Fix a bug with where the executors get inserted during EXTENDED_ARG

Author: Fidget-Spinner

Include/internal/pycore_uop.h

@@ -36,7 +36,7 @@ typedef struct _PyUOpInstruction{
 } _PyUOpInstruction;
 
 // This is the length of the trace we translate initially.
-#define UOP_MAX_TRACE_LENGTH 1400
+#define UOP_MAX_TRACE_LENGTH 1500
 #define UOP_BUFFER_SIZE (UOP_MAX_TRACE_LENGTH * sizeof(_PyUOpInstruction))
 
 /* Bloom filter with m = 256

Python/bytecodes.c

@@ -2979,7 +2979,13 @@ dummy_func(
                 }
                 int _is_sys_tracing = (tstate->c_tracefunc != NULL) || (tstate->c_profilefunc != NULL);
                 if (!_is_sys_tracing) {
-                    _PyJIT_InitializeTracing(tstate, frame, this_instr, next_instr, STACK_LEVEL(), 0, NULL);
+                    /* Back up over EXTENDED_ARGs so executor is inserted at the corret place */
+                    _Py_CODEUNIT *insert_exec_at = this_instr;
+                    while (oparg > 255) {
+                        oparg >>= 8;
+                        insert_exec_at--;
+                    }
+                    _PyJIT_InitializeTracing(tstate, frame, insert_exec_at, next_instr, STACK_LEVEL(), 0, NULL);
                     ENTER_TRACING();
                 }
                 int _jump_taken = false;