This patch fixes an ASAN-found issue in std::basic_filebuf where we'd
check the wrong size before proceeding to set our internal buffer to
the externally-provided buffer, leading to the library trying to read
from the incorrect buffer in underflow().
Thanks to Andrey Semin for the patch.
Differential Revision: https://reviews.llvm.org/D154514
We did not mark std::begin/std::end as noexcept for C-style arrays, we
did not have conditional noexcept on cbegin/cend, and we did not mark
array cbegin/cend as constexpr in all Standard modes. Since this is a
LWG issue, we should implement it as a DR in all Standard modes as
usual.
This patch fixes these issues and adds test coverage. Fixes#67471.
Resolves Issue #69083
The `_CopySegment` helper for `ranges::copy(join_view, out)` is not
`constexpr` causing rejection in `libc++` in a static assertion context
as in the issue snippet.
If we know that index is larger than SSO size, we know that we can't be
in SSO case, and should access the pointer. This removes extra check
from operator[] for inputs known at compile time to be larger than SSO.
This is an implementation detail for `move_only_function` (and potentially other type-erasing classes).
Reviewed By: #libc, ldionne
Spies: Mordante, ldionne, EricWF, libcxx-commits
Differential Revision: https://reviews.llvm.org/D140259
As explained in `__config`, we have an ABI tag that we use to ensure
that we don't run into ODR issues when mixing different versions of
libc++ in multiple TUs. However, the reasoning behind that extends not
only to different versions of libc++, but also to different
configurations of the same version of libc++. In fact, we've been aware
of this for a while but never really bothered to make the change because
ODR issues are often thought to be benign.
Well, it turns out that I just spent over an hour banging my head
against an issue that boils down to our lack of encoding of some ODR
properties in the ABI tag, so here's the patch we should have done a
long time ago.
For now, the ODR properties we encode in the ABI tag are:
- library version
- exceptions vs no-exceptions
- hardening mode
Those are all things that we support different values for on a per-TU
basis and they definitely affect ODR in a meaningful way. We can add
more properties later as we see fit.
There is currently a major problem with the CI output:
The information you need to see never appears in the visible log.
This is because our logs are very verbose, and list (A) every test as
they run it, and (B) every unsupported tests. This can be thousands of
lines.
(A) was introduced by me when I disabled the PTY progress bar, which
doesn't play nice with log files. That change was an improvement, but I
have now disabled the PTY on the builders, so we can go back to passing
`-s`. When `-s` is passed but no PTY is available, it prints a long
friendly progress indicator.
(B) is solved here by disabling the printing of unsupported tests at the
end of the test suite. While it can be useful on occasion to audit the
list of unsupported tests, it's far from a common operation. Instead
people want to see the log of their failure. We should upload the xml
results, so if auditing is needed, it can be done using that.
Hopefully this change will make it so that the test failures appear in
the actual log output
This list is a burden to maintain and provides very limited value. A
user wishing to know whether a particular function is declared with a
`nodiscard` extension would be better off grepping the headers.
Without this patch, the functions `executeScriptInternal` and thus
`runOnce` in `llvm/utils/lit/lit/TestRunner.py` return either a tuple like
`(out, err, exitCode, timeoutInfo)` or a `lit.Test.Result` object. They
return the latter only when there's a lit internal shell parse error in
a RUN line. In my opinion, a more straight-forward way to handle
exceptional cases like that is to use python exceptions.
For that purpose, this patch introduces `ScriptFatal`. Thus, this patch
changes `executeScriptInternal` to always either return the tuple or
raise the `ScriptFatal` exception. It updates `runOnce` and
`libcxx/utils/libcxx/test/format.py` to catch the exception rather than
check for the special return type.
This patch also changes `runOnce` to convert the exception to a
`Test.UNRESOLVED` result instead of `TEST.FAIL`. The former is the
proper result for such a malformed test, for which a rerun (given an
`ALLOW_RETRIES:`) serves no purpose. There are at least two benefits
from this change. First, `_runShTest` no longer has to specially and
cryptically handle this case to avoid unnecessary reruns. Second, an
`XFAIL:` directive no longer hides such a failure [as we saw
previously](https://reviews.llvm.org/D154987#4501125).
To facilitate the `_runShTest` change, this patch inserts the internal
shell parse error diagnostic into the format of the test's normal debug
output rather than suppressing the latter entirely. That change is also
important for [D154987](https://reviews.llvm.org/D154987), which
proposes to reuse `ScriptFatal` for python compile errors in PYTHON
lines or in `config.prologue`. In that case, the diagnostic might follow
debugging output from the test's previous RUN or PYTHON lines, so
suppressing the normal debug output would lose information.
This implements layout_stride for C++23 and with that completes the
implementation of the C++23 mdspan header. The feature test macro is
added, and the status pages updated.
Co-authored-by: Damien L-G <dalg24@gmail.com>
Differential Revision: https://reviews.llvm.org/D157171
If an iterator passed to std::uninitialized_copy & friends provided an
unconstrained comparison operator, we would trigger an ambiguous
overload resolution because we used to compare against
__unreachable_sentinel in our implementation.
This patch fixes that by only comparing the output iterator when it is
actually required, i.e. in the <ranges> versions of the algorithms.
Fixes#69334
Mark tests as necessary to accommodate Android L (5.0 / API 21) and up.
Add three Android lit features:
- android
- android-device-api=(21,22,23,...)
- LIBCXX-ANDROID-FIXME (for failures that need follow-up work)
Enable an AIX workaround in filesystem_test_helper.h for the broken
chmod on older Android devices.
Mark failing test with XFAIL or UNSUPPORTED:
- Mark modules tests as UNSUPPORTED, matching other configurations.
- Mark a gdb test as UNSUPPORTED.
- XFAIL tests for old devices that lack an API (fmemopen).
- XFAIL various FS tests (because SELinux blocks FIFO and hard linking,
because fchmodat is broken on old devices).
- XFAIL various locale tests (because Bionic has limited locale
support). (Also XFAIL an re.traits test.)
- XFAIL some print.fun tests because the error exception has no system
error string.
- Mark std::{cin,wcin} tests UNSUPPORTED because they hang with
adb_run.py on old devices.
- Mark a few tests UNSUPPORTED because they allocate too much memory.
- notify_one.pass.cpp is flaky on Android.
- XFAIL libc++abi demangler test because of Android's special long
double on x86[-64].
N.B. The `__ANDROID_API__` macro specifies a minimum required API level
at build-time, whereas the android-device-api lit feature is the
detected API level of the device at run-time. The android-device-api
value will be >= `__ANDROID_API__`.
This commit was split out from https://reviews.llvm.org/D139147.
Fixes: https://github.com/llvm/llvm-project/issues/69270
Add a Dockerfile for a new Docker image, libcxx-builder-android, that
extends libcxx-builder with support for testing Android.
The image includes these things:
* An Android Clang compiler and sysroot.
* The Android platform-tools (e.g. adb), so that an Android buildbot can
run programs on an Android device. At container startup, copy these
platform tools to an "android-platform-tools" Docker volume to share
them with an emulator container. This copying ensures that the emulator
and libcxx-builder containers avoid mismatched adb versions.
* Docker, so that an Android buildbot can manage a sibling Docker
container that runs the Android emulator.
Add an Android-specific run-buildbot-container script for local
development. Currently using this script requires building
libcxx-build-android and an emulator image locally.
Fixes: https://github.com/llvm/llvm-project/issues/69270
Differential Revision: https://reviews.llvm.org/D155271
I could probably break this commit into more pieces.
---
This patch adds libc++ support for Android L (Android 5.0+) and up,
tested using the Android team's current compiler, a recent version of
the AOSP sysroot, and the x86[-64] Android Emulator.
CMake and Lit Configuration:
Add runtimes/cmake/android/Arch-${ARCH}.cmake files that configure CMake
to cross-compile to Android without using CMake's built-in NDK support
(which only works with an actual packaged NDK).
Add libcxx/cmake/caches/AndroidNDK.cmake that builds and tests libc++
(and libc++abi) for Android. This file configures libc++ to match what
the NDK distributes, e.g.:
- libc++_shared.so (includes libc++abi objects, there is no
libc++abi.so). libunwind is linked statically but not exported.
- libc++_static.a (does not include libc++abi) and libc++abi.a
- `std::__ndk1` namespace
- All the libraries are built with `__ANDROID_API__=21`, even when they
are linked to something targeting a higher API level.
(However, when the Android LLVM team builds these components, they do
not use these CMake cache files. Instead they use Python scripts to
configure the builds. See
https://android.googlesource.com/toolchain/llvm_android/.)
Add llvm-libc++[abi].android-ndk.cfg.in files that test the Android
NDK's libc++_shared.so. These files can target old or new Android
devices. The Android LLVM team uses these test files to test libc++ for
both arm/arm64 and x86/x86_64 architectures.
The Android testing mode works by setting %{executor} to adb_run.py,
which uses `adb push` and `adb shell` to run tests remotely. adb_run.py
always runs tests as the "shell" user even on an old emulator where "adb
unroot" doesn't work. The script has workarounds for old Android
devices. The script uses a Unix domain socket on the host
(--job-limit-socket) to restrict concurrent adb invocations. Compiling
the tests is a major part of libc++ testing run-time, so it's desirable
to exploit all the host cores without overburdening the test devices,
which can have far fewer cores.
BuildKite CI:
Add a builder to run-buildbot, `android-ndk-*`, that uses Android Clang
and an Android sysroot to build libc++, then starts an Android emulator
container to run tests.
Run the emulator and an adb server in a separate Docker container
(libcxx-ci-android-emulator), and create a separate Docker image for
each emulator OS system image. Set ADB_SERVER_SOCKET to connect to the
container's adb server. Running the only adb server inside the container
makes cleanup more reliable between test runs, e.g. the adb client
doesn't create a `~/.android` directory and the adb server can be
restarted along with the emulator using docker stop/run. (N.B. The
emulator insists on connecting to an adb server and will start one
itself if it can't connect to one.)
The suffix to the android-ndk-* job is a label that concisely specifies
an Android SDK emulator image. e.g.:
- "system-images;android-21;default;x86" ==> 21-def-x86
- "system-images;android-33;google_apis;x86_64" ==> 33-goog-x86_64
Fixes: https://github.com/llvm/llvm-project/issues/69270
Differential Revision: https://reviews.llvm.org/D139147
Using `true` as a no-op unfortunately does not work on windows, which
fails libcxx lit tests on windows. Lit provides the `:` internal shell
builtin which is equivalent to `true`.
std::atomic is implemented with the following (confusing!) hierarchy of
types:
std::atomic<T> : std::__atomic_base<T> { ... };
std::__atomic_base<T> {
std::__cxx_atomic_impl<T> __impl;
};
std::__cxx_atomic_impl<T> {
_Atomic(T) __val;
};
Inside std::__atomic_base, we implement the is_lock_free() and
is_always_lock_free() functions. However, we used to implement them
inconsistently:
- is_always_lock_free() is based on whether __cxx_atomic_impl<T> is
always lock free (using the builtin), which means that we include any
potential padding added by _Atomic(T) into the determination.
- is_lock_free() was based on whether T is lock free (using the
builtin), which meant that we did not take into account any potential
padding added by _Atomic(T).
It is important to note that the padding added by _Atomic(T) can turn a
type that wouldn't be lock free into a lock free type, for example by
making its size become a power of two.
The inconsistency of how the two functions were implemented could lead
to cases where is_always_lock_free() would return true, but
is_lock_free() would then return false. This is the case for example of
the following type, which is always lock free on arm64 but was
incorrectly reported as !is_lock_free() before this patch:
struct Foo { float x[3]; };
This patch switches the determination of is_lock_free() to be based on
__cxx_atomic_impl<T> instead to match how we determine
is_always_lock_free().
rdar://115324353
This allows running these quick checks faster than in our Buildkite
pipeline, which has much more latency. This will also avoid blocking the
rest of the testing pipeline in case the generated-files checks are
failing.
This will make it easier to implement new(nothrow) without calling the
throwing version of new when exceptions are disabled. See
https://llvm.org/D150610 for the full discussion.
Specifically, the test std/input.output/string.streams/stringstream/stringstream.members/gcount.pass.cpp
allocates a std::string with INT_MAX-1 elements, and then writes this to
a std::stringstream. On Linux, running this test consumes around 5.0 GB
of memory; on Windows, it ends up using up to 6.8 GB of memory.
This limits whether such tests can run on e.g. GitHub Actions runners,
where the free runners are limited to 8 GB of memory.
This is somewhat similar to, but still notably different, from the
existing test parameter long_tests.
This PR addresses a smaller detail discussed in the code review for
https://github.com/llvm/llvm-project/pull/66968. Currently, some
functions in the `libc++` PSTL CPU backend have been appended with a
digit to indicate the number of input iterator arguments. However, there
is no need to change the name for each version as overloading can be
used instead. This PR will make the naming more consistent in the the
CPU and the proposed OpenMP backend.
We support GCC 13, which supports the flag. This allows simplifying the
CMake logic around the use of -nostdlib++. Note that there are other
places where we don't assume -nostdlib++ yet in the build, but this
patch is intentionally trying to be small because this part of our CMake
is pretty tricky.
This patch introduces the support::make_test_jthread utility which is
basically the same as support::make_test_thread but for std::jthread. It
allows vendors to maintain a downstream way to create threads for use
within the test suite, which is especially useful for embedded
platforms.
This allows smaller allocations to occur, closer to the actual
std::string's required size. This is particularly effective in
decreasing the allocation size upon initial construction (where
__recommend is called to determine the size).
Although the memory savings per-string are never more than 8 bytes per
string initially, this quickly adds up. And has lead to not insigficant
memory savings at Google.
Unfortunately, this change is ABI breaking because it changes the value
returned by max_size. So it has to be guarded.
This patch removes undefined behavior in list and forward_list and __hash_table
caused by improperly beginning and ending the lifetime of the various node
classes. It allows removing the _LIBCPP_STANDALONE_DEBUG macro from
these node types since we now properly begin and end their lifetime,
meaning that we won't trip up constructor homing.
See https://reviews.llvm.org/D98750 for more information on what prompted
this patch.
This commit re-applies 0687e4d9f3, which had been reverted in b935882bdc
because it broke the LLDB build. LLDB folks tell me I can go ahead and
re-commit this now.
Differential Revision: https://reviews.llvm.org/D101206
Co-authored-by: Amy Kwan <amy.kwan1@ibm.com>
This updates the clang-format we use in libc++ to 17. This is necessary
to start running the generated-files checks in GitHub Actions (in
#68920). In fact this is a pre-existing issue regardless of #68920 --
right now our ignore_format.txt job disagrees with the LLVM-wide
clang-format job.
extents CTAD was requiring default constructibility of the extent
arguments due to the way we implemented a pack expansion. This
requirement is not in the standard.
Reported in issue #68671https://github.com/llvm/llvm-project/issues/68671 by @hewillk.
Fixes#68671
The step that checked for ignore_format.txt being consistent with the
tree wouldn't print any explicit diagnostic when failing, which led to
confusion. After this patch, an explicit diagnostic will be printed by
the job along with the required diff to ignore_format.txt.
The previous example wasn't a configuration that we actually test, but
was a simplistic configuration with libstdc++ as ABI library.
The previous example configuration has bitrotted and broken in a couple
different ways since it was added:
- In b0fd9497af, libcxx added uses of weak
symbols that can be overridden by the user. GNU ld fails to export such
weak symbols on MinGW, resulting in errors like:
C:/msys64/mingw64/bin/ld: cannot export _ZNSt3__126__libcpp_assertion_handlerEPKciS1_S1_: symbol wrong type (2 vs 3)
By switching to using LLD, -DLLVM_ENABLE_LLD=ON, alternatively
-DCMAKE_SHARED_LINKER_FLAGS=-fuse-ld=lld, linking still succeeds.
- In a80e65e00a, building with
-DLIBCXX_CXX_ABI=libstdc++ in this configuration started failing. Switch
to showing how to build with libcxxabi instead, which is the tested
configuration and also is the typical configuration in how it most
usually would be used on Windows.
These updates allow removing the mentions of the old caveats with the
build configuration.
Previously, libcxx forced all strings created during constant evaluation
to point to allocated memory. That was done due to implementation
difficultites, but it turns out not to be necessary. This patch permits
the use of SSO strings during constant evaluation, and also simplifies
the implementation.
This does have a downside in terms of enabling users to accidentally
write non-portable code, however, which I've documented in
UsingLibcxx.rst.
In particular, whether `constinit std::string x = "...";` will
successfully compile now depends on whether the string is smaller than
the SSO capacity -- in libc++, up to 22 bytes on 64-bit platforms, and
up to 10 bytes on 32-bit platforms. By comparison, libstdc++ and MSVC
have an SSO capacity of 15 bytes, except that in libstdc++,
constant-initialized strings cannot be used as function-locals because
the object contains a pointer to itself.
Closes#68434
To allow for a smoother transition, keep the safe mode working as is in
the LLVM 18 release (the first release that aims to make hardening
available), then deprecate it in LLVM 19.
Fixes#64619
Clang warns diagnostic for non-standard layout types in `offsetof` only
if they are in evaluated context. With this patch, you'll also get
diagnostic if you use `offsetof` on non-standard layout types in any
other contexts
This reapplies the change from
c218c80c73, which was lost in the
refactoring in 78d649a417.
On Windows, the Python interpreter is by default installed in a path
like "C:\Program Files\Python38\python.exe", which requires quoting when
included in a concatenated command string (as opposed to a command list,
where each argument is a separate element).
This doesn't show up in the CI environments as they use Python installed
in a different directory, like C:\Python.
In addition to the python executable, also quote the path to run.py, if
needed.
This makes exception handling a lot simpler, since we don't have to convert any exceptions this way. Is also properly handles all the user-thrown exceptions.
Reviewed By: ldionne, #libc
Spies: arichardson, mstorsjo, libcxx-commits
Differential Revision: https://reviews.llvm.org/D154238
Adding additional instantiations to the dylib isn't actually an ABI break as long as programs targeting an older dylib don't start to depend on them. Making additional instantiations a matter of availability allows us to add them without an ABI break.
Reviewed By: #libc, ldionne, Mordante
Spies: arichardson, ldionne, Mordante, libcxx-commits
Differential Revision: https://reviews.llvm.org/D154796
Standard says that implementation of math functions that have
floating-point-type parameter should provide an "overload for each
cv-unqualified floating-point type".
The diagnostics for `enable_if_t` are extremely opaque:
```
error: no matching function for call to 'bind_front'
note: candidate template ignored: requirement 'integral_constant<bool, false>::value' was not satisfied
```
Using requires-expressions gives us a little more context:
```
error: no matching function for call to 'bind_front'
note: candidate template ignored: constraints not satisfied
note: because 'is_constructible_v<decay_t<T &>, T &>' evaluated to false
```
Pull request: #68249
This patch removes undefined behavior in list and forward_list and __hash_table
caused by improperly beginning and ending the lifetime of the various node
classes. It allows removing the _LIBCPP_STANDALONE_DEBUG macro from
these node types since we now properly begin and end their lifetime,
meaning that we won't trip up constructor homing.
See https://reviews.llvm.org/D98750 for more information on what prompted
this patch.
Differential Revision: https://reviews.llvm.org/D101206
Co-authored-by: Amy Kwan <amy.kwan1@ibm.com>
Fixes#68051.
Current implementation passes the number of `_AlignedStorage` objects
when it calls to `allocate` and the number of **bytes** on `deallocate`.
This only applies to allocations that allocate control block and the
storage together, i.e. `make_shared` and `allocate_shared`.
Found by ASan under Clang combined with `-fsized-deallocation`.
This patch removes the non compliant constructor of std::future_error
and adds the standards compliant constructor in C++17 instead.
Note that we can't support the constructor as an extension in all
standard modes because it uses delegating constructors, which require
C++11. We could in theory support the constructor as an extension in
C++11 and C++14 only, however I believe it is acceptable not to do that
since I expect the breakage from this patch will be minimal.
If it turns out that more code than we expect is broken by this, we can
reconsider that decision.
This was found during D99515.
Differential Revision: https://reviews.llvm.org/D99567
Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
We were incorrectly deducing the return type of size() because we were
not using ternary operators in the implementation (as the spec says).
Instead of deducing the common type of the expressions in the spec, we
would deduce potentially different return types and fail to compile.
Fixes#67551
The _LIBCPP_PSTL_CUSTOMIZATION_POINT macro was assuming that the policy
was called _RawPolicy and referencing it by name. It happened to always
work but this was definitely accidental and an oversight in the original
implementation. This patch fixes that by passing the policy to the macro
explicitly. Noticed while reviewing #66968.
Since 78d649a417 the recommended way to
pass an executor is to use the _TEST_PARAMS variable, which means we now
pass more complicated value (including ones that may contain multiple
`=`) as part of this variable. However, the `REGEX REPLACE` being used
has greedy matches so everything up to the last = becomes part of the
variable name which results in invalid syntax in the generated lit
config file.
This was noticed due to builder failures for those using the
CrossWinToARMLinux.cmake cache file.
---------
Co-authored-by: Vladimir Vereschaka <vvereschaka@accesssoftek.com>
We removed all traces of the legacy debug mode a while back, but we
forgot to remove the actual `.cpp` file that implemented the legacy
debug handler. The file is not referenced from anywhere so this is
effectively a NFC.
This PR amends the `XFAIL` directive to clang < 18 in
no_unique_address.compile.pass.cpp tests.
Since PR 67199 implements the [[msvc::no_unique_address]] attribute,
these tests pass in a windows-x64 build against ToT clang-cl:
```
********************
Unexpectedly Passed Tests (3):
llvm-libc++-static-clangcl.cfg.in :: libcxx/ranges/range.adaptors/range.lazy.split/no_unique_address.compile.pass.cpp
llvm-libc++-static-clangcl.cfg.in :: libcxx/ranges/range.adaptors/range.split/no_unique_address.compile.pass.cpp
llvm-libc++-static-clangcl.cfg.in :: libcxx/ranges/range.factories/range.istream.view/no_unique_address.compile.pass.cpp
```
---------
Co-authored-by: Caslyn Tonelli <caslyn@google.com>
Credits: this change is based on analysis and a proof of concept by
gerbens@google.com.
Before, the compiler loses track of end as 'this' and other references
possibly escape beyond the compiler's scope. This can be see in the
generated assembly:
16.28 │200c80: mov %r15d,(%rax)
60.87 │200c83: add $0x4,%rax
│200c87: mov %rax,-0x38(%rbp)
0.03 │200c8b: → jmpq 200d4e
...
...
1.69 │200d4e: cmp %r15d,%r12d
│200d51: → je 200c40
16.34 │200d57: inc %r15d
0.05 │200d5a: mov -0x38(%rbp),%rax
3.27 │200d5e: mov -0x30(%rbp),%r13
1.47 │200d62: cmp %r13,%rax
│200d65: → jne 200c80
We fix this by always explicitly storing the loaded local and pointer
back at the end of push back. This generates some slight source 'noise',
but creates nice and compact fast path code, i.e.:
32.64 │200760: mov %r14d,(%r12)
9.97 │200764: add $0x4,%r12
6.97 │200768: mov %r12,-0x38(%rbp)
32.17 │20076c: add $0x1,%r14d
2.36 │200770: cmp %r14d,%ebx
│200773: → je 200730
8.98 │200775: mov -0x30(%rbp),%r13
6.75 │200779: cmp %r13,%r12
│20077c: → jne 200760
Now there is a single store for the push_back value (as before), and a
single store for the end without a reload (dependency).
For fully local vectors, (i.e., not referenced elsewhere), the capacity
load and store inside the loop could also be removed, but this requires
more substantial refactoring inside vector.
Differential Revision: https://reviews.llvm.org/D80588
The tests were a bit of a mess -- the testing coverage wasn't bad but it
was extremely difficult to see what was being tested and where. I split
up the tests to make them easier to audit for completeness and did such
an audit, adding a few missing tests (e.g. the conditional noexcept-ness
of std::cbegin and std::cend). I also audited the synopsis and adjusted
it where it needed to be adjusted.
This patch is in preparation of fixing #67471.
Even though the underlying issue was fixed in #65177 when we made
std::pointer_traits SFINAE-friendly, it is worth adding a regression
test since it's so easy to do.
This fixes a size regression in Fuchsia when building a static libc++
multilib with exceptions disabled. Referring to `system_category` in
`__throw_system_error` brings in a relatively large amount of additional
exception classes into the link without substantially improving the
error message.
While doing this, I also found a few tests that were either clearly
incorrect (e.g. testing the wrong function) or that lacked basic test
coverage like testing std::string itself (e.g. the test was only checking
std::basic_string with a custom allocator). In these cases, I did a few
conservative drive-by changes.
Differential Revision: https://reviews.llvm.org/D140550
Co-authored-by: Brendan Emery <brendan.emery@esrlabs.com>
This is a more correct way of linking against a specific runtime
library with the GCC-like clang frontend.
This avoids having to pass -D_DEBUG (and passes flags like -D_DLL,
which we should be passing if linking against the dynamic CRT).
When -fms-runtime-lib= is specified, each compiled object file gets
embedded directives instructing the linker about what CRT it should
link against. The -nostdlib we pass to the compiler driver only
inhibits the libs that the compiler driver passes to the linker
command.
Thus, this also avoids having to specify -lmsvcrt, -lmsvcrtd,
-llibcmt or -llibcmtd, and -loldnames.
Based on a patch by Andrew Ng.
The -fms-runtime-lib= option was added in Clang 16, but libcxx
now has dropped support for Clang 15.
Differential Revision: https://reviews.llvm.org/D155562
This patch makes clang diagnose extensive cases of consteval if and is_constant_evaluated usage that are tautologically true or false.
This introduces a new IsRuntimeEvaluated boolean flag to Sema::ExpressionEvaluationContextRecord that means the immediate appearance of if consteval or is_constant_evaluated are tautologically false(e.g. inside if !consteval {} block or non-constexpr-qualified function definition body)
This patch also pushes new expression evaluation context when parsing the condition of if constexpr and initializer of constexpr variables so that Sema can be aware that the use of consteval if and is_consteval are tautologically true in if constexpr condition and constexpr variable initializers.
BEFORE this patch, the warning for is_constant_evaluated was emitted from constant evaluator. This patch moves the warning logic to Sema in order to diagnose tautological use of is_constant_evaluated in the same way as consteval if.
This patch separates initializer evaluation context from InitializerScopeRAII.
This fixes a bug that was happening when user takes address of function address in initializers of non-local variables.
Fixes https://github.com/llvm/llvm-project/issues/43760
Fixes https://github.com/llvm/llvm-project/issues/51567
Reviewed By: cor3ntin, ldionne
Differential Revision: https://reviews.llvm.org/D155064
This is useful when trying to run multiple tests with different
arguments to the executor script. This is needed in my case since I
do not know the correct ssh connection arguments when building libc++.
The testing script I have spawns multiple QEMU instances that listen on
a given port on localhost and runs lit with the --num-shards/--run-shard
argument. In order to connect each shard to the right QEMU instances I
need to customize the arguments passed to ssh.py (--extra-ssh-args and
--extra-scp-args) but can't do this at configure time since the target
port is only known when running the tests but not when calling CMake.
This change allows me to pass `executor=ssh.py <args>` to lit once I
know
the right hostname/port for running the tests.
This also deprecates the `LIB{CXX,CXXABI,UNWIND}_EXECUTOR` CMake
variable
as the same can be achieved by adding `executor=...` to the
`LIB{CXX,CXXABI,UNWIND}_TEST_PARAMS` variable.
This reduces the number of instantiations and also avoid blowing up
past the fold-expression limit of Clang.
This is NOT a general statement that we should strive to stay within
Clang's (sometimes way too small) limits, however in this case the
change will reduce the number of template instantiations while at the
same time doing that, which is good.
Differential Revision: https://reviews.llvm.org/D132509
Since LLVM 17 has been branched and is on the verge of being released,
we can drop the CI job that tests against Clang 15. I think the number
of cherry-picks to `release/17.x` will be a lot smaller now, so keeping
a Clang 15 job around for that purpose seems unnecessary.
As a fly-by, this patch also removes some Clang 15 workarounds and test
suite annotations as we usually do. It also removes some slightly older
gcc test suite annotations that were missed.
Now that we have a LLVM-wide GitHub action, it doesn't make sense to
have a custom job for libc++ in our BuildKite CI pipeline. This only creates
confusion as to which CI job should be acted upon.
The new formatting job enabled with GitHub actions checks for mis-formatted
files everywhere, not only in the select subdirectories we checked previously.
As a result, we'd get spurious clang-format failure reports whenever a
test was modified. This patch fixes that.
After landing the implementation of LWG3545, I realized that the tests
for std::pointer_traits had become a bit disorganized. This patch is a
NFC that refactors the tests:
- Move compile-only tests to `.compile.pass.cpp` tests
- Re-create the clear distinction between tests for the std::pointer_traits base tempate and for the T* specialization.
- De-duplicate test coverage -- we had a bunch of things that were tested in duplication.
AppleClang 15 was released on September 18th and is now stable. Per our
policy, we're bumping the supported AppleClang compiler to the latest
release. This allows cleaning up the test suite, but most importantly
unblocking various other patches that are blocked on bumping the
compiler requirements.
Previously, assignment to a std::basic_string type with a _custom_
allocator could under certain conditions attempt to interpret part of
the target string's "short" string-content as if it was a "long" data
pointer, and attempt to deallocate a garbage value.
This is a serious bug, but code in which it might happen is rare. It
required:
1. the basic_string must be using a custom allocator type which sets the
propagate_on_container_copy_assignment trait to true (thus, it does not
affect the default allocator, nor most custom allocators).
2. the allocator for the target string must compare not equal to the
allocator for the source string (many allocators always compare equal).
3. the source of the copy must currently contain a "long" string, and
the assignment-target must currently contain a "short" string.
Finally, the issue would've typically been innocuous when the bytes
misinterpreted as a pointer were all zero, as deallocating a nullptr is
typically a no-op. This is why existing test cases did not exhibit an
issue: they were all zero-length strings, which do not have data in the
bytes interpreted as a pointer.
Since GCC now supports -nostdlib++, we can remove some complexity in the
test configurations and do the same as Clang. However, we can't fully
remove the GCC test configuration for libc++ because we apparently need
to explicitly link against libm for some tests to work.
Temporary workaround for the following CMake error:
```
CMake Error in /llvm/libcxx/benchmarks/CMakeLists.txt:
The compiler feature "cxx_std_23" is not known to CXX compiler
"IBMClang"
```
This patch is a melting pot of changes picked up from
https://llvm.org/D61878. It adds a few tests checking corner cases of
unordered containers comparison and adds benchmarks for a few
unordered_set operations.
On Apple platforms, we always support the -nostdlib++ flag. Hence, it is
not necessary to manually link against system libraries. In fact, doing
so causes us to link against libSystem explicitly, which messes up with
the order of libraries we should use. Indeed:
Before patch, using the system unwinder (LIBCXXABI_USE_LLVM_UNWINDER = OFF)
===========================================================================
$ otool -L lib/{libc++.1.dylib,libc++abi.1.dylib,libunwind.1.dylib}
lib/libc++.1.dylib:
@rpath/libc++.1.dylib
/usr/lib/libSystem.B.dylib
@rpath/libc++abi.1.dylib
lib/libc++abi.1.dylib:
@rpath/libc++abi.1.dylib
/usr/lib/libSystem.B.dylib
lib/libunwind.1.dylib:
@rpath/libunwind.1.dylib
/usr/lib/libSystem.B.dylib
After patch, using the system unwinder (LIBCXXABI_USE_LLVM_UNWINDER = OFF)
===========================================================================
$ otool -L lib/{libc++.1.dylib,libc++abi.1.dylib,libunwind.1.dylib}
lib/libc++.1.dylib:
@rpath/libc++.1.dylib
@rpath/libc++abi.1.dylib
/usr/lib/libSystem.B.dylib
lib/libc++abi.1.dylib:
@rpath/libc++abi.1.dylib
/usr/lib/libSystem.B.dylib
lib/libunwind.1.dylib:
@rpath/libunwind.1.dylib
/usr/lib/libSystem.B.dylib
Before patch, with the LLVM unwinder (LIBCXXABI_USE_LLVM_UNWINDER = ON)
=======================================================================
$ otool -L lib/{libc++.1.dylib,libc++abi.1.dylib,libunwind.1.dylib}
lib/libc++.1.dylib:
@rpath/libc++.1.dylib
/usr/lib/libSystem.B.dylib
@rpath/libc++abi.1.dylib
@rpath/libunwind.1.dylib
lib/libc++abi.1.dylib:
@rpath/libc++abi.1.dylib
/usr/lib/libSystem.B.dylib
@rpath/libunwind.1.dylib
lib/libunwind.1.dylib:
@rpath/libunwind.1.dylib
/usr/lib/libSystem.B.dylib
After patch, with the LLVM unwinder (LIBCXXABI_USE_LLVM_UNWINDER = ON)
======================================================================
$ otool -L lib/{libc++.1.dylib,libc++abi.1.dylib,libunwind.1.dylib}
lib/libc++.1.dylib:
@rpath/libc++.1.dylib
@rpath/libc++abi.1.dylib
@rpath/libunwind.1.dylib
/usr/lib/libSystem.B.dylib
lib/libc++abi.1.dylib:
@rpath/libc++abi.1.dylib
@rpath/libunwind.1.dylib
/usr/lib/libSystem.B.dylib
lib/libunwind.1.dylib:
@rpath/libunwind.1.dylib
/usr/lib/libSystem.B.dylib
As we can see, libSystem appears before the just-built libraries before
the patch, which causes the libunwind.dylib bundled in libSystem.dylib
to be used instead of the just-built libunwind.dylib.
We didn't notice the issue until recently when I tried to update the
macOS CI builders to macOS 13.5, where it is necessary to use the right
libunwind library (the exact reason still needs to be investigated).
This makes it obvious that libc++ is used in an unsupported configuration,
and the compiler probably has to be updated. It often happens that people
try to use libc++ and don't realize that their compiler is too old.
Differential Revision: https://reviews.llvm.org/D158214
Since we are defining these typedefs inside namespace std, we need to
refer to ::once_flag (the C Standard Library version). Otherwise
'once_flag' refers to 'std::once_flag', and that's not something we can
pass to the C Standard Library '::call_once()' function later on.
__call_once is large and cluttered with #ifdef preprocessor guards. This
cleans it up a bit by using an exception guard instead of try-catch.
Differential Revision: https://reviews.llvm.org/D112319
Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
This patch and D154984 were discussed in
<https://discourse.llvm.org/t/rfc-improving-lits-debug-output/72839>.
Motivation
----------
D154984 removes the "Script:" section that lit prints along with a
test's output, and it makes -v and -a imply -vv. For example, after
D154984, the "Script:" section below is never shown, but -v is enough
to produce the execution trace following it:
```
Script:
--
: 'RUN: at line 1'; echo hello | FileCheck bogus.txt && echo success
--
Exit Code: 2
Command Output (stdout):
--
$ ":" "RUN: at line 1"
$ "echo" "hello"
# command output:
hello
$ "FileCheck" "bogus.txt"
# command stderr:
Could not open check file 'bogus.txt': No such file or directory
error: command failed with exit status: 2
--
```
In the D154984 review, some reviewers point out that they have been
using the "Script:" section for copying and pasting a test's shell
commands to a terminal window. The shell commands as printed in the
execution trace can be harder to copy and paste for the following
reasons:
- They drop redirections and break apart RUN lines at `&&`, `|`, etc.
- They add `$` at the start of every command, which makes it hard to
copy and paste multiple commands in bulk.
- Command stdout, stderr, etc. are interleaved with the commands and
are not clearly delineated.
- They don't always use proper shell quoting. Instead, they blindly
enclose all command-line arguments in double quotes.
Changes
-------
D154984 plus this patch converts the above example into:
```
Exit Code: 2
Command Output (stdout):
--
# RUN: at line 1
echo hello | FileCheck bogus-file.txt && echo success
# executed command: echo hello
# .---command stdout------------
# | hello
# `-----------------------------
# executed command: FileCheck bogus-file.txt
# .---command stderr------------
# | Could not open check file 'bogus-file.txt': No such file or directory
# `-----------------------------
# error: command failed with exit status: 2
--
```
Thus, this patch addresses the above issues as follows:
- The entire execution trace can be copied and pasted in bulk to a
terminal for correct execution of the RUN lines, which are printed
intact as they appeared in the original RUN lines except lit
substitutions are expanded. Everything else in the execution trace
appears in shell comments so it has no effect in a terminal.
- Each of the RUN line's commands is repeated (in shell comments) as
it executes to show (1) that the command actually executed (e.g.,
`echo success` above didn't) and (2) what stdout, stderr, non-zero
exit status, and output files are associated with the command, if
any. Shell quoting in the command is now correct and minimal but is
not necessarily the original shell quoting from the RUN line.
- The start and end of the contents of stdout, stderr, or an output
file is now delineated clearly in the trace.
To help produce some of the above output, this patch extends lit's
internal shell with a built-in `@echo` command. It's like `echo`
except lit suppresses the normal execution trace for `@echo` and just
prints its stdout directly. For now, `@echo` isn't documented for use
in lit tests.
Without this patch, libcxx's custom lit test format tries to parse the
stdout from `lit.TestRunner.executeScriptInternal` (which runs lit's
internal shell) to extract the stdout and stderr produced by shell
commands, and that parse no longer works after the above changes.
This patch makes a small adjustment to
`lit.TestRunner.executeScriptInternal` so libcxx can just request
stdout and stderr without an execution trace.
(As a minor drive-by fix that came up in testing: lit's internal `not`
command now always produces a numeric exit status and never `True`.)
Caveat
------
This patch only makes the above changes for lit's internal shell. In
most cases, we do not know how to force external shells (e.g., bash,
sh, window's `cmd`) to produce execution traces in the manner we want.
To configure a test suite to use lit's internal shell (which is
usually better for test portability than external shells anyway), add
this to the test suite's `lit.cfg` or other configuration file:
```
config.test_format = lit.formats.ShTest(execute_external=False)
```
Reviewed By: MaskRay, awarzynski
Differential Revision: https://reviews.llvm.org/D156954
The helper function `__pair_like_explicit_wknd` is only SFINAE-ed with
`tuple_size<remove_cvref_t<_PairLike>>::value == 2`, but its function
body assumes `std::get` being valid.
Fixes#65620
When running multiple shards in parallel, one shard might write to the
cache while another one is reading this cache. Instead of updating the
file in place, write to a temporary file and swap the cache file using
os.replace(). This is an atomic operation and means shards will either
see the old state or the new one.
Thanks to Giuseppe D'Angelo for pointing this out on the cpplang Slack!
The example implementation in https://eel.is/c++draft/string.view.comparison#example-1
was necessary when it was written, in C++17, but in C++20 we don't need that
complexity anymore, because of the reversed candidates that are
synthesized by the compiler.
This gets rid of the separate parameter enable_modules_lsv in favor of
adding a named option to the enable_modules parameter. The patch also
removes the getModuleFlag helper, which was just a really complicated
way of hardcoding "none".
If an assertion fails during the configuration of the test suite because
the compiler doesn't support a flag (or the compiler configuration is
borked entirely), we will now print additional context to help debug the
issue instead of just saying "The compiler doesn't support the flag" and
bailing.
We were running the test suite in C++20 mode, but we should really be
tracking the latest supported version. We can't do C++26 because the
latest stable AppleClang doesn't support it yet.
This removes a lot of code duplication, makes the code simpler and
prepares the terrain for https://reviews.llvm.org/D101206, which will
fix some UB in the node-based containers.
This also allows removing the dependency of list and forward_list on
unique_ptr by using __allocation_guard instead.
Notice that because Holder<Incomplete> is _possible_ to complete, but
_unsafe_ to complete, that means that Holder<Incomplete>* is basically
not an iterator and it's not even safe to ask if
input_iterator<Holder<Incomplete>*> because that _will_ necessarily
complete the type. So it's totally expected that we still cannot safely
ask e.g.
static_assert(std::indirect_unary_predicate<bool(&)(Holder<Incomplete>&),
Holder<Incomplete>*>);
or even
static_assert(!std::indirect_unary_predicate<int, Holder<Incomplete>*>);
This was originally uploaded as https://reviews.llvm.org/D119029 and I
picked it up here as part of the Github PR transition.
Co-authored-by: Arthur O'Dwyer <arthur.j.odwyer@gmail.com>
When we implemented C++20's P0674R1, we didn't enable the part of
P0674R1 that was resolving LWG2070 as a DR. This patch fixes that and
makes sure that we consistently go through the allocator when
constructing and destroying the underlying object in
std::allocate_shared.
Fixes#54365.
This fixes some unused variable and "comparison of integers of different
signs" warnings in the benchmarks.
Differential Revision: https://reviews.llvm.org/D156613
This partially reverts commit e30a148b09, which removed the base
template for std::char_traits. That base template had been marked as
deprecated since LLVM 16 and we were planning to remove it in LLVM 18.
However, as explained in the post-commit comments in
https://reviews.llvm.org/D157058, the deprecation mechanism didn't work
as expected. Basically, the deprecation warnings were never shown to
users since libc++ headers are system headers and Clang doesn't show
warnings in system headers.
As a result, this removal came with basically no lead time as far as
users are concerned, which is a poor experience. For this reason, I am
re-introducing the deprecated char_traits specialization until we have a
proper way of phasing it out in a way that is not a surprise for users.
Since we use C++20 to build the dylib, we can use a lambda to do the
first-time initialization instead of emulating std::bind. This should
not change the behavior of the code at all, it merely simplifies it.
This removes a symbol from the dylib, however that symbol was only ever
used inside the dylib so it shouldn't break the ABI for anyone. I
confirmed that by searching for that symbol on the ABI boundary of a
large number of programs and couldn't find any references to that
function.
This "bug" was probably not noticed because it doesn't affect any integer
type we currently support. It requires integers with more than 2x the
size of `unsigned long long`. However, with such types, the algorithm
used to break down the large integer into groups of size `unsigned long long`
didn't work because we rotated in the wrong direction.
For example, the 256 bit number (1 << 255) would yield the wrong answer
when used with the algorithm before this patch.
In particular, note that the current rotation happens to work for 128 bit
integers because it just swaps the halves in this case.
Differential Revision: https://reviews.llvm.org/D134625
Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
The safe mode is in-between the hardened and the debug modes, extending
the checks contained in the hardened mode with certain checks that are
relatively cheap and prevent common sources of errors but aren't
security-critical. Thus, the safe mode trades off some performance for
a wider set of checks, but unlike the debug mode, it can still be used
in production.
Differential Revision: https://reviews.llvm.org/D158823
Was part of D118114. Unify test cases for `std::map::count` method and
add test case for the comparator that marked `final`
Differential Revision: https://reviews.llvm.org/D118232
[libcxx] <experimental/simd> Added internal storage type for class simd/simd_mask
[libcxx] <experimental/simd> Added all constructors of class simd/simd_mask and related tests
[libcxx] <experimental/simd> Added basic simd reference implementation, subscript operators of class simd/simd_mask and related tests
Reviewed By: #libc, philnik
Differential Revision: https://reviews.llvm.org/D144364
The dylib contains multiple global variables of type locale::id. Those
can be marked as constinit to make it clear that static initialization
is performed.
This is to fix an error that occurs when the char type is a class type.
Thanks to Yichen Yan for the patch.
Differential Revision: https://reviews.llvm.org/D100005
It's important that the arch directory be included first so that
its header files which interpose on the default include dir
be included instead of the default ones. The clang driver [1] does
this when not building with -nostdinc, the libcxx build should
do the same.
We found this after https://reviews.llvm.org/D154282 when cross
compiling from non Linux to Linux. If the host machine was not
Linux, _LIBCPP_HAS_NO_TIME_ZONE_DATABASE would be defined in
the default include dir __config_site, while it was undefined
in the arch specific one causing build failures.
The test is hardcoded to fail after passing `test_ill_formed_utf16()`. It passes on 32-bit AIX if we remove this.
Reviewed By: Mordante, #libc, ldionne
Differential Revision: https://reviews.llvm.org/D150273
This commit contains refactorings around __dynamic_cast without changing
its behavior. Some important changes include:
- Refactor __dynamic_cast into various small helper functions;
- Move dynamic_cast_stress.pass.cpp to libcxx/benchmarks and refactor
it into a benchmark. The benchmark performance numbers are updated
as well.
Differential Revision: https://reviews.llvm.org/D138006
On GLibc, FreeBSD and macOS systems nl_catd is a pointer type, and
round-tripping this in a variable of ptrdiff_t is not portable.
In fact such a round-trip yields a non-dereferenceable pointer on
CHERI-enabled architectures such as Arm Morello. There pointers (and
therefore intptr_t) are twice the size of ptrdiff_t, which means casting
to ptrdiff_t strips the high (metadata) bits (as well as a hidden pointer
validity bit).
Since catalog is now guaranteed to be the same size or larger than nl_catd,
we can store all return values safely and the shifting workaround from
commit 0c68ed006d should not be needed
anymore (this is also not portable to CHERI systems on since shifting a
valid pointer right will create a massively out-of-bounds pointer that
may not be representable).
This can be fixed by using intptr_t which should be the same type as
ptrdiff_t on all currently supported architectures.
See also: https://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#2028
Differential Revision: https://reviews.llvm.org/D134420
Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
Should be the same logic, but hopefully easier to read this way. Gets
rid of some superfluous state variables, and uses early returns.
Differential Revision: https://reviews.llvm.org/D112956
Fixes a conflict with adding the no experimental module build.
Disables some tests that need further investigation, this should fix the
CI runners.
These issues were reported on Discord and in D154282.
Add a warning to the `Building Libcxx` documentation about the
limitations of the utility of `--fresh` at the top level.
Signed-off-by: Will Hawkins <hawkinsw@obs.cr>
Add a [-b|--bootstrap] flag to change cxx-test-depends to
runtimes-test-depends when testing a build of libcxx from bootstrapping.
Signed-off-by: Will Hawkins <hawkinsw@obs.cr>
Differential Revision: https://reviews.llvm.org/D157045
After many years of using the really old std::pair ABI which did not yet
have a trivial copy constructor, FreeBSD 14 and later will finally get
rid of it. Only use the old ABI for FreeBSD 13 and earlier.
Note: on the FreeBSD side, we will bump our libc++.so version for this,
and keep an old compatibility library in a separate package.
Differential Revision: https://reviews.llvm.org/D126462
Prior to this patch, libcxx/test/libcxx/strings/basic.string/string.capacity/max_size.pass.cpp
would fail for RISC-V. The __riscv macro is defined for both RV32 and RV64.
Differential Revision: https://reviews.llvm.org/D143158
