5.C: New test.
             * g++.old-deja/g++.abi/vtable3.h: New test.
             * g++.old-deja/g++.abi/vtable3a.C: New test.
             * g++.old-deja/g++.abi/vtable3b.C: New test.
             * g++.old-deja/g++.abi/vtable3c.C: New test.
             * g++.old-deja/g++.abi/vtable3d.C: New test.
             * g++.old-deja/g++.abi/vtable3e.C: New test.
             * g++.old-deja/g++.abi/vtable3f.C: New test.
             * g++.old-deja/g++.abi/vtable3g.C: New test.
             * g++.old-deja/g++.abi/vtable3h.C: New test.
             * g++.old-deja/g++.abi/vtable3i.C: New test.
             * g++.old-deja/g++.abi/vtable3j.C: New test.
             * g++.old-deja/g++.oliva/thunk1.C: Remove XFAIL.
     
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=1864&database=gcc
>From nobu.nokada@softhome.net Sun Apr 01 00:00:00 2001
From: nobu.nokada@softhome.net
To: gcc-gnats@gcc.gnu.org
Subject: optimization/2077: __builtin_frame_address(2) emits wrong codes.
Date: Sun, 01 Apr 2001 00:00:00 -0000
Message-id: <20010223224923.10630.qmail@sourceware.cygnus.com>
X-SW-Source: 2001-q1/msg01648.html
Content-length: 1191

>Number:         2077
>Category:       optimization
>Synopsis:       __builtin_frame_address(2) emits wrong codes.
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Feb 23 14:56:00 PST 2001
>Closed-Date:
>Last-Modified:
>Originator:     nobu.nokada@softhome.net
>Release:        gcc version 2.95.3 19991030 (prerelease)/ gcc version 2.95.2
>Organization:
>Environment:
Linux/ix86
cygwin/ix86
HP-UX
MIPS
>Description:
`__builtin_frame_address(2);' with `-fomit-frame-pointer' command line option emits wrong codes.

ex.

|void* foo(void)
|{
|    return __builtin_frame_address(2);
|}

This is compiled as followings, and causes SEGV or returns unknown value.

|	.file	"bug.c"
|	.version	"01.01"
|gcc2_compiled.:
|.text
|	.align 4
|.globl foo
|	.type	 foo,@function
|foo:
|	movl -12(%esp),%eax
|	movl (%eax),%eax
|	ret
|.Lfe1:
|	.size	 foo,.Lfe1-foo
|	.ident	"GCC: (GNU) 2.95.3 19991030 (prerelease)"

  Also, on some RISC architectures, I've heard it fails without `-fomit-frame-pointer'.
>How-To-Repeat:

>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted:

Re: c++/1864

[nathan]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2335 bytes --]

The following reply was made to PR c++/1864; it has been noted by GNATS.

From: nathan@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, snyder@fnal.gov
Cc:  
Subject: Re: c++/1864
Date: 5 Feb 2001 09:36:00 -0000

 Old Synopsis: critical
 New Synopsis: V3 ABI bug
 
 State-Changed-From-To: open->analyzed
 State-Changed-By: nathan
 State-Changed-When: Mon Feb  5 01:35:59 2001
 State-Changed-Why:
     I'm confirming this as a bug, but I'm fairly certain its
     related to an ABI defect, for which I'm about to commit a
     patch. I'll check when that's done.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=1864&database=gcc
>From wanderer@rsu.ru Sun Apr 01 00:00:00 2001
From: "Vladimir A. Merzliakov" <wanderer@rsu.ru>
To: nobody@gcc.gnu.org
Cc: gcc-prs@gcc.gnu.org
Subject: Re: libstdc++/2071: cin bug on solaris 2.6
Date: Sun, 01 Apr 2001 00:00:00 -0000
Message-id: <20010223182601.11490.qmail@sourceware.cygnus.com>
X-SW-Source: 2001-q1/msg01627.html
Content-length: 1322

The following reply was made to PR libstdc++/2071; it has been noted by GNATS.

From: "Vladimir A. Merzliakov" <wanderer@rsu.ru>
To: <Theodore.Papadopoulo@sophia.inria.fr>, <gcc-gnats@gcc.gnu.org>
Cc:  
Subject: Re: libstdc++/2071: cin bug on solaris 2.6
Date: Fri, 23 Feb 2001 21:18:51 +0300

 ----- Original Message -----
 From: <Theodore.Papadopoulo@sophia.inria.fr>
 To: <gcc-gnats@gcc.gnu.org>
 Sent: 23 ÆÅ×ÒÁÌÑ 2001 Ç. 20:59
 Subject: libstdc++/2071: cin bug on solaris 2.6
 
 
 >
 > >Number:         2071
 > >Category:       libstdc++
 > >Synopsis:       cin bug on solaris 2.6
 > >Confidential:   no
 > >Severity:       serious
 > >Priority:       medium
 > >Responsible:    unassigned
 > >State:          open
 > >Class:          sw-bug
 > >Submitter-Id:   net
 > >Arrival-Date:   Fri Feb 23 10:06:00 PST 2001
 > >Closed-Date:
 > >Last-Modified:
 > >Originator:     Theo Papadopoulo
 > >Release:        20010122
 > >Organization:
 > >Environment:
 > Solaris 2.6 machine with g++ 20010122
 > >Description:
 > The program below compiles successfully but requires
 > return to be typed twice to complete. This behaviour seems
 > specific to Solaris2.6 and dos not appear with linux
 > (same g++ version).
 
 
 This behaviour specific also to FreeBSD4.2
 (see closed but exist for me bug LIBSTDC++/1620)
 
 

Re: c++/1864

[nathan]

The following reply was made to PR c++/1864; it has been noted by GNATS.

From: nathan@gcc.gnu.org
To: gcc-gnats@gcc.gnu.org, nathan@gcc.gnu.org, nobody@gcc.gnu.org,
  snyder@fnal.gov
Cc:  
Subject: Re: c++/1864
Date: 5 Feb 2001 09:36:51 -0000

 Synopsis: V3 ABI bug
 
 Responsible-Changed-From-To: unassigned->nathan
 Responsible-Changed-By: nathan
 Responsible-Changed-When: Mon Feb  5 01:36:51 2001
 Responsible-Changed-Why:
     patch in progress
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=1864&database=gcc
>From rodrigc@mediaone.net Sun Apr 01 00:00:00 2001
From: Craig Rodrigues <rodrigc@mediaone.net>
To: nobody@gcc.gnu.org
Cc: gcc-prs@gcc.gnu.org
Subject: Re: c++/2017
Date: Sun, 01 Apr 2001 00:00:00 -0000
Message-id: <20010311164601.32576.qmail@sourceware.cygnus.com>
X-SW-Source: 2001-q1/msg02190.html
Content-length: 1132

The following reply was made to PR libstdc++/2017; it has been noted by GNATS.

From: Craig Rodrigues <rodrigc@mediaone.net>
To: Rafael Jesus Alcantara Perez <rafa@dedalo-ing.com>
Cc: Craig Rodrigues <rodrigc@mediaone.net>
Subject: Re: c++/2017
Date: Tue, 20 Feb 2001 12:29:13 -0500

 On Tue, Feb 20, 2001 at 10:24:48AM +0100, Rafael Jesus Alcantara Perez wrote:
 > Thanks, I have compiled GCC in a GLIBC 2.2 system, and that problem has
 > disappeared. Although, I have found another with streams (#2017).
 
 I looked at 2017.  I will get a more definitive answer, but
 from section 21.6.2 of The C++ Programming Language, 3rd ed. by Stroustrup:
 
 "Unfortunately backing up an input streams is not always possible....
  What is guaranteed is that you can back up one character after a successful
  read.  The peek() reads the next character but leaves
  it in the streambuf so that it can be read again."
 
 By the way, I tried your example under C++ compilers from
 Sun and Compaq, and they both produced the same result as
 gcc 3.0.
 -- 
 Craig Rodrigues        
 http://www.gis.net/~craigr    
 rodrigc@mediaone.net