[Lldb-commits] [lldb] [lldb-dap] Adjusting how repl-mode auto determines commands vs variable expressions. (PR #78005)

[Walter Erquinigo via lldb-commits]

================
@@ -1125,21 +1126,33 @@ void request_completions(const llvm::json::Object &request) {
   }
   llvm::json::Array targets;
 
-  if (g_dap.DetectExpressionContext(frame, text) ==
-      ExpressionContext::Variable) {
-    char command[] = "expression -- ";
-    text = command + text;
-    offset += strlen(command);
-  }
-  lldb::SBStringList matches;
-  lldb::SBStringList descriptions;
+  if (!text.empty() &&
+      llvm::StringRef(text).starts_with(g_dap.command_escape_prefix)) {
+    text = text.substr(g_dap.command_escape_prefix.size());
+  }
+
+  // While the user is typing then we likely have an incomplete input and cannot
+  // reliably determine the precise intent (command vs variable), try completing
+  // the text as both a command and variable expression, if applicable.
----------------
walter-erquinigo wrote:

good idea!

https://github.com/llvm/llvm-project/pull/78005