[Lldb-commits] [lldb] [lldb] Change the implementation of Status to store an llvm::Error (NFC) (PR #106774)

[Pavel Labath via lldb-commits]

================
@@ -37,48 +39,75 @@ class raw_ostream;
 using namespace lldb;
 using namespace lldb_private;
 
-Status::Status() {}
+char CloneableError::ID;
+char MachKernelError::ID;
+char Win32Error::ID;
+char ExpressionError::ID;
+
+namespace {
+/// A std::error_code category for eErrorTypeGeneric.
+class GenericCategory : public std::error_category {
+  const char *name() const override { return "LLDBGenericCategory"; }
+  std::string message(int __ev) const override { return "generic LLDB error"; };
+};
+GenericCategory &generic_category() {
+  static GenericCategory g_generic_category;
+  return g_generic_category;
+}
+
+/// A std::error_code category for eErrorTypeExpression.
+class ExpressionCategory : public std::error_category {
+  const char *name() const override { return "LLDBExpressionCategory"; }
+  std::string message(int __ev) const override {
+    return ExecutionResultAsCString(static_cast<lldb::ExpressionResults>(__ev));
+  };
+};
+ExpressionCategory &expression_category() {
+  static ExpressionCategory g_expression_category;
+  return g_expression_category;
+}
----------------
labath wrote:

Do we really need this? My understanding was that the std::error_code conversion is for compatibility, and so we may not need it if we can ensure that everyone gets the error string via `info->message()` rather than `info->convertToErrorCode().message()`.

Since we did have this class up to this point, I don't think we have any instances of the second pattern.

https://github.com/llvm/llvm-project/pull/106774