fix: prevent race condition by setting placeholder before processing …

…task Co-authored-by: aider (gemini/gemini-2.5-pro) <aider@aider.chat>
cecli-dev · ErichBSchulz · Oct 29, 2025 · Oct 29, 2025 · Oct 29, 2025 · Oct 29, 2025
commit c7757a7e5880beb9fbcf97c06ec0d9db2cfbcade
diff --git a/aider/coders/base_coder.py b/aider/coders/base_coder.py
@@ -1099,7 +1099,7 @@ async def _run_patched(self, with_message=None, preproc=True):
                                 f"input_task is None={input_task is None}, "
                                 f"user_message is None={user_message is None}, "
                                 f"processing_task is None={processing_task is None}, "
-                                f"placeholder='{self.io.placeholder}'"
+                                f"placeholder={self.io.placeholder!r}"
                             )
                             self.logger.debug("Getting input from user")
                         if not self.suppress_announcements_for_next_prompt:
@@ -1156,6 +1156,8 @@ async def _run_patched(self, with_message=None, preproc=True):
                             self.io.stop_spinner()
 
                     if user_message and self.run_one_completed and self.compact_context_completed:
+                        if self.commands.is_command(user_message):
+                            self.io.placeholder = "Running command..."
                         processing_task = asyncio.create_task(
                             self._processing_logic(user_message, preproc)
                         )

diff --git a/aider/main.py b/aider/main.py
@@ -11,7 +11,7 @@
 from dataclasses import fields
 from pathlib import Path
 
-DEBUG_COUNTER = 4
+DEBUG_COUNTER = 5
 
 try:
     import git

diff --git a/debug_plan.md b/debug_plan.md
@@ -66,11 +66,23 @@ tracking to ensure clarity and correctness.
     message while a command is running. This will prevent the main loop from asking for new
     input until the command is finished. I will also increment the debug counter to version 3.
     *Status: This fix was not successful and led to the application hanging.*
-*   `[ ]` **Add more logging to diagnose race condition**
+*   `[✓]` **Add more logging to diagnose race condition**
     The application still tries to get input while a command is running. To understand why the
     control flow logic is failing, I will add more detailed logging to `aider/coders/base_coder.py`.
     This will capture the state of key variables (`confirmation_in_progress`, `input_task`,
     `user_message`, `processing_task`, and `self.io.placeholder`) just before the decision to
     request new user input is made. This should reveal the nature of the race condition or logical
     flaw. The debug counter will be incremented to version 4.
+*   `[✓]` **Analyze logs and identify race condition**
+    The logs from debug counter 4 confirm a race condition. The main loop in `_run_patched` checks
+    whether to ask for new user input immediately after creating a `processing_task` for the
+    command. However, the `io.placeholder` which should prevent this is only set *inside* the
+    `processing_task`. By the time the placeholder is set, the main loop has already incorrectly
+    decided to ask for new user input.
+*   `[ ]` **Propose and Test a Fix**
+    To fix the race condition, I will modify `aider/coders/base_coder.py` to set a generic
+    placeholder (`"Running command..."`) *before* the `processing_task` is created. This will
+    ensure the main loop sees the placeholder and waits for the command to complete. The more
+    specific placeholder and its cleanup will still be handled within `aider/commands.py`. The
+    debug counter will be incremented to version 5.