Feature Request: Support for debugging multiprocessing applications

[TommasoBendinelli]

Hi pdbpp maintainers,

First of all, I'd like to thank you for your work on this fantastic debugger. It's a game-changer. I'm working extensively with PyTorch and often use multiple workers in the DataLoader. When I try to set a breakpoint inside the code run by multiple workers, I get the BdbQuit exception (on Ubuntu 22.04):

  raise exception
bdb.BdbQuit: Caught BdbQuit in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/csem/divr/users/tbe/repo/3dmoco/env/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 302, in _worker_loop
    data = fetcher.fetch(index)
  ...
  File "/local/user/toolsr64/linux64_u22_python38/usr/local/lib/python3.8/bdb.py", line 113, in dispatch_line
    if self.quitting: raise BdbQuit
bdb.BdbQuit

I believe that debugging such a setting would greatly benefit developers working on machine learning pipelines. Are there any plans to support this use case? I'm willing to contribute to the implementation of this feature, and I'd appreciate any guidance on where to start.

Thank you,
Tommaso