Python3.12 Unable to attach to PID (Jupyter python process)

[fleming79]

With VScode I normally debug IpyWidgets by attaching to a Jupyter Kenrnel using PID. With python 3.11 this works with an external Jupyter kernel or one managed by VScode.

Switching to using Python 3.12 it no longer connects to the PID. The only indication is a message in the console Python:....

Attaching to PID: 19800 (elapsed: 1.00s).
Attaching to PID: 19800 (elapsed: 2.00s).

Stops counting at 2.00s.

With Python 3.11 it counts up to about 8s then connects.

Environment data

• debugpy version: 1.8.0
• OS and version: Windows 10
• Python version (& distribution if applicable, e.g. Anaconda): 3.12 WinPython distribution
• Using VS Code:
  Version: 1.84.2
  Commit: 1a5daa3a0231a0fbba4f14db7ec463cf99d7768e
  Date: 2023-11-09T10:51:52.184Z
  Electron: 25.9.2
  ElectronBuildId: 24603566
  Chromium: 114.0.5735.289
  Node.js: 18.15.0
  V8: 11.4.183.29-electron.0
  OS: Windows_NT x64 10.0.19045

Actual behavior

• Debugger doesn't attach (Timeout error).

Expected behavior

• Debugger attaches to PID after about 8s.

Steps to reproduce:

1. Start VScode.
2. Start Jupyter and obtain PID for a Python 3.12 kernel.
3. Attach to PID using Python debugger.

Launch settings

    {
        "name": "Python: Attach using Process Id",
        "type": "python",
        "request": "attach",
        "processId": "${command:pickProcess}",
        "subProcess": false
    }

[int19h]

Could you enable logs using "logToFile": true in your launch.json, and share debugpy.server*.log?

[fleming79]

Unfortunately I don't see a debugpy.server*.log file.

These may be relevant.

debugger.vscode_24ababd2-a99b-44d7-a395-aea0bf4197c5.log
debugpy.adapter-15968.log

For my ref: how to find log file

[int19h]

Indeed, my apologies; the "logToFile" setting wouldn't be applied in this case. Instead, set the DEBUGPY_LOG_DIR environment variable for the process you're attaching to point at the directory where it should place the logs.

[fleming79]

Thank you for the guidance.
debugpy.server-21680.log

[int19h]

Hm, was there only one server log generated? Normally there should be two in this scenario - the one above is from the process that performs the injection, but the second one should be from bits of the debugger that get injected.

[fleming79]

There were two files created at the same time.

Python 3.12 kernel.

debugpy.server-6620.log
debugpy.pydevd.6620.log

Python 3.11 kernel.

Similar log files (with same prefix) are generated when successfully connecting.
debugpy.server-19904.log
debugpy.pydevd.19904.log

[fleming79]

I switched to using a system environment variable instead of using WinPython to add the variable and now get three files:

debugpy.adapter-4172.log
debugpy.pydevd.6836.log
debugpy.server-6836.log

[fleming79]

As another hint, the console in which Jupyter is running emits the following message:

Python version unknown!
Error when injecting code in target process. Error code (on windows): 5

[omasoud]

I ran into the same error. Tried with the process being just python.exe too; same thing.

Python version unknown!
Error when injecting code in target process. Error code (on windows): 5

debugpy.adapter-22480.log
debugger.vscode_236ca7a3-324e-4f22-870f-7f2842027d18.log

I'm guessing it's ending up here.

[tonyroberts]

Seems like Python 3.12 isn't supported at the moment? GetPythonVersion returns PythonVersion_Unknown for Python 3.12

debugpy/src/debugpy/_vendored/pydevd/pydevd_attach_to_process/common/py_version.hpp

Line 74 in 42853a9

return PythonVersion_Unknown; // we don't care about 3.1 anymore...

[NitroTamagotschi]

I ran into the same error by using a simple Python script instead of a jupyter notebook.

[xuweiwang2021]

Any solution for this issue? I got the same error in Anaconda Prompt running python 3.12 when debug in VS Code by attach using Process Id.
conda 24.3.0
python 3.12.2, python 3.12.3
VS Code 1.89.0
Python version unknown!
Error when injecting code in target process. Error code (on windows): 5

[AdamYoblick]

Yeah this is a known issue. Attach to a python 3.12 process doesn't work yet because of all the internal changes in 3.12 that we haven't caught up with. In the meantime, 3.11 should work.