Injection stuck bug after making injection unsafe again

madbg/debugger.py

@@ -86,12 +86,13 @@ def __del__(self):
         self.pty.close()
 
     def attach(self):
-
         def set():
-            f = currentframe().f_back.f_back
+            f = currentframe().f_back.f_back.f_back
             f.f_globals[self._DEBUGGING_GLOBAL] = True
             self.check_debugging_global = True
             self.set_trace(f)
+        # If this gets stuck, it's probably because the debugger kicks in before the injection finishes.
+        # The debugging global must be pushed to the right frame to prevent that.
         run_in_thread(self.thread, set)
 
     def notify_client_connect(self, tty_config: TTYConfig):

madbg/server.py

@@ -158,14 +158,9 @@ def make_sure_listening_at(cls, addr: Addr):
 
 
 """
-app - show periodically updated stack traces
-deadlock detection support
-
 Next steps:
-    - revert safe injection diff in hypno
+    - sqlite errors
     - when writing ? in the terminal, "Object `` not found." is printed to stdout 
-    - no \r printed...
-    - continue raises exceptions
     - run in thread using ptrace - better than signal??
         signal: interfering with signal handlers
         ptrace: invoke subprocess, load dll
@@ -183,6 +178,8 @@ def make_sure_listening_at(cls, addr: Addr):
       the app could be configured with the first user's terminal, and connected through a raw pty to each
       user's pty
       so each thread has a pty, and each user has a pty
+    - app - show periodically updated stack traces
+    - deadlock detection support
     - pyinjector issues:
         - getting the python error back to us or at least know that it failed
         - threads