REPL-driven, multi-process-native debugger features

[goodboy]

In tractor every actor is both a (potential) client and server. Concocting a "native" feeling "remote debugger" shouldn't be that bad (they thought naively..), right?

I did a little digging and asked all the cool peeps what they use. Well here's a list of interesting stuff:

popular Python remote debuggers

• qdb from quantopian
• ptvsd the python debugger for VS from the microsoft
• celery.contrib.rdb from the celery peeps
• python-remote-pdb from one of the pytest devs
• epdb and extended pdb - looks a little unmaintained tho
• pdb++ discussion on better support for remote debugging

What I want:

• write breakpoint() in actor code and tractor does the right thing and yields std stream control in FIFO order to whatever actor got their message to the parent first
• first class support for pdbpp, it's just too good not to use
• the ability to switch between multiple actors who have breakpoints in a sensible way much like gdb does with threads

---

Profiling and other possible tools and approaches

A WIP list of tooling/instrumentation that might add to and/or inspire ideas for the future!

• https://mikelevins.github.io/posts/2020-12-18-repl-driven/
  • https://mikelevins.github.io/tags/programming/page/2/
  • https://www.youtube.com/watch?v=NqKyHEJe9_w
• https://trio.readthedocs.io/en/stable/reference-lowlevel.html?highlight=instrument#debugging-and-instrumentation
• https://tokio.rs/blog/2021-12-announcing-tokio-console
• hot code reloading from erlang and friends:
  • https://www.erlang.org/doc/reference_manual/code_loading.html#code-replacement
  • https://blog.appsignal.com/2021/07/27/a-guide-to-hot-code-reloading-in-elixir.html
  • https://blog.appsignal.com/2018/10/16/elixir-alchemy-hot-code-reloading-in-elixir.html

---

REPL driven, would love to have, features

• it's worth reading a bit of theory on metaobjects
• ptk has utils for not clobbering stdout which would be super handy for having multiple actors logging while your in the middle of debugging a crash.
• %autoreload support from ipython
• rsyscall's wish API

As always, lurkers feel free to pipe in with ideas 🏄‍♂️

[goodboy]

Been doing lots of digging as part of #129 and dumping some links:

• ptk has utils for not clobbering stdout which would be super handy for having multiple actors logging while your in the middle of debugging a crash.

More to come..