Use version namespace in normal mode

[Jonathan Wakely]

Some quick comments before I go offline ...


On 21/10/16 21:21 +0200, François Dumont wrote:
>Hi
>
>    I configured libstdc++ to use gnu-version-namespace and there are 
>a number of failures, see below. But none of them related to this 
>patch so is it ok to commit ?
>
>The results:
>
>FAIL: libstdc++-abi/abi_check
>
>    3709 symbols reported as added. I don't know what to think about 
>it. I see a gnu-versioned-namespace.ver in config/abi/pre, is it the 
>list of symbols to support when versioned namespace is activated ? The 
>list looks pretty limited.

Because everything gets matched by widlcards and added to a single
symbol version.

I just ignore abi-check for the versioned namespace, since it's
explicitly not ABI compatible. Does the test even work for the
versioned namespace or does it only use the normal baselines?


>FAIL: 18_support/headers/limits/synopsis.cc (test for excess errors)
>FAIL: 20_util/headers/functional/synopsis.cc (test for excess errors)
>FAIL: 20_util/headers/memory/synopsis.cc (test for excess errors)
>FAIL: 20_util/headers/utility/synopsis.cc (test for excess errors)
>FAIL: 21_strings/headers/string/synopsis.cc (test for excess errors)
>FAIL: 22_locale/headers/locale/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/bitset/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/deque/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/forward_list/synopsis.cc (test for excess 
>errors)
>FAIL: 23_containers/headers/list/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/map/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/queue/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/set/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/stack/synopsis.cc (test for excess errors)
>FAIL: 23_containers/headers/vector/synopsis.cc (test for excess errors)
>FAIL: 24_iterators/headers/iterator/synopsis.cc (test for excess errors)
>FAIL: 24_iterators/headers/iterator/synopsis_c++11.cc (test for excess 
>errors)
>FAIL: 24_iterators/headers/iterator/synopsis_c++14.cc (test for excess 
>errors)
>FAIL: 24_iterators/headers/iterator/synopsis_c++17.cc (test for excess 
>errors)
>FAIL: 26_numerics/headers/complex/synopsis.cc (test for excess errors)
>FAIL: 26_numerics/headers/valarray/synopsis.cc (test for excess errors)
>FAIL: 27_io/headers/fstream/synopsis.cc (test for excess errors)
>FAIL: 27_io/headers/ios/synopsis.cc (test for excess errors)
>FAIL: 27_io/headers/istream/synopsis.cc (test for excess errors)
>FAIL: 27_io/headers/ostream/synopsis.cc (test for excess errors)
>FAIL: 27_io/headers/sstream/synopsis.cc (test for excess errors)
>FAIL: 27_io/headers/streambuf/synopsis.cc (test for excess errors)
>FAIL: tr1/2_general_utilities/headers/functional/synopsis.cc (test for 
>excess errors)
>FAIL: tr1/2_general_utilities/headers/memory/synopsis.cc (test for 
>excess errors)
>FAIL: tr1/3_function_objects/headers/functional/synopsis.cc (test for 
>excess errors)
>FAIL: tr1/4_metaprogramming/headers/type_traits/synopsis.cc (test for 
>excess errors)
>FAIL: tr1/6_containers/headers/array/synopsis.cc (test for excess errors)
>FAIL: tr1/6_containers/headers/unordered_map/synopsis.cc (test for 
>excess errors)
>FAIL: tr1/6_containers/headers/unordered_set/synopsis.cc (test for 
>excess errors)
>
>    All those failures are coming from declaration or explicit 
>instantiations of template types expected to be in std but being in 
>fact in std::__7. Should I add usage of 
>_GLIBCXX_BEGIN_NAMESPACE_VERSION/_GLIBCXX_END_NAMESSPACE_VERSION in 
>those files ? Or introduce a dg-require-no-versioned-namespace ?

I've just been ignoring those failures, as the reason is known. Either
of your suggestions would work, although I've been thinking we should
avoid using _GLIBCXX_ macros in the tests, so they are independent of
our implementation details.

We could define GLIBCXX_TEST_ macros for use in the tests, and define
them independently, so we could add a GLIBCXX_TEST_INLINE_NS to those
tests which would add the "inline namespace __7" bit.

We don't need to worry about it for now though.


>FAIL: 17_intro/using_namespace_std_tr1_neg.cc  (test for errors, line 65)
>FAIL: 21_strings/basic_string/cons/char/7.cc (test for excess errors)
>FAIL: 21_strings/basic_string/cons/wchar_t/7.cc (test for excess errors)
>FAIL: 21_strings/basic_string/lwg2758.cc (test for excess errors)
>FAIL: 21_strings/basic_string/modifiers/append/char/4.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/append/wchar_t/4.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/assign/char/4.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/assign/wchar_t/4.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/insert/char/3.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/insert/wchar_t/3.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/replace/char/7.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/modifiers/replace/wchar_t/7.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/operations/compare/char/2.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/operations/compare/wchar_t/2.cc (test 
>for excess errors)
>FAIL: 21_strings/basic_string/operations/find/char/5.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/operations/find/wchar_t/5.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string/operators/char/5.cc (test for excess errors)
>FAIL: 21_strings/basic_string/operators/wchar_t/5.cc (test for excess 
>errors)
>FAIL: 21_strings/basic_string_view/cons/char/1.cc (test for excess errors)
>FAIL: 21_strings/basic_string_view/cons/wchar_t/1.cc (test for excess 
>errors)
>FAIL: 21_strings/basic_string_view/inserters/char/1.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string_view/inserters/char/2.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string_view/inserters/char/3.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string_view/inserters/wchar_t/1.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string_view/inserters/wchar_t/2.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string_view/inserters/wchar_t/3.cc (test for 
>excess errors)
>FAIL: 21_strings/basic_string_view/operations/string_conversion/1.cc 
>(test for excess errors)
>FAIL: 23_containers/map/modifiers/erase/abi_tag.cc scan-assembler _ZNSt3mapIiiSt4lessIiESaISt4pairIKiiEEE5eraseB5cxx11ESt17_Rb_tree_iteratorIS4_E
>FAIL: 23_containers/multimap/modifiers/erase/abi_tag.cc scan-assembler _ZNSt8multimapIiiSt4lessIiESaISt4pairIKiiEEE5eraseB5cxx11ESt17_Rb_tree_iteratorIS4_E
>FAIL: 23_containers/multiset/modifiers/erase/abi_tag.cc scan-assembler _ZNSt8multisetIiSt4lessIiESaIiEE5eraseB5cxx11ESt23_Rb_tree_const_iteratorIiE
>FAIL: 23_containers/multiset/modifiers/erase/abi_tag.cc scan-assembler _ZNSt8multisetIiSt4lessIiESaIiEE5eraseB5cxx11ESt23_Rb_tree_const_iteratorIiES5_
>FAIL: 23_containers/set/modifiers/erase/abi_tag.cc scan-assembler _ZNSt3setIiSt4lessIiESaIiEE5eraseB5cxx11ESt23_Rb_tree_const_iteratorIiE
>FAIL: 23_containers/set/modifiers/erase/abi_tag.cc scan-assembler _ZNSt3setIiSt4lessIiESaIiEE5eraseB5cxx11ESt23_Rb_tree_const_iteratorIiES5_
>FAIL: 26_numerics/complex/abi_tag.cc scan-assembler 
>_ZNKSt7complexIfE4realB5cxx11Ev
>
>    Those failures are due to dual abi being disable when versioned 
>namespace is being used, in configure:
>
>  if test x$enable_symvers = xgnu-versioned-namespace; then
>    # gnu-versioned-namespace is incompatible with the dual ABI.
>    enable_libstdcxx_dual_abi="no"
>  fi
>
>    Should I add a dg-require-cxx11-abi ?

I have a patch to do that already, but it needs a bit more work.


>FAIL: 20_util/allocator/overaligned.cc (test for excess errors)
>FAIL: ext/bitmap_allocator/overaligned.cc (test for excess errors)
>FAIL: ext/mt_allocator/overaligned.cc (test for excess errors)
>FAIL: ext/new_allocator/overaligned.cc (test for excess errors)
>FAIL: ext/pool_allocator/overaligned.cc (test for excess errors)
>
>Failures coming from always the same undefined symbols:
>/home/fdt/dev/gcc/build_git/x86_64-pc-linux-gnu/libstdc++-v3/include/ext/new_allocator.h:108: 
>undefined reference to `operator new(unsigned long, std::align_val_t)'
>
>    I don't understand those failures for now but same failures 
>without patch so unrelated.

Those new symbols need to be added to the linker script,
gnu-versioned-namespace.ver

>FAIL: libstdc++-prettyprinters/cxx11.cc print efl
>...
>
>Of course many pretty printers failures. IMO should be set as 
>unsupported in this mode but I don't know how to do so.

No, we should fix the printers (they already try to work, but have
bitrotted).