[df] Ensure there is code to JIT before calling the interpreter

In an explicit multithreading scenario inside the `RLoopManager::Jit` method, a thread might move the global string with the code to JIT into its own function stack while another thread might still believe that string is not empty. This would make the other thread call `InterpreterCalc` with an empty string which would eventually lead to an exception. Check for the string emptiness twice, once at the beginning of the function, then again after taking the write lock so that if another thread already moved the string then the current thread will not continue with JITting.

Author: vepadulano

tree/dataframe/src/RLoopManager.cxx

@@ -805,18 +805,21 @@ void RLoopManager::CleanUpTask(TTreeReader *r, unsigned int slot)
 /// This method also clears the contents of GetCodeToJit().
 void RLoopManager::Jit()
 {
-   {
-      R__READ_LOCKGUARD(ROOT::gCoreMutex);
-      if (GetCodeToJit().empty()) {
-         R__LOG_INFO(RDFLogChannel()) << "Nothing to jit and execute.";
-         return;
-      }
+   R__READ_LOCKGUARD(ROOT::gCoreMutex);
+   if (GetCodeToJit().empty()) {
+      R__LOG_INFO(RDFLogChannel()) << "Nothing to jit and execute.";
+      return;
    }
 
-   const std::string code = []() {
-      R__WRITE_LOCKGUARD(ROOT::gCoreMutex);
-      return std::move(GetCodeToJit());
-   }();
+   R__WRITE_LOCKGUARD(ROOT::gCoreMutex);
+   // Check again if another thread has already cleared the global string
+   // with the code to JIT. Without this check, we could end up calling
+   // InterpreterCalc with an empty string, which would raise an exception.
+   if (GetCodeToJit().empty()) {
+      R__LOG_INFO(RDFLogChannel()) << "Nothing to jit and execute.";
+      return;
+   }
+   const std::string code = std::move(GetCodeToJit());
 
    TStopwatch s;
    s.Start();