Clang

Written by the LLVM Team

PreRelease

Warning

These are in-progress notes for the upcoming Clang release. Release notes for previous releases can be found on the Releases Page.

Introduction

This document contains the release notes for the Clang C/C++/Objective-C frontend, part of the LLVM Compiler Infrastructure, release . Here we describe the status of Clang in some detail, including major improvements from the previous release and new feature work. For the general LLVM release notes, see the LLVM documentation. For the libc++ release notes, see this page. All LLVM releases may be downloaded from the LLVM releases web site.

For more information about Clang or LLVM, including information about the latest release, please see the Clang Web Site or the LLVM Web Site.

Potentially Breaking Changes

These changes are ones which we think may surprise users when upgrading to Clang because of the opportunity they pose for disruption to existing code bases.

C/C++ Language Potentially Breaking Changes

Indirect edges of asm goto statements under certain circumstances may now be split. In previous releases of clang, that means for the following code the two inputs may have compared equal in the inline assembly. This is no longer guaranteed (and necessary to support outputs along indirect edges, which is now supported as of this release). This change is more consistent with the behavior of GCC.
```
foo: asm goto ("# %0 %1"::"i"(&&foo)::foo);
```
__builtin_object_size and __builtin_dynamic_object_size now add the sizeof the elements specified in designated initializers of flexible array members for structs that contain them. This change is more consistent with the behavior of GCC.

C++ Specific Potentially Breaking Changes

Clang won't search for coroutine_traits in std::experimental namespace any more. Clang will only search for std::coroutine_traits for coroutines then.
Clang no longer allows dereferencing of a void * as an extension. Clang 16 converted this to a default-error as -Wvoid-ptr-dereference, as well as a SFINAE error. This flag is still valid however, as it disables the equivalent warning in C.

ABI Changes in This Version

A bug in evaluating the ineligibility of some special member functions has been fixed. This can make some classes trivially copyable that were not trivially copyable before. (#62555)

What's New in Clang ?

Some of the major new features and improvements to Clang are listed here. Generic improvements to Clang as a whole or to its underlying infrastructure are described first, followed by language-specific sections with improvements to Clang's support for those languages.

C++ Language Changes

Improved -O0 code generation for calls to std::forward_like. Similarly to std::move, std::forward et al. it is now treated as a compiler builtin and implemented directly rather than instantiating the definition from the standard library.
Implemented CWG2518 which allows static_assert(false) to not be ill-formed when its condition is evaluated in the context of a template definition.

C++20 Feature Support

Implemented the rule introduced by CA104 for comparison of constraint-expressions. Improved support for out-of-line definitions of constrained templates. This fixes: #49620, #60231, #61414, #61809.
Lambda templates with a requires clause directly after the template parameters now parse correctly if the requires clause consists of a variable with a dependent type. (#61278)
Announced C++20 Coroutines is fully supported on all targets except Windows, which still has some stability and ABI issues.
Downgraded use of a reserved identifier in a module export declaration from an error to a warning under the -Wreserved-module-identifier warning group. This warning is enabled by default. This addresses #61446 and allows easier building of standard modules. This diagnostic may be strengthened into an error again in the future once there is a less fragile way to mark a module as being part of the implementation rather than a user module.
Clang now implements [temp.deduct]p9. Substitution failures inside lambdas from unevaluated contexts will be surfaced as errors. They were previously handled as SFINAE.
Clang now supports requires cplusplus20 for module maps.
Implemented missing parts of P2002R1: Consistent comparison operators
Clang now defines __cpp_consteval macro.
Implemented P1816R0: and P2082R1:, which allows CTAD for aggregates.

C++23 Feature Support

Implemented P2036R3: Change scope of lambda trailing-return-type and P2579R0 Mitigation strategies for P2036. These proposals modify how variables captured in lambdas can appear in trailing return type expressions and how their types are deduced therein, in all C++ language versions.
Implemented partial support for P2448R2: Relaxing some constexpr restrictions Explicitly defaulted functions no longer have to be constexpr-compatible but merely constexpr suitable. We do not support outside of defaulted special memeber functions the change that constexpr functions no longer have to be constexpr compatible but rather support a less restricted requirements for constexpr functions. Which include allowing non-literal types as return values and parameters, allow calling of non-constexpr functions and constructors.
Clang now supports requires cplusplus23 for module maps.
Implemented P2564R3: consteval needs to propagate up.

C++2c Feature Support

Compiler flags -std=c++2c and -std=gnu++2c have been added for experimental C++2c implementation work.
Implemented P2738R1: constexpr cast from void*.
Partially implemented P2361R6: Unevaluated strings. The changes to attributes declarations are not part of this release.
Implemented P2741R3: user-generated static_assert messages.

Resolutions to C++ Defect Reports

Implemented DR2397 which allows auto specifier for pointers and reference to arrays.

Implemented CWG2521 which reserves using __ in user-defined literal suffixes and deprecates literal operator function declarations using an identifier. Taught -Wuser-defined-literals for the former, on by default, and added -Wdeprecated-literal-operator for the latter, off by default for now.

// What follows is warned by -Wuser-defined-literals
// albeit "ill-formed, no diagnostic required".
// Its behavior is undefined, [reserved.names.general]p2.
string operator ""__i18n(const char*, std::size_t);

// Assume this declaration is not in the global namespace.
// -Wdeprecated-literal-operator diagnoses the extra space.
string operator "" _i18n(const char*, std::size_t);
//                ^ an extra space

C Language Changes

Support for outputs from asm goto statements along indirect edges has been added. (#53562)
Fixed a bug that prevented initialization of an _Atomic-qualified pointer from a null pointer constant.
Fixed a bug that prevented casting to an _Atomic-qualified type. (#39596)

Added an extension to _Generic which allows the first operand to be a type rather than an expression. The type does not undergo any conversions, which makes this feature suitable for matching qualified types, incomplete types, and function or array types.

const int i = 12;
_Generic(i, int : 0, const int : 1); // Warns about unreachable code, the
                                     // result is 0, not 1.
_Generic(typeof(i), int : 0, const int : 1); // Result is 1, not 0.

C2x Feature Support

Implemented the unreachable macro in freestanding <stddef.h> for WG14 N2826
Removed the ATOMIC_VAR_INIT macro in C2x and later standards modes, which implements WG14 N2886
Implemented WG14 N2934 which introduces the bool, static_assert, alignas, alignof, and thread_local keywords in C2x.
Implemented WG14 N2900 and WG14 N3011 which allows for empty braced initialization in C.
```
struct S { int x, y } s = {}; // Initializes s.x and s.y to 0
```
As part of this change, the -Wgnu-empty-initializer warning group was removed, as this is no longer a GNU extension but a C2x extension. You can use -Wno-c2x-extensions to silence the extension warning instead.

Updated the implementation of WG14 N3042 based on decisions reached during the WG14 CD Ballot Resolution meetings held in Jan and Feb 2023. This should complete the implementation of nullptr and nullptr_t in C. The specific changes are:

void func(nullptr_t);
func(0); // Previously required to be rejected, is now accepted.
func((void *)0); // Previously required to be rejected, is now accepted.

nullptr_t val;
val = 0; // Previously required to be rejected, is now accepted.
val = (void *)0; // Previously required to be rejected, is now accepted.

bool b = nullptr; // Was incorrectly rejected by Clang, is now accepted.

Implemented WG14 N3124 <https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3124.pdf>_, which allows any universal character name to appear in character and string literals.

Non-comprehensive list of changes in this release

Clang now saves the address of ABI-indirect function parameters on the stack, improving the debug information available in programs compiled without optimizations.
Clang now supports __builtin_nondeterministic_value that returns a nondeterministic value of the same type as the provided argument.
Clang now supports __builtin_FILE_NAME() which returns the same information as the __FILE_NAME__ macro (the presumed file name from the invocation point, with no path components included).
Clang now supports __builtin_assume_separate_storage that indicates that its arguments point to objects in separate storage allocations.
Clang now supports expressions in #pragma clang __debug dump.
Clang now supports declaration of multi-dimensional arrays with __declspec(property).
A new builtin type trait __is_trivially_equality_comparable has been added, which checks whether comparing two instances of a type is equivalent to memcmp(&lhs, &rhs, sizeof(T)) == 0.
Clang now ignores null directives outside of the include guard when deciding whether a file can be enabled for the multiple-include optimization.
Clang now support __builtin_FUNCSIG() which returns the same information as the __FUNCSIG__ macro (available only with -fms-extensions flag). This fixes (#58951).
Clang now supports the NO_COLOR environment variable as a way to disable color diagnostics.
Clang now supports __builtin_isfpclass, which checks if the specified floating-point value falls into any of the specified data classes.
Added __builtin_elementwise_round for builtin for floating point types. This allows access to llvm.round for arbitrary floating-point and vector of floating-point types.
Added __builtin_elementwise_rint for floating point types. This allows access to llvm.rint for arbitrary floating-point and vector of floating-point types.
Added __builtin_elementwise_nearbyint for floating point types. This allows access to llvm.nearbyint for arbitrary floating-point and vector of floating-point types.

New Compiler Flags

The flag -std=c++23 has been added. This behaves the same as the existing flag -std=c++2b.
-dumpdir has been implemented to specify auxiliary and dump output filenames for features like -gsplit-dwarf.
-fcaret-diagnostics-max-lines= has been added as a driver options, which lets users control the maximum number of source lines printed for a caret diagnostic.

Deprecated Compiler Flags

Modified Compiler Flags

clang -g -gsplit-dwarf a.c -o obj/x (compile and link) now generates the .dwo file at obj/x-a.dwo, instead of a file in the temporary directory (/tmp on *NIX systems, if none of the environment variables TMPDIR, TMP, and TEMP are specified).
-ffat-lto-objects can now be used to emit object files with both object code and LLVM bitcode. Previously this flag was ignored for GCC compatibility. (See related patch).

Removed Compiler Flags

The deprecated flag -fmodules-ts is removed. Please use -std=c++20 or higher to use standard C++ modules instead.
The deprecated flag -fcoroutines-ts is removed. Please use -std=c++20 or higher to use standard C++ coroutines instead.
The CodeGen flag -lower-global-dtors-via-cxa-atexit which affects how global destructors are lowered for MachO is removed without replacement. The default of -lower-global-dtors-via-cxa-atexit=true is now the only supported way.
The cc1 flag -no-opaque-pointers has been removed.

Attribute Changes in Clang

Introduced a new function attribute __attribute__((unsafe_buffer_usage)) to be worn by functions containing buffer operations that could cause out of bounds memory accesses. It emits warnings at call sites to such functions when the flag -Wunsafe-buffer-usage is enabled.
__declspec attributes can now be used together with the using keyword. Before the attributes on __declspec was ignored, while now it will be forwarded to the point where the alias is used.
Introduced a new USR (unified symbol resolution) clause inside of the existing __attribute__((external_source_symbol)) attribute. Clang's indexer uses the optional USR value when indexing Clang's AST. This value is expected to be generated by an external compiler when generating C++ bindings during the compilation of the foreign language sources (e.g. Swift).
The __has_attribute, __has_c_attribute and __has_cpp_attribute preprocessor operators now return 1 also for attributes defined by plugins.
Improve the AST fidelity of alignas and _Alignas attribute. Before, we model alignas(type-id) as though the user wrote alignas(alignof(type-id)), now we directly use alignas(type-id).

Improvements to Clang's diagnostics

We now generate a diagnostic for signed integer overflow due to unary minus in a non-constant expression context. (#31643)
Clang now warns by default for C++20 and later about deprecated capture of this with a capture default of =. This warning can be disabled with -Wno-deprecated-this-capture.
Clang had failed to emit some -Wundefined-internal for members of a local class if that class was first introduced with a forward declaration.
Diagnostic notes and fix-its are now generated for ifunc/alias attributes which point to functions whose names are mangled.
Diagnostics relating to macros on the command line of a preprocessed assembly file or precompiled header are now reported as coming from the file <command line> instead of <built-in>.
Clang constexpr evaluator now provides a more concise diagnostic when calling function pointer that is known to be null.
Clang now avoids duplicate warnings on unreachable [[fallthrough]]; statements previously issued from -Wunreachable-code and -Wunreachable-code-fallthrough by prioritizing -Wunreachable-code-fallthrough.
Clang now correctly diagnoses statement attributes [[clang::always_inline]] and [[clang::noinline]] when used on a statement with dependent call expressions.
Clang now checks for completeness of the second and third arguments in the conditional operator. (#59718)
There were some cases in which the diagnostic for the unavailable attribute might not be issued, this fixes those cases. (61815)
Clang now avoids unnecessary diagnostic warnings for obvious expressions in the case of binary operators with logical OR operations. (#57906)
Clang's "static assertion failed" diagnostic now points to the static assertion expression instead of pointing to the static_assert token. (#61951)
-Wformat now recognizes %lb for the printf/scanf family of functions. (#62247:).
Clang now diagnoses shadowing of lambda's template parameter by a capture. (#61105:).
Address a false positive in -Wpacked when applied to a non-pod type using Clang ABI >= 15. (#62353:, fallout from the non-POD packing ABI fix in LLVM 15).
Clang constexpr evaluator now prints subobject's name instead of its type in notes when a constexpr variable has uninitialized subobjects after its constructor call. (#58601)
Clang's -Wshadow warning now warns about shadowings by static local variables (#62850:).
Clang now warns when any predefined macro is undefined or redefined, instead of only some of them.
Clang now correctly diagnoses when the argument to alignas or _Alignas is an incomplete type. (#55175:, and fixes an incorrect mention of alignof in a diagnostic about alignas).
Clang will now show a margin with line numbers to the left of each line of code it prints for diagnostics. This can be disabled using -fno-diagnostics-show-line-numbers. At the same time, the maximum number of code lines it prints has been increased from 1 to 16. This can be controlled using -fcaret-diagnostics-max-lines=.
Clang no longer emits -Wunused-variable warnings for variables declared with __attribute__((cleanup(...))) to match GCC's behavior.
Clang now issues expected warnings for situations of comparing with NULL pointers. (#42992:)
Clang now diagnoses unused const-qualified variable template as "unused variable template" rather than "unused variable".
When diagnosing a constant expression where an enum without a fixed underlying type is set to a value outside the range of the enum's values, clang will now print the name of the enum in question.
Clang no longer diagnoses a read of an empty structure as use of an uninitialized variable. (#26842:)
The Fix-It emitted for unused labels used to expand to the next line, which caused visual oddities now that Clang shows more than one line of code snippet. This has been fixed and the Fix-It now only spans to the end of the :.
Clang now underlines the parameter list of function declaration when emitting a note about the mismatch in the number of arguments.
Clang now diagnoses unexpected tokens after a #pragma clang|GCC diagnostic push|pop directive. (#13920:)
Clang now does not try to analyze cast validity on variables with dependent alignment (#63007:).
Clang constexpr evaluator now displays member function calls more precisely by making use of the syntactical structure of function calls. This avoids display of syntactically invalid codes in diagnostics. (#57081:)
Clang no longer emits inappropriate notes about the loss of __unaligned qualifier on overload resolution, when the actual reason for the failure is loss of other qualifiers.

Clang's notes about unconvertible types in overload resolution failure now covers the source range of parameter declaration of the candidate function declaration.

Example Code:

void func(int aa, int bb);
void test() { func(1, "two"); }

BEFORE:

source:2:15: error: no matching function for call to 'func'
void test() { func(1, "two");  }
              ^~~~
source:1:6: note: candidate function not viable: no known conversion from 'const char[4]' to 'int' for 2nd argument
void func(int aa, int bb);
     ^

AFTER:

source:2:15: error: no matching function for call to 'func'
void test() { func(1, "two");  }
              ^~~~
source:1:6: note: candidate function not viable: no known conversion from 'const char[4]' to 'int' for 2nd argument
void func(int aa, int bb);
     ^            ~~~~~~

-Wformat cast fix-its will now suggest static_cast instead of C-style casts for C++ code.
-Wformat will no longer suggest a no-op fix-it for fixing scoped enum format warnings. Instead, it will suggest casting the enum object to the type specified in the format string.
Clang now emits -Wconstant-logical-operand warning even when constant logical operand is on left side. (#37919)

Bug Fixes in This Version

Added a new diagnostic warning group -Wdeprecated-redundant-constexpr-static-def, under the existing -Wdeprecated group. This controls warnings about out-of-line definitions of 'static constexpr' data members that are unnecessary from C++17 onwards.
Fix segfault while running clang-rename on a non existing file. (#36471)
Fix crash when diagnosing incorrect usage of _Nullable involving alias templates. (#60344)
Fix confusing warning message when /clang:-x is passed in clang-cl driver mode and emit an error which suggests using /TC or /TP clang-cl options instead. (#59307)
Fix assert that fails when the expression causing the this pointer to be captured by a block is part of a constexpr if statement's branch and instantiation of the enclosing method causes the branch to be discarded.
Fix __VA_OPT__ implementation so that it treats the concatenation of a non-placemaker token and placemaker token as a non-placemaker token. (#60268)
Fix crash when taking the address of a consteval lambda call operator. (#57682)
Clang now support export declarations in the language linkage. (#60405)
Fix aggregate initialization inside lambda constexpr. (#60936)
No longer issue a false positive diagnostic about a catch handler that cannot be reached despite being reachable. This fixes #61177 in anticipation of CWG2699 <https://wg21.link/CWG2699>_ being accepted by WG21.
Fix crash when parsing fold expression containing a delayed typo correction. (#61326)
Fix crash when dealing with some member accesses outside of class or member function context. (#37792) and (#48405)
Fix crash when using [[clang::always_inline]] or [[clang::noinline]] statement attributes on a call to a template function in the body of a template function.
Fix coroutines issue where get_return_object() result was always eagerly converted to the return type. Eager initialization (allowing RVO) is now only performed when these types match, otherwise deferred initialization is used, enabling short-circuiting coroutines use cases. This fixes (#56532) in anticipation of CWG2563 <https://cplusplus.github.io/CWG/issues/2563.html>_.
Fix highlighting issue with _Complex and initialization list with more than 2 items. (#61518)
Fix getSourceRange on VarTemplateSpecializationDecl and VarTemplatePartialSpecializationDecl, which represents variable with the initializer, so it behaves consistently with other VarDecls and ends on the last token of initializer, instead of right angle bracket of the template argument list.
Fix false-positive diagnostic issued for consteval initializers of temporary objects. (#60286)
Correct restriction of trailing requirements clauses on a templated function. Previously we only rejected non-'templated' things, but the restrictions ALSO need to limit non-defined/non-member functions as well. Additionally, we now diagnose requires on lambdas when not allowed, which we previously missed. (#61748)
Fix confusing diagnostic for incorrect use of qualified concepts names.
Fix handling of comments in function like macros so they are ignored in -CC mode. (#60887)
Fix incorrect merging of lambdas across modules. (#60985)
Fix crash when handling nested immediate invocations in initializers of global variables. (#58207)
Fix crash when generating code coverage information for PseudoObjectExpr in Clang AST. (#45481)
Fix the assertion hit when a template consteval function appears in a nested consteval/constexpr call chain. (#61142)
Clang now better diagnose placeholder types constrained with a concept that is not a type concept.
Fix crash when a doc comment contains a line splicing. (#62054)
Work around with a clang coverage crash which happens when visiting expressions/statements with invalid source locations in non-assert builds. Assert builds may still see assertions triggered from this.
Fix a failed assertion due to an invalid source location when trying to form a coverage report for an unresolved constructor expression. (#62105)
Fix defaulted equality operator so that it does not attempt to compare unnamed bit-fields. This fixes: (#61355) and (#61417)
Fix crash after suggesting typo correction to constexpr if condition. (#61885)
Clang constexpr evaluator now treats comparison of [[gnu::weak]]-attributed member pointer as an invalid expression.
Fix crash when member function contains invalid default argument. (#62122)
Fix crash when handling undefined template partial specialization (#61356)
Fix premature substitution into the constraints of an inherited constructor.
Fix crash when attempting to perform parenthesized initialization of an aggregate with a base class with only non-public constructors. (#62296)
Fix crash when handling initialization candidates for invalid deduction guide. (#62408)
Fix crash when redefining a variable with an invalid type again with an invalid type. (#62447)
Fix a stack overflow issue when evaluating consteval default arguments. (#60082)
Fix the assertion hit when generating code for global variable initializer of _BitInt(1) type. (#62207)
Fix lambdas and other anonymous function names not respecting -fdebug-prefix-map (#62192)
Fix crash when attempting to pass a non-pointer type as first argument of __builtin_assume_aligned. (#62305)
A default argument for a non-type template parameter is evaluated and checked at the point where it is required. This fixes: (#62224) and (#62596)
Fix an assertion when instantiating the body of a Class Template Specialization when it had been instantiated from a partial template specialization with different template arguments on the containing class. This fixes: (#60778).
Fix a crash when an enum constant has a dependent-type recovery expression for C. (#62446).
Propagate the value-dependent bit for VAArgExpr. Fixes a crash where a __builtin_va_arg call has invalid arguments. (#62711).
Fix crash on attempt to initialize union with flexible array member. (#61746).
Clang TextNodeDumper enabled through -ast-dump flag no longer evaluates the initializer of constexpr VarDecl if the declaration has a dependent type.
Match GCC's behavior for __builtin_object_size and __builtin_dynamic_object_size on structs containing flexible array members. (#62789).
Fix a crash when instantiating a non-type template argument in a dependent scope. (#62533).
Fix crash when diagnosing default comparison method. (#62791) and (#62102).
Fix crash when passing a braced initializer list to a parentehsized aggregate initialization expression. (#63008).
Reject increment of bool value in unevaluated contexts after C++17. (#47517).
Fix assertion and quality of diagnostic messages in a for loop containing multiple declarations and a range specifier (#63010).
Fix rejects-valid when consteval operator appears inside of a template. (#62886).
Fix crash for code using _Atomic types in C++ (See patch).
Fix crash when passing a value larger then 64 bits to the aligned attribute. (#50534).
CallExpr built for C error-recovery now is always type-dependent. Fixes a crash when we encounter a unresolved TypoExpr during diagnostic emission. (#50244).
Apply -fmacro-prefix-map to anonymous tags in template arguments (#63219).
Clang now properly diagnoses format string mismatches involving scoped enumeration types. A scoped enumeration type is not promoted to an integer type by the default argument promotions, and thus this is UB. Clang's behavior now matches GCC's behavior in C++. (#38717).
Fixed a failing assertion when implicitly defining a function within a GNU statement expression that appears outside of a function block scope. The assertion was benign outside of asserts builds and would only fire in C. (#48579).
Fixed a failing assertion when applying an attribute to an anonymous union. The assertion was benign outside of asserts builds and would only fire in C++. (#48512).
Fixed a failing assertion when parsing incomplete destructor. (#63503)
Fix C++17 mode assert when parsing malformed code and the compiler is attempting to see if it could be type template for class template argument deduction. This fixes (Issue 57495)
Fix missing destructor calls and therefore memory leaks in generated code when an immediate invocation appears as a part of an expression that produces temporaries. (#60709).
Fixed a missed integer overflow warning with temporary values. (#63629)
Fixed parsing of elaborated type specifier inside of a new expression. (#34341)
Clang now correctly evaluates __has_extension (cxx_defaulted_functions) and __has_extension (cxx_default_function_template_args) to 1. (#61758)
Stop evaluating a constant expression if the condition expression which in switch statement contains errors. (#63453 <https://github.com/llvm/llvm-project/issues/63453>_)
Fixed false positive error diagnostic when pack expansion appears in template parameters of a member expression. (#48731)
Fix the contains-errors bit not being set for DeclRefExpr that refers to a VarDecl with invalid initializer. This fixes: (#50236), (#50243), (#48636), (#50320).
Fix an assertion when using \u0024 ($) as an identifier, by disallowing that construct (#62133 <https://github.com/llvm/llvm-project/issues/38717>_).
Fix crash caused by PseudoObjectExprBitfields: NumSubExprs overflow. (#63169 <https://github.com/llvm/llvm-project/issues/63169>_)
Fix crash when casting an object to an array type. (#63758 <https://github.com/llvm/llvm-project/issues/63758>_)

Bug Fixes to Compiler Builtins

Bug Fixes to Attribute Support

Fixed a bug where attribute annotations on type specifiers (enums, classes, structs, unions, and scoped enums) were not properly ignored, resulting in misleading warning messages. Now, such attribute annotations are correctly ignored. (#61660)
GNU attributes preceding C++ style attributes on templates were not properly handled, resulting in compilation error. This has been corrected to match the behavior exhibited by GCC, which permits mixed ordering of GNU and C++ attributes.

Bug Fixes to C++ Support

Fix crash on invalid code when looking up a destructor in a templated class inside a namespace. (#59446)
Fix crash when evaluating consteval constructor of derived class whose base has more than one field. (#60166)
Fix an issue about decltype in the members of class templates derived from templates with related parameters. (#58674)
Fix incorrect deletion of the default constructor of unions in some cases. (#48416)
No longer issue a pre-C++23 compatibility warning in -pedantic mode regarding overloaded operator[] with more than one parameter or for static lambdas. (#61582)
Stop stripping CV qualifiers from the type of this when capturing it by value in a lambda. (#50866)
Fix ordering of function templates by constraints when they have template template parameters with different nested constraints.
Fix type equivalence comparison between auto types to take constraints into account.
Fix bug in the computation of the __has_unique_object_representations builtin for types with unnamed bitfields. (#61336)
Fix default member initializers sometimes being ignored when performing parenthesized aggregate initialization of templated types. (#62266)
Fix overly aggressive lifetime checks for parenthesized aggregate initialization. (#61567)
Fix a crash when expanding a pack as the index of a subscript expression.
Fix handling of constexpr dynamic memory allocations in template arguments. (#62462)
Some predefined expressions are now treated as string literals in MSVC compatibility mode. (#114)
Fix parsing of auto(x), when it is surrounded by parentheses. (#62494)
Fix handling of generic lambda used as template arguments. (#62611)
Allow omitting typename in the parameter declaration of a friend constructor declaration. (#63119)
Fix access of a friend class declared in a local class. Clang previously emitted an error when a friend of a local class tried to access it's private data members.
Allow abstract parameter and return types in functions that are either deleted or not defined. (#63012)
Fix handling of using-declarations in the init statements of for loop declarations. (#63627)
Fix crash when emitting diagnostic for out of order designated initializers in C++. (#63605)
Fix crash when using standard C++ modules with OpenMP. (#62359)
Fix crash when using consteval non static data member initialization in standard C++ modules. (#60275)
Fix handling of ADL for dependent expressions in standard C++ modules. (#60488)
Fix crash when combining -ftime-trace within standard C++ modules. (#60544)
Don't generate template specializations when importing standard C++ modules. (#60693)
Fix the visibility of initializer list in the importer of standard C++ modules. This addresses (#60775)
Allow the use of constrained friend in standard C++ modules. (#60890)
Don't evaluate initializer of used variables in every importer of standard C++ modules. (#61040)
Fix the issue that the default operator== in standard C++ modules will cause duplicate symbol linker error. (#61067)
Fix the false positive ODR check for template names. This addresses the issue that we can't include <ranges> in multiple module units. (#61317)
Fix crash for inconsistent exported declarations in standard C++ modules. (#61321)
Fix ignoring #pragma comment and #pragma detect_mismatch directives in standard C++ modules. (#61733)
Don't generate virtual tables if the class is defined in another module units for Itanium ABI. (#61940)
Fix false postive check for constrained satisfaction in standard C++ modules. (#62589)
Serialize the evaluated constant values for variable declarations in standard C++ modules. (#62796)
Merge lambdas in require expressions in standard C++ modules. (#63544)
Fix location of default member initialization in parenthesized aggregate initialization. (#63903)

Bug Fixes to AST Handling

Preserve namespace definitions that follow malformed declarations.

Miscellaneous Bug Fixes

Miscellaneous Clang Crashes Fixed

Dumping the AST to JSON no longer causes a failed assertion when targetting the Microsoft ABI and the AST to be dumped contains dependent names that would not typically be mangled. (#61440)

Target Specific Changes

AMDGPU Support

Linking for AMDGPU now uses --no-undefined by default. This causes undefined symbols in the created module to be a linker error. To prevent this, pass -Wl,--undefined if compiling directly, or -Xoffload-linker --undefined if using an offloading language.
The deprecated -mcode-object-v3 and -mno-code-object-v3 command-line options have been removed.
A new option -mprintf-kind has been introduced that controls printf lowering scheme. It is currently supported only for HIP and takes following values, hostcall - printing happens during kernel execution via series of hostcalls, The scheme requires the system to support pcie atomics.(default) buffered - Scheme uses a debug buffer to populate printf varargs, does not rely on pcie atomics support.

X86 Support

Add ISA of AMX-COMPLEX which supports tcmmimfp16ps and tcmmrlfp16ps.
Support ISA of SHA512.
- Support intrinsic of _mm256_sha512msg1_epi64.
- Support intrinsic of _mm256_sha512msg2_epi64.
- Support intrinsic of _mm256_sha512rnds2_epi64.
Support ISA of SM3.
- Support intrinsic of _mm_sm3msg1_epi32.
- Support intrinsic of _mm_sm3msg2_epi32.
- Support intrinsic of _mm_sm3rnds2_epi32.
Support ISA of SM4.
- Support intrinsic of _mm(256)_sm4key4_epi32.
- Support intrinsic of _mm(256)_sm4rnds4_epi32.
Support ISA of AVX-VNNI-INT16.
- Support intrinsic of _mm(256)_dpwsud(s)_epi32.
- Support intrinsic of _mm(256)_dpwusd(s)_epi32.
- Support intrinsic of _mm(256)_dpwuud(s)_epi32.

Arm and AArch64 Support

The hard-float ABI is now available in Armv8.1-M configurations that have integer MVE instructions (and therefore have FP registers) but no scalar or vector floating point computation. Previously, trying to select the hard-float ABI on such a target (via -mfloat-abi=hard or a triple ending in hf) would silently use the soft-float ABI instead.
Clang builtin __arithmetic_fence and the command line option -fprotect-parens are now enabled for AArch64.
Clang supports flag output operands by which conditions in the NZCV could be outputs of inline assembly for AArch64. This change is more consistent with the behavior of GCC.
```
// int a = foo(); int* b = bar();
asm("ands %w[a], %w[a], #3" : [a] "+r"(a), "=@cceq"(*b));
```
Fix a crash when preserve_all calling convention is used on AArch64. Issue 58145

Windows Support

LoongArch Support

Patchable function entry (-fpatchable-function-entry) is now supported on LoongArch.
Unaligned memory accesses can be toggled by -m[no-]unaligned-access or the aliases -m[no-]strict-align.

RISC-V Support

Added -mrvv-vector-bits= option to give an upper and lower bound on vector length. Valid values are powers of 2 between 64 and 65536. A value of 32 should eventually be supported. We also accept "zvl" to use the Zvl*b extension from -march or -mcpu to the be the upper and lower bound.
Fixed incorrect ABI lowering of _Float16 in the case of structs containing _Float16 that are eligible for passing via GPR+FPR or FPR+FPR.
Removed support for __attribute__((interrupt("user"))). User-level interrupts are not in version 1.12 of the privileged specification.
Added attribute(riscv_rvv_vector_bits(__riscv_v_fixed_vlen)) to allow the size of a RVV (RISC-V Vector) scalable type to be specified. This allows RVV scalable vector types to be used in structs or in global variables.
The rules for ordering of extensions in -march strings were relaxed. A canonical ordering is no longer enforced on z*, s*, and x* prefixed extensions.

CUDA/HIP Language Changes

Clang has been updated to align its default language standard for CUDA/HIP with that of C++. The standard has now been enhanced to gnu++17, supplanting the previously used c++14.

CUDA Support

Clang now supports CUDA SDK up to 12.1

AIX Support

Add an AIX-only link-time option, -mxcoff-build-id=0xHEXSTRING, to allow users to embed a hex id in their binary such that it's readable by the program itself. This option is an alternative to the --build-id=0xHEXSTRING GNU linker option which is currently not supported by the AIX linker.
Introduced the -mxcoff-roptr option to place constant objects with relocatable address values in the read-only data section. This option should be used with the -fdata-sections option, and is not supported with -fno-data-sections. When -mxcoff-roptr is in effect at link time, read-only data sections with relocatable address values that resolve to imported symbols are made writable.

WebAssembly Support

Shared library support (and PIC code generation) for WebAssembly is no longer limited to the Emscripten target OS and now works with other targets such as wasm32-wasi. Note that the format is not yet stable and may change between LLVM versions. Also, WASI does not yet have facilities to load dynamic libraries.

AVR Support

The definition of USHRT_MAX in the freestanding <limits.h> no longer overflows on AVR (where sizeof(int) == sizeof(unsigned short)). The type of USHRT_MAX is now unsigned int instead of int, as required by the C standard.

DWARF Support in Clang

Floating Point Support in Clang

Add __builtin_elementwise_log builtin for floating point types only.
Add __builtin_elementwise_log10 builtin for floating point types only.
Add __builtin_elementwise_log2 builtin for floating point types only.
Add __builtin_elementwise_exp builtin for floating point types only.
Add __builtin_elementwise_exp2 builtin for floating point types only.
Add __builtin_set_flt_rounds builtin for X86, x86_64, Arm and AArch64 only.

AST Matchers

Add coroutineBodyStmt matcher.
The hasBody matcher now matches coroutine body nodes in CoroutineBodyStmts.

clang-format

Add NextLineOnly style to option PackConstructorInitializers. Compared to NextLine style, NextLineOnly style will not try to put the initializers on the current line first, instead, it will try to put the initializers on the next line only.
Add additional Qualifier Ordering support for special cases such as templates, requires clauses, long qualified names.
Fix all known issues associated with LambdaBodyIndentation: OuterScope.
Add BracedInitializerIndentWidth which can be used to configure the indentation level of the contents of braced init lists.
Add KeepEmptyLinesAtEOF to keep empty lines at end of file.
Add RemoveParentheses to remove redundant parentheses.
Add TypeNames to treat listed non-keyword identifiers as type names.

libclang

Introduced the new function clang_CXXMethod_isExplicit, which identifies whether a constructor or conversion function cursor was marked with the explicit identifier.
Introduced the new CXIndex constructor function clang_createIndexWithOptions, which allows storing precompiled preambles in memory or overriding the precompiled preamble storage path.
Deprecated two functions clang_CXIndex_setGlobalOptions and clang_CXIndex_setInvocationEmissionPathOption in favor of the new function clang_createIndexWithOptions in order to improve thread safety.
Added check in clang_getFieldDeclBitWidth for whether a bit-field has an evaluable bit width. Fixes undefined behavior when called on a bit-field whose width depends on a template parameter.
Added CXBinaryOperatorKind and CXUnaryOperatorKind. (#29138)

Static Analyzer

Fix incorrect alignment attribute on the this parameter of certain non-complete destructors when using the Microsoft ABI. (#60465)

Sanitizers

Python Binding Changes

The following methods have been added:

clang_Location_isInSystemHeader exposed via the is_in_system_header property of the Location class.

Additional Information

A wide variety of additional information is available on the Clang web page. The web page contains versions of the API documentation which are up-to-date with the Git version of the source code. You can access versions of these documents specific to this release by going into the "clang/docs/" directory in the Clang tree.

If you have any questions or comments about Clang, please feel free to contact us on the Discourse forums (Clang Frontend category).

Files

ReleaseNotes.rst

Latest commit

History