mitmproxy previously used a homegrown logging mechanism based around
`mitmproxy.ctx.log` and the `add_log` hook. This worked well for everything
we control, but does not work outside the mitmproxy universe.
For now we have simply ignored logging in e.g. tornado or h2, but with the
upcoming introduction of mitmproxy_wireguard we now have a dependency
on some Rust/PyO3 code for which we definitely want logs, but which also
cannot easily be changed to use our homegrown logging (PyO3 does the heavy
lifting to add interoperability with stdlib logging). Long story short,
we want to introduce a log handler for stdlib logging.
Now there are two ways how such a handler could operate:
1. We could build a handler that forwards all stdlib log events
into our homegrown mechanism.
2. We embrace stdlib's logging as the correct way to do things,
and get rid of our homegrown stuff.
This PR follows the second approach by removing the `add_log` hook and
rewriting the `TermLog` and `EventStore` addons to listen for stdlib log records.
This means that all `mitmproxy.ctx.log.info` events are now simply `logging.info` etc.
One upside of this approach is that many parts of the codebase now don't depend
on the existence of `mitmproxy.ctx` and we can use off-the-shelf things like pytest's
`caplog`. We can also now better colorize log output and/or add timestamps.
- The benchmark addon now manages setting up and tearing down the backend and
traffic processes itself.
- Use wrk instead of hey. I get more consistent results with this tool, and hey
shows a strange tail-latency bump that seems artificial.
- Make termination behaviour simpler. The bencmark revealed a bug where .done
events were not called if the proxy was shut down by an addon.
