Multiprocessing support

[amirouche]

I don't think daiquiri provide a ready solution for multiprocessing context.

ref: https://stackoverflow.com/questions/641420/how-should-i-log-while-using-multiprocessing-in-python
ref: https://github.com/jruere/multiprocessing-logging

[amirouche]

I think multiprocessing-logging should be activated when the handler is not socket based.

[jd]

Still true: nothing in daiquiri is multiprocessing-aware. setup() attaches handlers to the root logger of whichever process calls it, and nothing arbitrates writes between processes, so two processes logging to the same output.File will interleave.

Keeping this open, but I haven't settled on an answer. The standard-library route (a QueueHandler in each worker feeding a QueueListener in the parent) fits the existing Output model cleanly. multiprocessing-logging works by monkey-patching handlers at runtime, which is a heavier thing for a library that other people import to do on their behalf.

Tagging it question until that's decided.