Summary:
This patch is very closely related to D4859. Please see http://reviews.llvm.org/D4859 for more information.
This patch adds support for "fancy" pointers and allocators to promise and packaged_task. The changes made to support this are exactly the same as in D4859.
Test Plan: "fancy" pointer tests were added to each constructor affected by the change.
Reviewers: danalbert, mclow.lists
Reviewed By: mclow.lists
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4862
llvm-svn: 220471
Summary:
This patch add support for "fancy pointers/allocators" as well as fixing support for shared_pointer and "minimal" allocators.
Fancy pointers are class types that meet the NullablePointer requirements. In our case they are created by fancy allocators. `support/min_allocator.h` is an archetype for these types.
There are three types of changes made in this patch:
1. `_Alloc::template rebind<T>::other` -> `__allocator_traits_rebind<_Alloc, T>::type`. This change was made because allocators don't need a rebind template. `__allocator_traits_rebind` is used instead of `allocator_traits::rebind` because use of `allocator_traits::rebind` requires a workaround for when template aliases are unavailable.
2. `a.deallocate(this, 1)` -> `a.deallocate(pointer_traits<self>::pointer_to(*this), 1)`. This change change is made because fancy pointers aren't always constructible from raw pointers.
3. `p.get()` -> `addressof(*p.get())`. Fancy pointers aren't actually a pointer. When we need a "real" pointer we take the address of dereferencing the fancy pointer. This should give us the actual raw pointer.
Test Plan: Tests were added using `support/min_allocator.h` to each affected shared_ptr overload and creation function. These tests can only be executed in C++11 or greater since min_allocator is only available then. A extra test was added for the non-variadic versions of allocate_shared.
Reviewers: danalbert, mclow.lists
Reviewed By: mclow.lists
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4859
llvm-svn: 220469
Clang 3.6 no longer links the sanitizer runtime library dependancies when
-nodefaultlibs is used. This patch manually links in a missing dependancy.
llvm-svn: 220463
Summary:
This is the second attempt at allowing for the use of libraries that the linker cannot find. The first attempt used `CMAKE_LIBRARY_PATH` and `find_library` to select which ABI library should be used. There were a number of problems with this approach:
- `find_library` didn't work with cmake targets (ie in-tree libcxxabi build)
- It wasn't always possible to determine where `find_library` actually found your library.
- `target_link_libraries` inserted the path of the ABI library into libc++'s RPATH when `find_library` was used.
- Linking libc++ and it's ABI library is a special case. It's a lot easier to keep it simple.
After discussion with @cbergstrum a new approach was decided upon.
This patch achieve the same ends by simply using `LIBCXX_CXX_ABI_LIBRARY_PATH` to specify where to find the library (if the linker won't find it). When this variable is defined it is simply added as a library search path when linking libc++. It is a lot easier to duplicate this behavior in LIT. It also prevents libc++ from being linked with an RPATH.
Reviewers: mclow.lists, cbergstrom, chandlerc, danalbert
Reviewed By: chandlerc, danalbert
Subscribers: chandlerc, cfe-commits
Differential Revision: http://reviews.llvm.org/D5860
llvm-svn: 220157
When libcxx is built in-tree with libcxxabi it links against libcxxabi using
the name of the cmake target and not the actual library name. The cmake target
will not work with `find_library()`, so it needs special case handling.
llvm-svn: 220121
Summary:
This patch adds support for building/testing libc++ with an ABI library that the linker would not normally find.
- `CMAKE_LIBRARY_PATH` is used to specify the list of search directories.
- The ABI library is now found using `find_library` instead of assuming its along the linker's search path.
- `CMAKE_LIBRARY_PATH` is passed to our LIT config as `library_paths`.
- For each path in `library_paths` the following flags are added `-L<path> -Wl,-rpath -Wl,<path>`
Some changes in existing behavior were also added:
- `target_link_libraries` is now passed the ABI library file instead of the library name. Ex `target_link_libraries(cxx "/usr/lib/libc++abi.so")` vs `target_link_libraries(cxx "c++abi")`.
- `-Wl,-rpath -Wl,<path>` is now used on OSX to link to libc++ instead of env['DYLD_LIBRARY_PATH'] if `use_system_lib=False`.
Reviewers: mclow.lists, danalbert, EricWF
Reviewed By: EricWF
Subscribers: emaste, cfe-commits
Differential Revision: http://reviews.llvm.org/D5038
llvm-svn: 220118
Delay instantiation of `__numeric_type` within <cmath>,
don't instantiate it when the `is_arithmetic` conditions do not hold as it causes
errors with user-defined types with ambiguous conversions. Fixes PR21083.
llvm-svn: 219998
LLVM_USE_SANITIZER=Undefined support was added to the LLVM CMake configuration.
Update libc++'s handling of LLVM_USE_SANITIZER to support this as well.
llvm-svn: 219987
With clang, the header atomic requires __has_feature(cxx_atomic), which is only
true in c++11 mode. Because of this, when using modules in c++98 with libc++
compilation of the std module would fail without this change, PR21002.
(With gcc, only gcc4.7+ is needed, no c++11. But gcc doesn't have modules yet,
and the module.modulemap language can't express things like "this is only
required if the compiler is clang". If gcc gets module support, we'd probably
have a module.modulemap file for each compiler that libc++ supports?)
llvm-svn: 218372
Summary:
I changed the build so that each ABI header gets its own install rule. This gives us the flexibility to install different headers in different directories.
This also fixes the problem where libstdc++ bits/<header>'s were not being installed under a bits directory.
Test Plan: I tested this patch on linux against libstdc++ and libcxxabi.
Reviewers: danalbert, mclow.lists, jroelofs
Reviewed By: jroelofs
Subscribers: jhunold, cfe-commits
Differential Revision: http://reviews.llvm.org/D5454
llvm-svn: 218309
GCC 4.9 fails to inline these functions at -O1 because they are used
indirectly. Declare them as inline instead of always_inline. Discussion
in GCC bugreport: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63220
llvm-svn: 217961
These calls are allowed to fail spuriously.
29.6.5.25:
Remark: A weak compare-and-exchange operation may fail spuriously.
That is, even when the contents of memory referred to by expected
and object are equal, it may return false and store back to expected
the same memory contents that were originally there. [ Note: This
spurious failure enables implementation of compare and-exchange on a
broader class of machines, e.g., load-locked store-conditional
machines. A consequence of spurious failure is that nearly all uses
of weak compare-and-exchange will be in a loop.
To fix this, we replace any assert() that expects
std::atomic::compare_exchange_weak() to return true with a loop. If the
call does not return true within N runs (with N currently equal to 10),
then the test fails.
http://llvm.org/bugs/show_bug.cgi?id=20546
llvm-svn: 217319
If you're crazy enough to want this sort of thing, then add
-D_LIBCPP_HAS_NO_THREADS to your CXXFLAGS and
--param=additiona_features=libcpp-has-no-threads to your lit commnad line.
http://reviews.llvm.org/D3969
llvm-svn: 217271
This modifies the use_clang_verify parameter I added in r217009 to
only apply to tests that specifically ask for it via // USE_VERIFY.
This allows us to incrementally convert tests, but start enjoying the
benefits right away.
Suggested by Eric Fiselier in code review.
llvm-svn: 217017
Currently, failure tests work by checking that compilation exits 1.
This can lead to tests that fail for the wrong reason, so it'd be
preferable to convert them to check for specific errors.
This adds use_clang_verify parameter that runs failure tests using
clang's -verify flag. I'll convert some tests in subsequent commits,
and once all of the tests are converted we should key this on whether
cxx_under_test is clang.
I've also converted one of the unique.ptr tests, since it's the one
that motivated the idea of using clang -verify when possible in the
review of r216317.
llvm-svn: 217009
The way the standard currently specifies the default constructor for atomic<T>
requires T to be nothrow default constructible. This patch makes our test types
meet this requirement.
Note: The nothrow default constructible requirment is subject to the outcome of
LWG issue 1265.
llvm-svn: 216561
After discussing implementing more tests for this with @danalbert & @mclow, I
realized this change is not correct.
The C++ standard requires do_is() to behave as if it were a loop that checked
is(). Furthermore, it requires is() to check "The first form returns the result
of the expression (M & m) != 0; i.e., true if the character has the
characteristics specified"... which the reverted patch definitely does not
conform to. Even further, furthermore, this requires that ctype's mask be an
actual bitmask, unlike what android and newlib provide for _ctype_.
Fixing the original bug that instigated this patch remains TBD.
llvm-svn: 216508
Summary:
In order to gather more information about testsuite results these flags should be added to LIT's default args.
These new switches were recently added to LIT. It been more than two weeks since both switches were added.
I think its time we add these to our LIT flags.
Reviewers: mclow.lists, danalbert
Reviewed By: danalbert
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D5037
llvm-svn: 216315
The purely imperative format of the configuration performed in lit.cfg
was making merge conflicts with changes I have for Android an
unbelievable pain in the ass. I've moved all of the configuration into a
Configuration class, with each piece of configuration happening in a
different method. This way I can avoid merge conflicts, and any new
features that get added (as have been done with the sanitizers, the -std
flag, etc.) can be easily applied to Android as well.
Reviewers: mclow.lists, EricWF
Reviewed By: EricWF
Differential Revision: http://reviews.llvm.org/D4952
llvm-svn: 216196
This marks some of the localization test XFAIL on linux.
There has been some discussion on D4861 about doing this.
Please let me know if any of these tests for you on linux.
llvm-svn: 216151
This marks some of the localization test XFAIL on linux.
There has been some discussion on D4861 about doing this.
Please let me know if any of these tests for you on linux.
llvm-svn: 216150
This marks some of the localization test XFAIL on linux.
There has been some discussion on D4861 about doing this.
Please let me know if any of these tests for you on linux.
llvm-svn: 216148
Summary:
This patch adds support for // UNSUPPORTED: feature. If an excluded feature is found in the list of available features then the test is marked unsupported.
I hope to use this to mark test unsupported if the fail with msan on asan. As well as tests that fail in particular standard modes.
Reviewers: mclow.lists, danalbert
Reviewed By: danalbert
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4950
llvm-svn: 215883
Summary:
Instead of having to update test results for different platforms it would be helpful to just supply links to buildbots that build+test libcxx.
For the FreeBSD bots Pawel Worach gave me the OK to link to them.
Reviewers: mclow.lists, danalbert
Reviewed By: danalbert
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4947
llvm-svn: 215882
Summary:
This patch adds support for LLVM_USE_SANITIZER when being built in-tree and standalone.
This patch does the following things:
1. define the LLVM_USE_SANITIZER option to "" when being built standalone. This also helps show we support it.
2. Translate LLVM_USE_SANITIZER when standalone in a very similar way done in llvm/cmake/HandleLLVMOptions.cmake.
3. Add config.llvm_use_sanitizer to lit.site.cfg.in
4. Add code to translate config.llvm_use_sanitizer's value into the needed compile flags in lit.cfg.
Currently lit.cfg assumes that that the compiler supports '-fno-omit-frame-pointer' while CMakeLists.txt actually checks to see if its supported. We could pass this information to lit but I'm not sure its needed.
Reviewers: mclow.lists, danalbert
Reviewed By: danalbert
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4949
llvm-svn: 215872
Summary:
This patch does two things:
CMake Update:
- Add compiler flag checks for -std=c++11 and -std=c++1y and remove check for -std=c++0x.
- Add configuration option LIBCXX_ENABLE_CXX1Y to prevent/allow -std=c++1y from being chosen as the std version. LIBCXX_ENABLE_CXX1Y is set to OFF by default.
- if LIBCXX_ENABLE_CXX1Y is enabled then set LIBCXX_STD_VERSION to c++1y and fail if the compiler does not support -std=c++1y
- If c++1y is not enabled then use c++11 and fail if the compiler does not support c++11.
Lit Update:
- Update lit.site.cfg.in to capture LIBCXX_STD_VERSION information as config.std.
- Remove mentions of has_cxx0X configuration option.
- Check for `--param std=X' passed to lit on the command line.
- Choose the std for the tests either from command line parameter or (if it doesn't exist) the lit.site.cfg.
Reviewers: mclow.lists, danalbert
Reviewed By: danalbert
Subscribers: emaste, rnk, ajwong, danalbert, cfe-commits
Differential Revision: http://reviews.llvm.org/D4329
llvm-svn: 215802
Add initial support for using LIT to run the tests on FreeBSD.
More work may need to be done to add support for FreeBSD but this is a
good initial step.
llvm-svn: 215742
Turning off explicit template instantiation leads to a pretty
significant build time and code size cost. We're better off dealing
with ABI incompatibility issues that come up in a less heavy handed
way.
This reverts commit r189610.
llvm-svn: 215740
The build file was trying to use LIBCXX_VERSION, which isn't set
anywhere, and also forgot to split the 'compile_flags' list.
Differential Revision: http://reviews.llvm.org/D4860
llvm-svn: 215463
This patch removes the use of the "%c" specifier for getting/setting times.
The semantics of this specifier differ between linux and Mac. I don't believe
the use of this specifier was important to the test.
The following tests now pass on linux.
test/input.output/iostream.format/ext.manip/get_time.pass.cpp
test/input.output/iostream.format/ext.manip/put_time.pass.cpp
llvm-svn: 215417
This patch just adds the required return statements to slice_array::operator=
and mask_array::operator=.
Tests were added to check that the return value is the same as the object assigned
to.
llvm-svn: 215414
Okay, so this actually does more than just that. I've rearranged most of
the information on the page to try to make it more helpful and flow
better. Essentially, the differences between Mac and Linux, the various
ABI libraries, and in-tree versus out-of-tree builds were cluttering
things. To clean up, I've done the following:
* Only describe the cmake process. buildit doesn't work out of the box
on Linux, and we need to stop having duplicates for every process.
* Use libc++abi for the default instructions. This works on the major
platforms.
* Describe both in-tree and out-of-tree builds. Previously it wasn't
clear that in-tree builds were even possible for libc++.
* Separate the documentation about using libc++ from that about
building and testing libc++.
llvm-svn: 215358
Things done in this patch:
1. Make __debug include __config since it uses macros from it.
2. The current method of defining _LIBCPP_ASSERT is prone to redefinitions. Move
the null _LIBCPP_ASSERT definition into the __debug header to prevent this.
3. Remove external <__debug> include gaurds. <__debug> guards almost all of its
contents internally. There is no reason to be doing it externally.
This patch should not change any functionality.
llvm-svn: 215332
gcc 4.7 and above has atomic built-ins which slightly different APIs
from those provided by clang. Add proxy functions that wrap the gcc
built-ins to produce a symbol that is API equivalent to the clang
built-ins. This allows libc++'s atomic library to be used with gcc-4.7
and newer.
Patch contributed by Albert Wong.
llvm-svn: 215305
Some tests were constructing it with 0, so use -1 as the invalid state
instead.
Reviewers: Marshall Clow
Differential Revision: http://reviews.llvm.org/D4095
llvm-svn: 215301
These tests were marked as failing because the libc implementation does not
use the proper return type on certain math functions. There is other weirdness
that prevents us from defining our own version of these functions.
The failing tests in cmake were separated into their own files so that the rest
failures in the rest of the cmath tests were not hidden.
This was not done for the math.h test since we don't actually supply math.h
llvm-svn: 215275
Linux has a lot of failures caused by not having support for certain
locales. Since these come out as a lot of noise in the test results,
have lit.cfg detect the presence of the various locales used in the
tests and add them to config.available_features as locale.LOCALE_NAME.
This patch also adds REQUIRES: locale.REQUIRED_LOCALE to every test that
I saw failing in this manner. We probably need to add more for all the
tests requiring en_US.UTF-8, but we can do that on an as-needed basis.
One thing that concerns me is how many tests get skipped because of
missing locales (especially in regex/). We should make a point of
splitting up any tests that test default behavior _and_ behavior under a
given locale so that we aren't losing coverage for default behavior.
llvm-svn: 214753
Summary: This patch moves the SFINAE for __is_destructor_welformed out of the function template parameters. type_traits must compile in c++03 mode since it is included in c++03 headers.
Test Plan: No tests have been added.
Reviewers: danalbert, mclow.lists
Reviewed By: danalbert
Subscribers: K-ballo, cfe-commits
Differential Revision: http://reviews.llvm.org/D4735
llvm-svn: 214422
__get_classname() and __bracket_expression were assuming that
char_class_type was ctype_base::mask rather than using
regex_traits<_CharT>::char_class_type.
This change allows char_class_type to be defined to something other than
ctype_base::mask so that the implementation will still work for
platforms with an 8-bit ctype mask (such as Android and OpenBSD).
llvm-svn: 214201
Pass target name ("cxxabi") to target_link_libraries(cxx ...) to ensure
that linker is able to locate just-built libc++abi in the build tree,
instead of relying on "-lc++abi" linker flag.
llvm-svn: 214114
This will detect if you are building libcxx in-tree and libcxxabi is
available. If so, it will default to using the in-tree libcxxabi by
setting LIBCXX_CXX_ABI to "libcxxabi", LIBCXX_LIBCXXABI_INCLUDE_PATHS to
"${CMAKE_SOURCE_DIR}/projects/libcxxabi/include" and will add "cxxabi"
as a proper dependency.
Patch by Russell Harmon.
llvm-svn: 214037
Summary: The polymorphic allocator implementation would greatly benefit by defining virtual functions in the dynlib instead of inline. In order to do that some types are going to have to be available outside of c++1y. This is the first step.
Reviewers: mclow.lists, EricWF
Reviewed By: EricWF
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4554
llvm-svn: 213889
std::make_heap is currently implemented by iteratively applying a
siftup-type algorithm. Since sift-up is O(ln n), this gives
std::make_heap a worst case time complexity of O(n ln n).
The C++ standard mandates that std::make_heap make no more than O(3n)
comparisons, this makes our std::make_heap out of spec.
Fix this by introducing an implementation of __sift_down and switch
std::make_heap to create the heap using it.
This gives std::make_heap linear time complexity in the worst case.
This fixes PR20161.
llvm-svn: 213615
Summary:
This patch adds the `<experimental/utility>` header as specified in the latest draft of the library fundamentals TS.
`<experimental/utility>` only contains `class erased_type`.
This patch also updates the documentation to list the `erased_type` class as "initial implementation complete".
Test Plan:
Three test cases where added:
1. Test that `_LIBCPP_VERSION` is defined.
2. Test that `<utility>` has been included.
3. Test that `erased_type` is in the correct namespace and is constexpr default constructible.
Reviewers: mclow.lists
Reviewed By: mclow.lists
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D4510
llvm-svn: 213226
Mark the base classes for time_get_byname and time_get as _LIBCPP_TYPE_VIS_ONLY
rather than _LIBCPP_TYPE_VIS. These base classes are templated types and cannot
be stored with export dll storage.
Fixes compilation with _LIBCPP_DLL for Windows when the time_get and
time_get_byname classes are used.
llvm-svn: 213116
Signals may result in nanosleep returning with only some of the
requested sleeping performed.
Utilize nanosleep's "time-remaining" out parameter to continue sleeping
when this occurs.
llvm-svn: 210210
libc++ currently relies on undefined initialization order of global
initializers when using gcc:
1. __start_std_streams in iostream.cpp calls locale:🆔:_init, which assigns
an id to each locale::facet in an initializer
2. Every facet has a static locale::id id, whose constructor sets the facet's
id to 0
If 2 runs after 1, it clobbers the facet's assigned consecutive id, causing
exceptions to be thrown when e.g. running code like "cout << endl".
To fix this, let _LIBCPP_CONSTEXPR evaluate to "constexpr" instead of nothing
with gcc. locale::id's constructor is marked _LIBCPP_CONSTEXPR, which ensures
that it won't get an initializer that could potentially run after the
iostream.cpp initializer. (This remains broken when building with msvc.)
Also switch constexpr-specific code in bitset to use __SIZEOF_SIZE_T__ instead
of __SIZE_WIDTH__, because gcc doesn't define the latter.
Pair-programmed/debugged with Dana Jansens.
llvm-svn: 210188
Make sure we appropriately retry calls to read if the return result is
less than what we asked for.
Additionally, check and handle IO errors: EINTR results in the read
operation getting restarted; other errors turn into exceptions.
llvm-svn: 210061
random_device::random_device(const string&) wrongly assumes that open
can only validly return a file descriptor greater than zero.
This results in random_device believing that it didn't successfully open
the device causing it to throw in it's constructor, this ends up leaking
a file descriptor.
The fix is simple, don't error on file descriptors which are zero.
llvm-svn: 210060
Other tests in this directory use this type, so it's probably copypasta from
there.
(test_buf only forwards to the superclass in all tests where it's used though,
so I wonder if it can be replaced with just using filebuf / wfilebuf
everywhere?)
llvm-svn: 210019
(clang doesn't complain about this, but gcc does. This is necessary for a
follow-up patch that will enable _LIBCPP_CONSTEXPR for gcc.)
llvm-svn: 209888
[syserr.errcat.objects]p4 specifies that
system_category().default_error_condition(ev) map to
error_condition(posv, generic_category()) if ev could map to a POSIX
errno.
Linux reserves up to and including 4095 for errno values, use this as a
bound.
This fixes syserr.errcat.objects/system_category.pass.cpp on Linux.
llvm-svn: 209795
r207606 changed the __need_foo macros to behave like they do with gcc: If they
are set, _only_ the __need_foo stuff gets defined. As a consequence, cstddef
no longer defined "offsetof". It looks like the __need_foo defines aren't
needed anymore, so just remove them.
Fixes PR19723.
llvm-svn: 208942
When testing against the system library, there is a relatively minor
ABI breakage that the std::future_errc values have been changed to
avoid using zero. Update the tests that rely on the values being
consistent.
llvm-svn: 208840
These tests haven't been failing on darwin11 or 12 since r189610 when
pr17027 was fixed, but they've been keeping the libc++ bot red by
XPASSing since then.
llvm-svn: 208831
compatibility to libstdc++. Move the implementation into a header for
easier sharing with libc++abi. Merge a number of improvements from that
version. Provide a POD definition for <stdexcept>'s public use to avoid
cast dances. Discussed with Marshall Clow.
llvm-svn: 207695
libc++ will not call address_sanitizer to detect addressing errors in the
standard library containers. This is a negative macro to enable users to
disable the libc++ checks even if they are compiling with address sanitizer
enabled by defining this macro.
At the present time, there is no code in libc++ that looks at this macro.
That will come soon. This is just infrastructure.
llvm-svn: 206184
It's identical to name() these days. (At one point it avoided masking
of the RTTI uniqueness bit because ARM64 ignored it architecturally,
but no longer).
llvm-svn: 205518
ARM64 generates RTTI with hidden visibility, which means that typeinfo
must be compared char-by-char since it's not guaranteed to be uniqued
across the whole program.
llvm-svn: 205139
to remove redundant wording, which required no changes to libc++. 2075 was a
rewrite of the requirements for forward progress, and again, requires no changes
to the library.
llvm-svn: 204724
Also mark #2104 as complete. Leave the implementation in libc++ as noexcept, since
implementations are allowed to add noexcept to non-virtual calls. If we throw from
unique_lock& operator=(unique_lock&& u), then that means the preconditions were violated,
and calling terminate() (as a result of throwing from a noexcept function) is as
good example of undefined behavior as any other.
llvm-svn: 204653
This is as straightforward as it sounds, a renamed from shared_mutex to
shared_timed_mutex.
Note that libcxx .dylib and .so files built with c++14 support need to
be rebuilt.
llvm-svn: 204078
My fix for PR15820 in r180132 inadvertently removed the exclusion for ".*".
This puts it back again. Thanks for Nico Weber for pointing this out!
llvm-svn: 203807
This is unnecessary now that the flag handling has been fixed. The flags will
be added properly in the main CMakeLists.txt after the config-ix inclusion which
performs the required check.
llvm-svn: 203639
Always use list(APPEND) as it will perform the desired action even if the list
is empty or previously unset. The first set is harmless, however, the
subsequent set was overwriting the previous flag setup resulting in an improper
compilation command being generated. This manifested as a build failure on
Linux when using cmake + ninja.
llvm-svn: 203638
add_definitions is meant for adding C preprocessor definitions. Modern cmake
suggests use of the CMAKE_CXX_FLAGS for the purposes of pushing flags to the
compilation commands. Simply switch to the modern form given that we are
already requiring a new enough cmake.
llvm-svn: 203637
Instead, define explicit specializations for the basic types listed in
the SGI documentation. This solves two problems:
1) Helps avoid silent ODR violations caused by the absence of a
user-supplied __gnu_cxx::hash specialization in cases where a std::hash
specialization exists (e.g. for std::string).
2) __gnu_cxx::hash semantics are slightly different to those of
std::hash (for example, the former may dereference a pointer argument)
so it is inappropriate for __gnu_cxx::hash to receive std::hash
specializations by default.
Differential Revision: http://llvm-reviews.chandlerc.com/D2747
llvm-svn: 203070
Eric Fiselier