Merge branch 'llvm:main' into nested-impl-do-data

Author: eugeneepshteyn

clang/docs/ClangFormatStyleOptions.rst

@@ -2582,9 +2582,9 @@ the configuration (without a prefix: ``Auto``).
 
 .. _BracedInitializerIndentWidth:
 
-**BracedInitializerIndentWidth** (``Unsigned``) :versionbadge:`clang-format 17` :ref:`¶ <BracedInitializerIndentWidth>`
+**BracedInitializerIndentWidth** (``Integer``) :versionbadge:`clang-format 17` :ref:`¶ <BracedInitializerIndentWidth>`
   The number of columns to use to indent the contents of braced init lists.
-  If unset, ``ContinuationIndentWidth`` is used.
+  If unset or negative, ``ContinuationIndentWidth`` is used.
 
   .. code-block:: c++
 

clang/docs/ReleaseNotes.rst

@@ -225,6 +225,8 @@ Bug Fixes in This Version
 
 - Clang now outputs correct values when #embed data contains bytes with negative
   signed char values (#GH102798).
+- Fixed rejects-valid problem when #embed appears in std::initializer_list or
+  when it can affect template argument deduction (#GH122306).
 
 Bug Fixes to Compiler Builtins
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

clang/include/clang/AST/Expr.h

@@ -5189,6 +5189,16 @@ class InitListExpr : public Expr {
 
   unsigned getNumInits() const { return InitExprs.size(); }
 
+  /// getNumInits but if the list has an EmbedExpr inside includes full length
+  /// of embedded data.
+  unsigned getNumInitsWithEmbedExpanded() const {
+    unsigned Sum = InitExprs.size();
+    for (auto *IE : InitExprs)
+      if (auto *EE = dyn_cast<EmbedExpr>(IE))
+        Sum += EE->getDataElementCount() - 1;
+    return Sum;
+  }
+
   /// Retrieve the set of initializers.
   Expr **getInits() { return reinterpret_cast<Expr **>(InitExprs.data()); }
 

clang/include/clang/Basic/BuiltinsAMDGPU.def

@@ -544,29 +544,29 @@ TARGET_BUILTIN(__builtin_amdgcn_wmma_f32_16x16x16_bf8_fp8_w64_gfx12, "V4fiiV4f",
 TARGET_BUILTIN(__builtin_amdgcn_wmma_f32_16x16x16_bf8_bf8_w64_gfx12, "V4fiiV4f", "nc", "gfx12-insts,wavefrontsize64")
 TARGET_BUILTIN(__builtin_amdgcn_wmma_i32_16x16x32_iu4_w64_gfx12, "V4iIbiIbiV4iIb", "nc", "gfx12-insts,wavefrontsize64")
 
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_f16_w32, "V8fV8hV16hV8fs", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf16_w32, "V8fV8sV16sV8fs", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f16_16x16x32_f16_w32, "V8hV8hV16hV8hs", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_bf16_16x16x32_bf16_w32, "V8sV8sV16sV8ss", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu8_w32, "V8iIbV2iIbV4iV8isIb", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu4_w32, "V8iIbiIbV2iV8isIb", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x64_iu4_w32, "V8iIbV2iIbV4iV8isIb", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_fp8_w32, "V8fV2iV4iV8fs", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_bf8_w32, "V8fV2iV4iV8fs", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_fp8_w32, "V8fV2iV4iV8fs", "nc", "gfx12-insts,wavefrontsize32")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w32, "V8fV2iV4iV8fs", "nc", "gfx12-insts,wavefrontsize32")
-
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_f16_w64, "V4fV4hV8hV4fs", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf16_w64, "V4fV4sV8sV4fs", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f16_16x16x32_f16_w64, "V4hV4hV8hV4hs", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_bf16_16x16x32_bf16_w64, "V4sV4sV8sV4ss", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu8_w64, "V4iIbiIbV2iV4isIb", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu4_w64, "V4iIbiIbiV4isIb", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x64_iu4_w64, "V4iIbiIbV2iV4isIb", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_fp8_w64, "V4fiV2iV4fs", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_bf8_w64, "V4fiV2iV4fs", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_fp8_w64, "V4fiV2iV4fs", "nc", "gfx12-insts,wavefrontsize64")
-TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w64, "V4fiV2iV4fs", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_f16_w32, "V8fV8hV16hV8fi", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf16_w32, "V8fV8sV16sV8fi", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f16_16x16x32_f16_w32, "V8hV8hV16hV8hi", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_bf16_16x16x32_bf16_w32, "V8sV8sV16sV8si", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu8_w32, "V8iIbV2iIbV4iV8iiIb", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu4_w32, "V8iIbiIbV2iV8iiIb", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x64_iu4_w32, "V8iIbV2iIbV4iV8iiIb", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_fp8_w32, "V8fV2iV4iV8fi", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_bf8_w32, "V8fV2iV4iV8fi", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_fp8_w32, "V8fV2iV4iV8fi", "nc", "gfx12-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w32, "V8fV2iV4iV8fi", "nc", "gfx12-insts,wavefrontsize32")
+
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_f16_w64, "V4fV4hV8hV4fi", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf16_w64, "V4fV4sV8sV4fi", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f16_16x16x32_f16_w64, "V4hV4hV8hV4hi", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_bf16_16x16x32_bf16_w64, "V4sV4sV8sV4si", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu8_w64, "V4iIbiIbV2iV4iiIb", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x32_iu4_w64, "V4iIbiIbiV4iiIb", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_i32_16x16x64_iu4_w64, "V4iIbiIbV2iV4iiIb", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_fp8_w64, "V4fiV2iV4fi", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_fp8_bf8_w64, "V4fiV2iV4fi", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_fp8_w64, "V4fiV2iV4fi", "nc", "gfx12-insts,wavefrontsize64")
+TARGET_BUILTIN(__builtin_amdgcn_swmmac_f32_16x16x32_bf8_bf8_w64, "V4fiV2iV4fi", "nc", "gfx12-insts,wavefrontsize64")
 
 TARGET_BUILTIN(__builtin_amdgcn_prng_b32, "UiUi", "nc", "prng-inst")
 TARGET_BUILTIN(__builtin_amdgcn_cvt_scalef32_pk32_fp6_f16, "V6UiV32hf", "nc", "f16bf16-to-fp6bf6-cvt-scale-insts")

clang/include/clang/Basic/DiagnosticDriverKinds.td

@@ -563,6 +563,11 @@ def err_test_module_file_extension_format : Error<
 def err_drv_module_output_with_multiple_arch : Error<
   "option '-fmodule-output' cannot be used with multiple arch options">;
 
+def err_drv_reduced_module_output_overrided : Warning<
+  "the implicit output of reduced BMI may be overrided by the output file specified by '--precompile'. "
+  "please consider use '-fmodule-output=' to specify the output file for reduced BMI explicitly">,
+  InGroup<DiagGroup<"reduced-bmi-output-overrided">>;
+
 def warn_drv_delayed_template_parsing_after_cxx20 : Warning<
   "-fdelayed-template-parsing is deprecated after C++20">,
   InGroup<DiagGroup<"delayed-template-parsing-in-cxx20">>;

clang/include/clang/Driver/Options.td

@@ -3176,7 +3176,7 @@ def modules_reduced_bmi : Flag<["-"], "fmodules-reduced-bmi">,
   HelpText<"Generate the reduced BMI">,
   MarshallingInfoFlag<FrontendOpts<"GenReducedBMI">>;
 
-def experimental_modules_reduced_bmi : Flag<["-"], "fexperimental-modules-reduced-bmi">, 
+def experimental_modules_reduced_bmi : Flag<["-"], "fexperimental-modules-reduced-bmi">,
   Group<f_Group>, Visibility<[ClangOption, CC1Option]>, Alias<modules_reduced_bmi>;
 
 def fmodules_embed_all_files : Joined<["-"], "fmodules-embed-all-files">,
@@ -7431,7 +7431,7 @@ def fuse_register_sized_bitfield_access: Flag<["-"], "fuse-register-sized-bitfie
 def relaxed_aliasing : Flag<["-"], "relaxed-aliasing">,
   HelpText<"Turn off Type Based Alias Analysis">,
   MarshallingInfoFlag<CodeGenOpts<"RelaxedAliasing">>;
-defm pointer_tbaa: BoolOption<"", "pointer-tbaa", CodeGenOpts<"PointerTBAA">, 
+defm pointer_tbaa: BoolOption<"", "pointer-tbaa", CodeGenOpts<"PointerTBAA">,
 DefaultTrue,
   PosFlag<SetTrue, [], [ClangOption], "Enable">,
   NegFlag<SetFalse, [], [ClangOption], "Disable">,

clang/include/clang/Format/Format.h

@@ -1289,7 +1289,7 @@ struct FormatStyle {
   BitFieldColonSpacingStyle BitFieldColonSpacing;
 
   /// The number of columns to use to indent the contents of braced init lists.
-  /// If unset, ``ContinuationIndentWidth`` is used.
+  /// If unset or negative, ``ContinuationIndentWidth`` is used.
   /// \code
   ///   AlignAfterOpenBracket: AlwaysBreak
   ///   BracedInitializerIndentWidth: 2
@@ -1319,7 +1319,7 @@ struct FormatStyle {
   ///   }
   /// \endcode
   /// \version 17
-  std::optional<unsigned> BracedInitializerIndentWidth;
+  int BracedInitializerIndentWidth;
 
   /// Different ways to wrap braces after control statements.
   enum BraceWrappingAfterControlStatementStyle : int8_t {

clang/lib/CodeGen/CGCoroutine.cpp

@@ -855,6 +855,20 @@ void CodeGenFunction::EmitCoroutineBody(const CoroutineBodyStmt &S) {
     // Create parameter copies. We do it before creating a promise, since an
     // evolution of coroutine TS may allow promise constructor to observe
     // parameter copies.
+    for (const ParmVarDecl *Parm : FnArgs) {
+      // If the original param is in an alloca, exclude it from the coroutine
+      // frame. The parameter copy will be part of the frame, but the original
+      // parameter memory should remain on the stack. This is necessary to
+      // ensure that parameters destroyed in callees, as with `trivial_abi` or
+      // in the MSVC C++ ABI, are appropriately destroyed after setting up the
+      // coroutine.
+      Address ParmAddr = GetAddrOfLocalVar(Parm);
+      if (auto *ParmAlloca =
+              dyn_cast<llvm::AllocaInst>(ParmAddr.getBasePointer())) {
+        ParmAlloca->setMetadata(llvm::LLVMContext::MD_coro_outside_frame,
+                                llvm::MDNode::get(CGM.getLLVMContext(), {}));
+      }
+    }
     for (auto *PM : S.getParamMoves()) {
       EmitStmt(PM);
       ParamReplacer.addCopy(cast<DeclStmt>(PM));
@@ -942,9 +956,16 @@ void CodeGenFunction::EmitCoroutineBody(const CoroutineBodyStmt &S) {
   if (Stmt *Ret = S.getReturnStmt()) {
     // Since we already emitted the return value above, so we shouldn't
     // emit it again here.
-    if (GroManager.DirectEmit)
+    Expr *PreviousRetValue = nullptr;
+    if (GroManager.DirectEmit) {
+      PreviousRetValue = cast<ReturnStmt>(Ret)->getRetValue();
       cast<ReturnStmt>(Ret)->setRetValue(nullptr);
+    }
     EmitStmt(Ret);
+    // Set the return value back. The code generator, as the AST **Consumer**,
+    // shouldn't change the AST.
+    if (PreviousRetValue)
+      cast<ReturnStmt>(Ret)->setRetValue(PreviousRetValue);
   }
 
   // LLVM require the frontend to mark the coroutine.

clang/lib/Driver/ToolChains/Clang.cpp

@@ -4250,10 +4250,18 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D,
 
     if (Args.hasArg(options::OPT_fmodule_output_EQ))
       Args.AddLastArg(CmdArgs, options::OPT_fmodule_output_EQ);
-    else
+    else {
+      if (Args.hasArg(options::OPT__precompile) &&
+          (!Args.hasArg(options::OPT_o) ||
+           Args.getLastArg(options::OPT_o)->getValue() ==
+               getCXX20NamedModuleOutputPath(Args, Input.getBaseInput()))) {
+        D.Diag(diag::err_drv_reduced_module_output_overrided);
+      }
+
       CmdArgs.push_back(Args.MakeArgString(
           "-fmodule-output=" +
           getCXX20NamedModuleOutputPath(Args, Input.getBaseInput())));
+    }
   }
 
   // Noop if we see '-fmodules-reduced-bmi' with other translation

clang/lib/Format/ContinuationIndenter.cpp

@@ -1921,9 +1921,9 @@ void ContinuationIndenter::moveStatePastScopeOpener(LineState &State,
       NewIndent = Style.IndentWidth +
                   std::min(State.Column, CurrentState.NestedBlockIndent);
     } else if (Current.is(tok::l_brace)) {
-      NewIndent =
-          CurrentState.LastSpace + Style.BracedInitializerIndentWidth.value_or(
-                                       Style.ContinuationIndentWidth);
+      const auto Width = Style.BracedInitializerIndentWidth;
+      NewIndent = CurrentState.LastSpace +
+                  (Width < 0 ? Style.ContinuationIndentWidth : Width);
     } else {
       NewIndent = CurrentState.LastSpace + Style.ContinuationIndentWidth;
     }