[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #11 from jakub at gcc dot gnu dot org 2008-02-27 09:09 --- Subject: Bug 35368 Author: jakub Date: Wed Feb 27 09:09:05 2008 New Revision: 132704 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=132704 Log: PR c++/35368 * rtti.c: Include c-pragma.h. (push_abi_namespace, pop_abi_namespace): New functions. (build_dynamic_cast_1, tinfo_base_init, get_pseudo_ti_index, create_tinfo_types, emit_support_tinfos): Use them. * Make-lang.in (cp/rtti.o): Depend on $(C_PRAGMA_H). * g++.dg/ext/visibility/typeinfo2.C: New test. * g++.dg/ext/visibility/typeinfo3.C: New test. Added: trunk/gcc/testsuite/g++.dg/ext/visibility/typeinfo2.C trunk/gcc/testsuite/g++.dg/ext/visibility/typeinfo3.C Modified: trunk/gcc/cp/ChangeLog trunk/gcc/cp/Make-lang.in trunk/gcc/cp/rtti.c trunk/gcc/testsuite/ChangeLog -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #12 from jakub at gcc dot gnu dot org 2008-02-27 09:27 --- Subject: Bug 35368 Author: jakub Date: Wed Feb 27 09:26:24 2008 New Revision: 132705 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=132705 Log: PR c++/35368 * rtti.c: Include c-pragma.h. (push_abi_namespace, pop_abi_namespace): New functions. (build_dynamic_cast_1, tinfo_base_init, get_pseudo_ti_index, create_tinfo_types, emit_support_tinfos): Use them. * Make-lang.in (cp/rtti.o): Depend on $(C_PRAGMA_H). * g++.dg/ext/visibility/typeinfo2.C: New test. * g++.dg/ext/visibility/typeinfo3.C: New test. Added: branches/gcc-4_3-branch/gcc/testsuite/g++.dg/ext/visibility/typeinfo2.C branches/gcc-4_3-branch/gcc/testsuite/g++.dg/ext/visibility/typeinfo3.C Modified: branches/gcc-4_3-branch/gcc/cp/ChangeLog branches/gcc-4_3-branch/gcc/cp/Make-lang.in branches/gcc-4_3-branch/gcc/cp/rtti.c branches/gcc-4_3-branch/gcc/testsuite/ChangeLog -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #5 from jakub at gcc dot gnu dot org 2008-02-26 16:30 --- Testing a patch. -- jakub at gcc dot gnu dot org changed: What|Removed |Added AssignedTo|unassigned at gcc dot gnu |jakub at gcc dot gnu dot org |dot org | Status|NEW |ASSIGNED Last reconfirmed|2008-02-25 15:54:38 |2008-02-26 16:30:23 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #6 from mmitchel at gcc dot gnu dot org 2008-02-26 17:06 --- We need to be careful about this. We have a lot of ways to specify visibility: dllimport/dllexport attributes, notshared attribute, visibility attributes on classes. I actually think the compiler is behaving as intended here. To me: #pragma GCC push visibility(hidden) is equivalent to applying: __attribute__((visibility(hidden))) to all declarations in the scope of the #pragma. And, a class with hidden visibility does have a hidden virtual table. Why do we think this is a bug? -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #7 from jakub at gcc dot gnu dot org 2008-02-26 17:19 --- Created an attachment (id=15233) -- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15233action=view) gcc43-pr35368.patch Patch I'm testing. The reasons it should IMNSHO have default visibility is: 1) these are compiler generated references and libsupc++.a/libstdc++.a provided symbols. For e.g. builtins we use default visibility as well, rather than the current default one 2) the same reason why do the typeinfo etc. headers have those #pragma GCC visibility push(default)/#pragma GCC visibility pop around them -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #8 from benjamin at smedbergs dot us 2008-02-26 17:25 --- Yes, to make it clear: the class typeinfo object may have hidden visibility... it's the __cxxabiv1::__class_type_info class that should have default visibility always. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #9 from jakub at gcc dot gnu dot org 2008-02-26 17:39 ---
BTW, not sure why 4.1.x/4.2.x is listed as broken. Only 4.3+ has H.J's:
http://gcc.gnu.org/viewcvs?root=gccview=revrev=119764
Without that change, although __cxxabiv1::* symbols are incorrectly marked
as hidden, GCC doesn't emit .hidden directives for the external symbols and all
these symbols of course are external, as they are defined in
libsupc++.a/libstdc++.{so,a}, and as they are referenced just in the RTTI
pointers, not in code directly, it makes zero difference to the generated code
whether they are hidden or not.
BTW, I've noticed that for _ZTI*/_ZTV* symbols defined in the assembly .hidden
directives are emitted twice, once at the definition spot and once at the end
of the file. Guess assemble_external_real should skip decls that lost
DECL_EXTERNAL flag (became defined in the current TU).
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #10 from mark at codesourcery dot com 2008-02-26 17:57 --- Subject: Re: [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation benjamin at smedbergs dot us wrote: --- Comment #8 from benjamin at smedbergs dot us 2008-02-26 17:25 --- Yes, to make it clear: the class typeinfo object may have hidden visibility... it's the __cxxabiv1::__class_type_info class that should have default visibility always. Oh, I see! Yes, __cxxabiv1::* should definitely have default visibility. Thanks, -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #3 from rguenth at gcc dot gnu dot org 2008-02-25 15:54 --- Doens't work for me with 4.1.3 20080114, 4.2.3 or 4.3.0 20080219. Works with 4.0.4 20060904. The bogus reference is here: .LEFDE5: .hidden DW.ref.__gxx_personality_v0 .weak DW.ref.__gxx_personality_v0 .section .data.DW.ref.__gxx_personality_v0,awG,@progbits,DW.ref.__gxx_personality_v0,comdat .align 4 .type DW.ref.__gxx_personality_v0, @object .size DW.ref.__gxx_personality_v0, 4 DW.ref.__gxx_personality_v0: .long __gxx_personality_v0 .hidden _ZTVN10__cxxabiv117__class_type_infoE .hidden _ZTI1A .hidden _ZTV1A Maybe related to PR19159, but that is for systems w/o weak support. This is a regression on the 4.1/4.2 branches. -- rguenth at gcc dot gnu dot org changed: What|Removed |Added CC||mmitchel at gcc dot gnu dot ||org Status|UNCONFIRMED |NEW Ever Confirmed|0 |1 Keywords||wrong-code Known to fail|4.3.0 |4.1.3 4.2.3 4.3.0 Known to work|4.2.1 |4.0.4 4.2.1 Priority|P3 |P1 Last reconfirmed|-00-00 00:00:00 |2008-02-25 15:54:38 date|| Summary|With #pragma visibility,|[4.1/4.2/4.3/4.4 Regression] |`vtable for |With #pragma visibility, |__cxxabiv1::__class_type_inf|`vtable for |o' is emitted as a hidden- |__cxxabiv1::__class_type_inf |visibility relocation |o' is emitted as a hidden- ||visibility relocation Target Milestone|--- |4.1.3 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
[Bug c++/35368] [4.1/4.2/4.3/4.4 Regression] With #pragma visibility, `vtable for __cxxabiv1::__class_type_info' is emitted as a hidden-visibility relocation
--- Comment #4 from matz at gcc dot gnu dot org 2008-02-25 16:18 --- For me it works with gcc42-4.2.1_20070724-17.1 (i686). With 4.3 it breaks. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35368
