[lldb] Inline expression evaluator error visualization

This patch is a reworking of Pete Lawrence's (@PortalPete) proposal
for better expression evaluator error messages:
#80938

Before:

```
$ lldb -o "expr a+b"
(lldb) expr a+b
error: <user expression 0>:1:1: use of undeclared identifier 'a'
a+b
^
error: <user expression 0>:1:3: use of undeclared identifier 'b'
a+b
  ^
```

After:

```
(lldb) expr a+b
            ^ ^
            │ ╰─ error: use of undeclared identifier 'b'
            ╰─ error: use of undeclared identifier 'a'
```

This eliminates the confusing `<user expression 0>:1:3` source
location and avoids echoing the expression to the console again, which
results in a cleaner presentation that makes it easier to grasp what's
going on.

Author: adrian-prantl

lldb/include/lldb/Expression/DiagnosticManager.h

@@ -39,6 +39,7 @@ struct DiagnosticDetail {
     unsigned line = 0;
     uint16_t column = 0;
     uint16_t length = 0;
+    bool hidden = false;
     bool in_user_input = false;
   };
   /// Contains {{}, 1, 3, 3, true} in the example above.

lldb/include/lldb/Interpreter/CommandObject.h

@@ -340,6 +340,13 @@ class CommandObject : public std::enable_shared_from_this<CommandObject> {
       return false;
   }
 
+  /// Set the command input as it appeared in the terminal. This
+  /// is used to have errors refer directly to the original command.
+  void SetOriginalCommandString(std::string s) { m_original_command = s; }
+
+  /// \param offset_in_command is on what column \c args_string
+  /// appears, if applicable. This enables diagnostics that refer back
+  /// to the user input.
   virtual void Execute(const char *args_string,
                        CommandReturnObject &result) = 0;
 
@@ -404,6 +411,7 @@ class CommandObject : public std::enable_shared_from_this<CommandObject> {
   std::string m_cmd_help_short;
   std::string m_cmd_help_long;
   std::string m_cmd_syntax;
+  std::string m_original_command;
   Flags m_flags;
   std::vector<CommandArgumentEntry> m_arguments;
   lldb::CommandOverrideCallback m_deprecated_command_override_callback;

lldb/source/Commands/CommandObjectExpression.cpp

@@ -10,6 +10,7 @@
 
 #include "CommandObjectExpression.h"
 #include "lldb/Core/Debugger.h"
+#include "lldb/Expression/DiagnosticManager.h"
 #include "lldb/Expression/ExpressionVariable.h"
 #include "lldb/Expression/REPL.h"
 #include "lldb/Expression/UserExpression.h"
@@ -398,6 +399,122 @@ CanBeUsedForElementCountPrinting(ValueObject &valobj) {
   return Status();
 }
 
+static llvm::raw_ostream &PrintSeverity(Stream &stream,
+                                        lldb::Severity severity) {
+  llvm::HighlightColor color;
+  llvm::StringRef text;
+  switch (severity) {
+  case eSeverityError:
+    color = llvm::HighlightColor::Error;
+    text = "error: ";
+    break;
+  case eSeverityWarning:
+    color = llvm::HighlightColor::Warning;
+    text = "warning: ";
+    break;
+  case eSeverityInfo:
+    color = llvm::HighlightColor::Remark;
+    text = "note: ";
+    break;
+  }
+  return llvm::WithColor(stream.AsRawOstream(), color, llvm::ColorMode::Enable)
+         << text;
+}
+
+namespace lldb_private {
+// Public for unittesting.
+void RenderDiagnosticDetails(Stream &stream,
+                             std::optional<uint16_t> offset_in_command,
+                             bool multiline,
+                             llvm::ArrayRef<DiagnosticDetail> details) {
+  if (details.empty())
+    return;
+
+  if (!offset_in_command) {
+    for (const DiagnosticDetail &detail : details) {
+      PrintSeverity(stream, detail.severity);
+      stream << detail.rendered << '\n';
+    }
+    return;
+  }
+
+  // Print a line with caret indicator(s) below the lldb prompt + command.
+  const size_t padding = *offset_in_command;
+  stream << std::string(padding, ' ');
+
+  size_t offset = 1;
+  std::vector<DiagnosticDetail> remaining_details, other_details,
+      hidden_details;
+  for (const DiagnosticDetail &detail : details) {
+    if (multiline || !detail.source_location) {
+      other_details.push_back(detail);
+      continue;
+    }
+    if (detail.source_location->hidden) {
+      hidden_details.push_back(detail);
+      continue;
+    }
+    if (!detail.source_location->in_user_input) {
+      other_details.push_back(detail);
+      continue;
+    }
+
+    auto &loc = *detail.source_location;
+    remaining_details.push_back(detail);
+    if (offset > loc.column)
+      continue;
+    stream << std::string(loc.column - offset, ' ') << '^';
+    if (loc.length > 1)
+      stream << std::string(loc.length - 1, '~');
+    offset = loc.column + 1;
+  }
+  stream << '\n';
+
+  // Work through each detail in reverse order using the vector/stack.
+  bool did_print = false;
+  for (auto detail = remaining_details.rbegin();
+       detail != remaining_details.rend();
+       ++detail, remaining_details.pop_back()) {
+    // Get the information to print this detail and remove it from the stack.
+    // Print all the lines for all the other messages first.
+    stream << std::string(padding, ' ');
+    size_t offset = 1;
+    for (auto &remaining_detail :
+         llvm::ArrayRef(remaining_details).drop_back(1)) {
+      uint16_t column = remaining_detail.source_location->column;
+      stream << std::string(column - offset, ' ') << "│";
+      offset = column + 1;
+    }
+
+    // Print the line connecting the ^ with the error message.
+    uint16_t column = detail->source_location->column;
+    if (offset <= column)
+      stream << std::string(column - offset, ' ') << "╰─ ";
+
+    // Print a colorized string based on the message's severity type.
+    PrintSeverity(stream, detail->severity);
+
+    // Finally, print the message and start a new line.
+    stream << detail->message << '\n';
+    did_print = true;
+  }
+
+  // Print the non-located details.
+  for (const DiagnosticDetail &detail : other_details) {
+    PrintSeverity(stream, detail.severity);
+    stream << detail.rendered << '\n';
+    did_print = true;
+  }
+
+  // Print the hidden details as a last resort.
+  if (!did_print)
+    for (const DiagnosticDetail &detail : hidden_details) {
+      PrintSeverity(stream, detail.severity);
+      stream << detail.rendered << '\n';
+    }
+}
+} // namespace lldb_private
+
 bool CommandObjectExpression::EvaluateExpression(llvm::StringRef expr,
                                                  Stream &output_stream,
                                                  Stream &error_stream,
@@ -486,19 +603,37 @@ bool CommandObjectExpression::EvaluateExpression(llvm::StringRef expr,
 
         result.SetStatus(eReturnStatusSuccessFinishResult);
       } else {
-        const char *error_cstr = result_valobj_sp->GetError().AsCString();
-        if (error_cstr && error_cstr[0]) {
-          const size_t error_cstr_len = strlen(error_cstr);
-          const bool ends_with_newline = error_cstr[error_cstr_len - 1] == '\n';
-          if (strstr(error_cstr, "error:") != error_cstr)
-            error_stream.PutCString("error: ");
-          error_stream.Write(error_cstr, error_cstr_len);
-          if (!ends_with_newline)
-            error_stream.EOL();
+        // Retrieve the diagnostics.
+        std::vector<DiagnosticDetail> details;
+        llvm::consumeError(
+            llvm::handleErrors(result_valobj_sp->GetError().ToError(),
+                               [&](DetailedExpressionError &error) {
+                                 details.push_back(error.GetDetail());
+                               }));
+        // Find the position of the expression in the command.
+        std::optional<uint16_t> expr_pos;
+        size_t nchar = m_original_command.find(expr);
+        if (nchar != std::string::npos)
+          expr_pos = nchar + GetDebugger().GetPrompt().size();
+
+        if (!details.empty()) {
+          bool multiline = expr.contains('\n');
+          RenderDiagnosticDetails(error_stream, expr_pos, multiline, details);
         } else {
-          error_stream.PutCString("error: unknown error\n");
+          const char *error_cstr = result_valobj_sp->GetError().AsCString();
+          if (error_cstr && error_cstr[0]) {
+            const size_t error_cstr_len = strlen(error_cstr);
+            const bool ends_with_newline =
+                error_cstr[error_cstr_len - 1] == '\n';
+            if (strstr(error_cstr, "error:") != error_cstr)
+              error_stream.PutCString("error: ");
+            error_stream.Write(error_cstr, error_cstr_len);
+            if (!ends_with_newline)
+              error_stream.EOL();
+          } else {
+            error_stream.PutCString("error: unknown error\n");
+          }
         }
-
         result.SetStatus(eReturnStatusFailed);
       }
     }

lldb/source/Interpreter/CommandInterpreter.cpp

@@ -1887,7 +1887,8 @@ bool CommandInterpreter::HandleCommand(const char *command_line,
                                        CommandReturnObject &result,
                                        bool force_repeat_command) {
   std::string command_string(command_line);
-  std::string original_command_string(command_line);
+  std::string original_command_string(command_string);
+  std::string real_original_command_string(command_string);
 
   Log *log = GetLog(LLDBLog::Commands);
   llvm::PrettyStackTraceFormat stack_trace("HandleCommand(command = \"%s\")",
@@ -2076,6 +2077,7 @@ bool CommandInterpreter::HandleCommand(const char *command_line,
     }
 
     ElapsedTime elapsed(execute_time);
+    cmd_obj->SetOriginalCommandString(real_original_command_string);
     cmd_obj->Execute(remainder.c_str(), result);
   }
 

lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionParser.cpp

@@ -215,8 +215,6 @@ class ClangDiagnosticManagerAdapter : public clang::DiagnosticConsumer {
     m_os->flush();
 
     DiagnosticDetail detail;
-    bool make_new_diagnostic = true;
-
     switch (DiagLevel) {
     case DiagnosticsEngine::Level::Fatal:
     case DiagnosticsEngine::Level::Error:
@@ -230,9 +228,6 @@ class ClangDiagnosticManagerAdapter : public clang::DiagnosticConsumer {
       detail.severity = lldb::eSeverityInfo;
       break;
     case DiagnosticsEngine::Level::Note:
-      m_manager->AppendMessageToDiagnostic(m_output);
-      make_new_diagnostic = false;
-
       // 'note:' diagnostics for errors and warnings can also contain Fix-Its.
       // We add these Fix-Its to the last error diagnostic to make sure
       // that we later have all Fix-Its related to an 'error' diagnostic when
@@ -250,7 +245,6 @@ class ClangDiagnosticManagerAdapter : public clang::DiagnosticConsumer {
       AddAllFixIts(clang_diag, Info);
       break;
     }
-    if (make_new_diagnostic) {
       // ClangDiagnostic messages are expected to have no whitespace/newlines
       // around them.
       std::string stripped_output =
@@ -270,6 +264,7 @@ class ClangDiagnosticManagerAdapter : public clang::DiagnosticConsumer {
           loc.line = fsloc.getSpellingLineNumber();
           loc.column = fsloc.getSpellingColumnNumber();
           loc.in_user_input = filename == m_filename;
+          loc.hidden = filename.starts_with("<lldb wrapper ");
 
           // Find the range of the primary location.
           for (const auto &range : Info.getRanges()) {
@@ -299,7 +294,6 @@ class ClangDiagnosticManagerAdapter : public clang::DiagnosticConsumer {
         AddAllFixIts(new_diagnostic.get(), Info);
 
       m_manager->AddDiagnostic(std::move(new_diagnostic));
-    }
   }
 
   void BeginSourceFile(const LangOptions &LO, const Preprocessor *PP) override {

lldb/source/Plugins/TraceExporter/ctf/CommandObjectThreadTraceExportCTF.h

@@ -48,7 +48,7 @@ class CommandObjectThreadTraceExportCTF : public CommandObjectParsed {
   Options *GetOptions() override { return &m_options; }
 
 protected:
-  void DoExecute(Args &command, CommandReturnObject &result) override;
+  void DoExecute(Args &args, CommandReturnObject &result) override;
 
   CommandOptions m_options;
 };

lldb/source/Utility/Status.cpp

@@ -8,6 +8,7 @@
 
 #include "lldb/Utility/Status.h"
 
+#include "lldb/Expression/DiagnosticManager.h"
 #include "lldb/Utility/LLDBLog.h"
 #include "lldb/Utility/Log.h"
 #include "lldb/Utility/VASPrintf.h"
@@ -279,8 +280,6 @@ ErrorType Status::GetType() const {
     else if (error.convertToErrorCode().category() == lldb_generic_category() ||
              error.convertToErrorCode() == llvm::inconvertibleErrorCode())
       result = eErrorTypeGeneric;
-    else
-      result = eErrorTypeInvalid;
   });
   return result;
 }

lldb/test/API/commands/expression/diagnostics/TestExprDiagnostics.py

@@ -183,3 +183,53 @@ def test_source_locations_from_objc_modules(self):
         # The NSLog definition source line should be printed. Return value and
         # the first argument are probably stable enough that this test can check for them.
         self.assertIn("void NSLog(NSString *format", value.GetError().GetCString())
+
+    def test_command_expr_formatting(self):
+        """Test that the source and caret positions LLDB prints are correct"""
+        self.build()
+
+        (target, process, thread, bkpt) = lldbutil.run_to_source_breakpoint(
+            self, "// Break here", self.main_source_spec
+        )
+        frame = thread.GetFrameAtIndex(0)
+
+        def check(input_ref):
+            self.expect(input_ref[0], error=True, substrs=input_ref[1:])
+
+        check(
+            [
+                "expression -- a+b",
+                "              ^ ^",
+                "              │ ╰─ error: use of undeclared identifier 'b'",
+                "              ╰─ error: use of undeclared identifier 'a'",
+            ]
+        )
+
+        check(
+            [
+                "expr -- a",
+                "        ^",
+                "        ╰─ error: use of undeclared identifier 'a'",
+            ]
+        )
+        check(
+            [
+                "expr -i 0 -o 0 -- a",
+                "                  ^",
+                "                  ╰─ error: use of undeclared identifier 'a'",
+            ]
+        )
+
+        self.expect(
+            "expression --top-level -- template<typename T> T FOO(T x) { return x/2;}"
+        )
+        check(
+            [
+                'expression -- FOO("")',
+                "              ^",
+                "              ╰─ note: in instantiation of function template specialization 'FOO<const char *>' requested here",
+                "error: <user expression",
+                "invalid operands to binary expression",
+            ]
+        )
+        check(["expression --\na\n+\nb", "error: <user", "a", "error: <user", "b"])

lldb/test/API/commands/expression/persistent_variables/TestPersistentVariables.py

@@ -56,7 +56,7 @@ def test_persistent_variables(self):
         self.expect(
             "expr int $i = 123",
             error=True,
-            substrs=["error: redefinition of persistent variable '$i'"],
+            substrs=["redefinition of persistent variable '$i'"],
         )
         self.expect_expr("$i", result_type="int", result_value="5")
 
@@ -65,7 +65,7 @@ def test_persistent_variables(self):
         self.expect(
             "expr long $i = 123",
             error=True,
-            substrs=["error: redefinition of persistent variable '$i'"],
+            substrs=["redefinition of persistent variable '$i'"],
         )
         self.expect_expr("$i", result_type="int", result_value="5")
 

lldb/test/API/commands/expression/static-initializers/TestStaticInitializers.py

@@ -35,7 +35,7 @@ def test_failing_init(self):
         self.expect(
             "expr -p -- struct Foo2 { Foo2() { do_abort(); } }; Foo2 f;",
             error=True,
-            substrs=["error: couldn't run static initializer:"],
+            substrs=["couldn't run static initializer:"],
         )
 
     def test_without_process(self):

lldb/test/API/lang/cpp/template-function/TestTemplateFunctions.py

@@ -21,7 +21,8 @@ def do_test_template_function(self, add_cast):
                 "expr b1 <=> b2",
                 error=True,
                 substrs=[
-                    "warning: <user expression 0>:1:4: '<=>' is a single token in C++20; add a space to avoid a change in behavior"
+                    "warning:",
+                    "'<=>' is a single token in C++20; add a space to avoid a change in behavior",
                 ],
             )
 

lldb/test/API/lang/mixed/TestMixedLanguages.py

@@ -40,7 +40,7 @@ def cleanup():
         self.runCmd("run")
         self.expect("thread backtrace", substrs=["`main", "lang=c"])
         # Make sure evaluation of C++11 fails.
-        self.expect("expr foo != nullptr", error=True, startstr="error")
+        self.expect("expr foo != nullptr", error=True, substrs=["error"])
 
         # Run to BP at foo (in foo.cpp) and test that the language is C++.
         self.runCmd("breakpoint set -n foo")