Merge branch 'trunk' into proposal-while-loops

.bazelrc

@@ -8,3 +8,9 @@ build --host_crosstool_top=@bazel_cc_toolchain
 build:force_local_bootstrap --repo_env=CARBON_FORCE_LOCAL_BOOTSTRAP_BUILD=1
 
 build:fuzzer --features=fuzzer
+
+# Force actions to have a UTF-8 language encoding.
+# TODO: Need to investigate what this should be on Windows, but at least for
+# Linux and macOS this seems strictly better than the Bazel default of just
+# `en_US`.
+build --action_env=LANG=en_US.UTF-8

.gitignore

@@ -10,3 +10,9 @@
 
 # VSCode creates this directory in ways that are hard to prevent.
 /.vscode/
+
+# Directories created by clangd
+/.cache/clangd
+
+# Compilation database used by clangd
+compile_commands.json

.pre-commit-config.yaml

@@ -17,7 +17,7 @@ repos:
       - id: check-executables-have-shebangs
       - id: check-merge-conflict
       - id: check-symlinks
-        exclude: '^website/jekyll/site/_includes$'
+        exclude: '^(website/jekyll/site/_includes|bazel-(clang-toolchain|execroot))$'
       - id: check-yaml
       - id: detect-private-key
       - id: end-of-file-fixer

WORKSPACE

@@ -6,26 +6,13 @@ workspace(name = "carbon")
 
 load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
 
-# We want to use LLVM via an external CMake build, so pull in the Bazel
-# infrastructure that provides direct CMake interfacing support.
-http_archive(
-    name = "rules_foreign_cc",
-    strip_prefix = "rules_foreign_cc-main",
-    url = "https://github.com/bazelbuild/rules_foreign_cc/archive/main.zip",
-)
-
 # Add Bazel's python rules.
 http_archive(
     name = "rules_python",
     sha256 = "b6d46438523a3ec0f3cead544190ee13223a52f6a6765a29eae7b7cc24cc83a0",
     url = "https://github.com/bazelbuild/rules_python/releases/download/0.1.0/rules_python-0.1.0.tar.gz",
 )
 
-# Set up necessary dependencies for working with the foreign C++ rules.
-load("@rules_foreign_cc//:workspace_definitions.bzl", "rules_foreign_cc_dependencies")
-
-rules_foreign_cc_dependencies()
-
 # Bootstrap a Clang and LLVM toolchain.
 load("//bazel/cc_toolchains:clang_bootstrap.bzl", "bootstrap_clang_toolchain")
 
@@ -69,8 +56,8 @@ llvm_zlib_system(name = "llvm_zlib")
 # Until then, this is pulling from that pull request's commit.
 http_archive(
     name = "rules_m4",
-    strip_prefix = "rules_m4-add-extra-copts",
     sha256 = "4d34917214e8890ad770bdf0c319c41c9201fffd770938b41a1d641d4b27e05c",
+    strip_prefix = "rules_m4-add-extra-copts",
     urls = ["https://github.com/chandlerc/rules_m4/archive/add-extra-copts.zip"],
 )
 
@@ -87,8 +74,8 @@ m4_register_toolchains(extra_copts = ["-w"])
 # Until then, this is pulling from that pull request's commit.
 http_archive(
     name = "rules_flex",
-    strip_prefix = "rules_flex-add-extra-copts",
     sha256 = "fd97c3ae23926507be1b95158a683cd41c628d201e852a325d38b5e9f821b752",
+    strip_prefix = "rules_flex-add-extra-copts",
     urls = ["https://github.com/chandlerc/rules_flex/archive/add-extra-copts.zip"],
 )
 
@@ -105,8 +92,8 @@ flex_register_toolchains(extra_copts = ["-w"])
 # Until then, this is pulling from that pull request's commit.
 http_archive(
     name = "rules_bison",
-    strip_prefix = "rules_bison-add-extra-copts",
     sha256 = "c6e926f15214d903966dc950d759ec69116db67f148be114c119e4def0551eaa",
+    strip_prefix = "rules_bison-add-extra-copts",
     urls = ["https://github.com/chandlerc/rules_bison/archive/add-extra-copts.zip"],
 )
 

bazel-clang-toolchain

@@ -0,0 +1 @@
+bazel-out/../../../external/bootstrap_clang_toolchain

bazel-execroot

@@ -0,0 +1 @@
+bazel-out/../../carbon

compile_flags.txt

@@ -43,35 +43,38 @@
 -iquote
 bazel-bin
 -iquote
-bazel-carbon-lang/external/llvm-project
+bazel-execroot/external/llvm-project
 -iquote
 bazel-bin/external/llvm-project
 -iquote
-bazel-carbon-lang/external/llvm_terminfo
+bazel-execroot/external/llvm_terminfo
 -iquote
 bazel-bin/external/llvm_terminfo
 -iquote
-bazel-carbon-lang/external/llvm_zlib
+bazel-execroot/external/llvm_zlib
 -iquote
 bazel-bin/external/llvm_zlib
 -iquote
-bazel-carbon-lang/external/bazel_tools
+bazel-execroot/external/bazel_tools
 -iquote
 bazel-bin/external/bazel_tools
--Ibazel-bin/external/llvm-project/llvm/_virtual_includes/gtest_internal_headers
+-Ibazel-execroot/external/llvm-project/llvm/utils/unittest/googletest/src
 -isystem
-bazel-carbon-lang/external/llvm-project/llvm/include
+bazel-execroot/external/llvm-project/llvm/include
 -isystem
 bazel-bin/external/llvm-project/llvm/include
 -isystem
-bazel-carbon-lang/external/llvm-project/llvm/utils/unittest/googlemock/include
+bazel-execroot/external/llvm-project/llvm/utils/unittest/googlemock/include
 -isystem
 bazel-bin/external/llvm-project/llvm/utils/unittest/googlemock/include
 -isystem
-bazel-carbon-lang/external/llvm-project/llvm/utils/unittest/googletest/include
+bazel-execroot/external/llvm-project/llvm/utils/unittest/googletest/include
 -isystem
 bazel-bin/external/llvm-project/llvm/utils/unittest/googletest/include
 -std=c++17
+-nostdinc++
+-isystem
+bazel-clang-toolchain/include/c++/v1
 -no-canonical-prefixes
 -Wno-builtin-macro-redefined
 -D__DATE__="redacted"

diagnostics/BUILD

@@ -13,11 +13,32 @@ cc_library(
     deps = ["@llvm-project//llvm:Support"],
 )
 
+cc_library(
+    name = "null_diagnostics",
+    hdrs = ["null_diagnostics.h"],
+    deps = [
+        ":diagnostic_emitter",
+        "@llvm-project//llvm:Support",
+    ],
+)
+
+cc_library(
+    name = "mocks",
+    testonly = 1,
+    hdrs = ["mocks.h"],
+    deps = [
+        ":diagnostic_emitter",
+        "@llvm-project//llvm:Support",
+        "@llvm-project//llvm:gmock",
+    ],
+)
+
 cc_test(
     name = "diagnostic_emitter_test",
     srcs = ["diagnostic_emitter_test.cpp"],
     deps = [
         ":diagnostic_emitter",
+        ":mocks",
         "@llvm-project//llvm:Support",
         "@llvm-project//llvm:gmock",
         "@llvm-project//llvm:gtest",

diagnostics/diagnostic_emitter.h

@@ -23,63 +23,119 @@ namespace Carbon {
 // TODO: turn this into a much more reasonable API when we add some actual
 // uses of it.
 struct Diagnostic {
+  struct Location {
+    // Name of the file or buffer that this diagnostic refers to.
+    std::string file_name;
+    // 1-based line number.
+    int32_t line_number;
+    // 1-based column number.
+    int32_t column_number;
+  };
+
+  Location location;
   llvm::StringRef short_name;
   std::string message;
 };
 
+// Receives diagnostics as they are emitted.
+class DiagnosticConsumer {
+ public:
+  virtual ~DiagnosticConsumer() = default;
+
+  // Handle a diagnostic.
+  virtual auto HandleDiagnostic(const Diagnostic& diagnostic) -> void = 0;
+};
+
+// An interface that can translate some representation of a location into a
+// diagnostic location.
+//
+// TODO: Revisit this once the diagnostics machinery is more complete and see
+// if we can turn it into a `std::function`.
+template <typename LocationT>
+class DiagnosticLocationTranslator {
+ public:
+  virtual ~DiagnosticLocationTranslator() = default;
+
+  [[nodiscard]] virtual auto GetLocation(LocationT loc)
+      -> Diagnostic::Location = 0;
+};
+
 // Manages the creation of reports, the testing if diagnostics are enabled, and
 // the collection of reports.
+//
+// This class is parameterized by a location type, allowing different
+// diagnostic clients to provide location information in whatever form is most
+// convenient for them, such as a position within a buffer when lexing, a token
+// when parsing, or a parse tree node when type-checking, and to allow unit
+// tests to be decoupled from any concrete location representation.
+template <typename LocationT>
 class DiagnosticEmitter {
  public:
-  using Callback = std::function<void(const Diagnostic&)>;
-
-  explicit DiagnosticEmitter(Callback callback)
-      : callback_(std::move(callback)) {}
+  // The `translator` and `consumer` are required to outlive the diagnostic
+  // emitter.
+  explicit DiagnosticEmitter(
+      DiagnosticLocationTranslator<LocationT>& translator,
+      DiagnosticConsumer& consumer)
+      : translator_(&translator), consumer_(&consumer) {}
   ~DiagnosticEmitter() = default;
 
   // Emits an error unconditionally.
   template <typename DiagnosticT>
-  auto EmitError(DiagnosticT diag) -> void {
-    callback_({.short_name = DiagnosticT::ShortName, .message = diag.Format()});
+  auto EmitError(LocationT location, DiagnosticT diag) -> void {
+    // TODO: Encode the diagnostic kind in the Diagnostic object rather than
+    // hardcoding an "error: " prefix.
+    consumer_->HandleDiagnostic({.location = translator_->GetLocation(location),
+                                 .short_name = DiagnosticT::ShortName,
+                                 .message = "error: " + diag.Format()});
   }
 
   // Emits a stateless error unconditionally.
   template <typename DiagnosticT>
-  auto EmitError() -> std::enable_if_t<std::is_empty_v<DiagnosticT>> {
-    EmitError<DiagnosticT>({});
+  auto EmitError(LocationT location)
+      -> std::enable_if_t<std::is_empty_v<DiagnosticT>> {
+    EmitError<DiagnosticT>(location, {});
   }
 
   // Emits a warning if `F` returns true.  `F` may or may not be called if the
   // warning is disabled.
   template <typename DiagnosticT>
-  auto EmitWarningIf(llvm::function_ref<bool(DiagnosticT&)> f) -> void {
-    // TODO(kfm): check if this warning is enabled
+  auto EmitWarningIf(LocationT location,
+                     llvm::function_ref<bool(DiagnosticT&)> f) -> void {
+    // TODO(kfm): check if this warning is enabled at `location`.
     DiagnosticT diag;
     if (f(diag)) {
-      callback_(
-          {.short_name = DiagnosticT::ShortName, .message = diag.Format()});
+      // TODO: Encode the diagnostic kind in the Diagnostic object rather than
+      // hardcoding a "warning: " prefix.
+      consumer_->HandleDiagnostic(
+          {.location = translator_->GetLocation(location),
+           .short_name = DiagnosticT::ShortName,
+           .message = "warning: " + diag.Format()});
     }
   }
 
  private:
-  Callback callback_;
+  DiagnosticLocationTranslator<LocationT>* translator_;
+  DiagnosticConsumer* consumer_;
 };
 
-inline auto ConsoleDiagnosticEmitter() -> DiagnosticEmitter& {
-  static auto* emitter = new DiagnosticEmitter(
-      [](const Diagnostic& d) { llvm::errs() << d.message << "\n"; });
-  return *emitter;
-}
+inline auto ConsoleDiagnosticConsumer() -> DiagnosticConsumer& {
+  struct Consumer : DiagnosticConsumer {
+    auto HandleDiagnostic(const Diagnostic& d) -> void override {
+      if (!d.location.file_name.empty()) {
+        llvm::errs() << d.location.file_name << ":" << d.location.line_number
+                     << ":" << d.location.column_number << ": ";
+      }
 
-inline auto NullDiagnosticEmitter() -> DiagnosticEmitter& {
-  static auto* emitter = new DiagnosticEmitter([](const Diagnostic&) {});
-  return *emitter;
+      llvm::errs() << d.message << "\n";
+    }
+  };
+  static auto* consumer = new Consumer;
+  return *consumer;
 }
 
 // CRTP base class for diagnostics with no substitutions.
 template <typename Derived>
 struct SimpleDiagnostic {
-  struct Substitutions {};
   static auto Format() -> std::string { return Derived::Message.str(); }
 };