Code on Windows expects to be able to do:
#define _USE_MATH_DEFINES
#include <math.h>
and receive the definitions of mathematical constants, even if <math.h>
has previously been included. To support this scenario, re-include
<math.h> every time the wrapper header is included.
Differential Revision: https://reviews.llvm.org/D42403
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323490 91177308-0d34-0410-b5e6-96231b3b80d8
There was a bug in the implementation of splice where the container
sizes were updated before decrementing one of the iterators. Afterwards,
the result of decrementing the iterator was flagged as UB by the debug
implementation because the container was reported to be empty.
This patch fixes that bug by delaying the updating of the container
sizes until after the iterators have been correctly constructed.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323390 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
See https://bugs.llvm.org/show_bug.cgi?id=20855
Libc++ goes out of it's way to diagnose `std::tuple` constructions which are UB due to lifetime bugs caused by reference creation. For example:
```
// The 'const std::string&' is created *inside* the tuple constructor, and its lifetime is over before the end of the constructor call.
std::tuple<int, const std::string&> t(std::make_tuple(42, "abc"));
```
However, we are over-aggressive and we incorrectly diagnose cases such as:
```
void foo(std::tuple<int const&, int const&> const&);
foo(std::make_tuple(42, 42));
```
This patch fixes the incorrectly diagnosed cases, as well as converting the diagnostic to use the newly added Clang trait `__reference_binds_to_temporary`. The new trait allows us to diagnose cases we previously couldn't such as:
```
std::tuple<int, const std::string&> t(42, "abc");
```
Reviewers: rsmith, mclow.lists
Reviewed By: rsmith
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D41977
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323380 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Currently when a regular expression contains an invalid character
class name std::regex constructors throw an std::regex_error with
std::regex_constants::error_brack code.
This patch changes the code to std::regex_constants::error_ctype and
adds a test.
Reviewers: EricWF, mclow.lists
Reviewed By: mclow.lists
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D42291
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323322 91177308-0d34-0410-b5e6-96231b3b80d8
Some users may have a custom build system which gives a different
name to the libc++ archive (or does not create an archive at all,
instead passing the object files directly to the linker). Give those
users a way to disable auto-linking.
Differential Revision: https://reviews.llvm.org/D42436
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323300 91177308-0d34-0410-b5e6-96231b3b80d8
The language standard does not define a function with this name,
so it is part of the user's namespace. This change fixes a duplicate
symbol error that occurs when a user attempts to define a function
with this name.
Differential Revision: https://reviews.llvm.org/D42405
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323237 91177308-0d34-0410-b5e6-96231b3b80d8
When CMAKE_SYSROOT or CMAKE_FIND_ROOT_PATH is set, cmake
recommends setting CMAKE_FIND_ROOT_PATH_MODE_INCLUDE=ONLY
globally which means find_path() always prepends CMAKE_SYSROOT or
CMAKE_FIND_ROOT_PATH to all paths used in the search.
However, this find_path() invocation is looking for a path in the
libcxxabi project on the host system, not the target system,
which can be done by passing NO_CMAKE_FIND_ROOT_PATH.
Differential Revision: https://reviews.llvm.org/D41622
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@323143 91177308-0d34-0410-b5e6-96231b3b80d8
This patch fixes almost all currently failing tests when
using GCC ToT.
The specific changes are:
(A) Workaround gcc.gnu.org/PR83921 which rejects variables w/o initializers
in constexpr contexts -- even when the variable is an empty class. This
bug has been worked around at all callsites by adding an initializer.
Additionally a new test, constexpr_init.pass.cpp, has been added to
test that Clang doesn't suffer from these bugs.
(B) Fix streambuf.assign/swap.pass.cpp. This test was never actually
calling the swap method as intended. In fact, the swap function it
intended to call was ill-formed when instantiated. GCC diagnosed
this ill-formedness w/o needing an instantiation.
(C) size_delete11.pass.cpp was fixed by adding c++2a to the list of
unsupported dialects.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@322810 91177308-0d34-0410-b5e6-96231b3b80d8
Previously .fail.cpp tests for nodiscard were run with -Wunused-result
being a warning, not an error, when the compiler didn't support -verify.
When -verify isn't enabled this change judiciously adds -Werror=unused-result
when to only the failure tests containing the // expected-error string for nodiscard.
As a drive-by change, this patch also adds a missing // UNSUPPORTED: c++2a to
a test which was only supposed to run in C++ <= 11.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@322776 91177308-0d34-0410-b5e6-96231b3b80d8
Inline the provided "fallback" definitions (which seem to always be
taken) that expand to __cdecl into users. The fallback definitions
for the *CRTIMP* macros were wrong in the case where the CRT is being
linked statically, so define our own macro as a replacement.
Differential Revision: https://reviews.llvm.org/D42158
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@322617 91177308-0d34-0410-b5e6-96231b3b80d8
type 'result_type' to 'double'. The only thing that we ever do with
these numbers is to promote them to 'double' and use them in a division.
For small result_types, the values were getting truncated, skewing the
results. Thanks to James Nagurne for the suggestion.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@322556 91177308-0d34-0410-b5e6-96231b3b80d8
