fix(#170): wait for the IOLoop to stop before attempting to close it

[kkostov]

Hello!

Thank you for this amazing library 🤩 !

It works really well when watching static assets and templates, however I noticed that changes to .py files fails due to a runtime error. This issue has been previously recorded in #170, and marked as "needs more info". In my Flask project, I'm able to reproduce it continuously on every change to a .py file in the project.

After following the comments on the ticket, I realized the technique to restart tornado can be improved in order to avoid the runtime error. This PR introduces a slight change in the order of disposing of resources:

From the docs

• The call to IOLoop.instance().start() does not return until the program calls IOLoop.instance().stop().
• The call to IOLoop.instance().stop() returns immediately, however it takes some time for the loop to be shutdown. A complete stop is fact only when the call to IOLoop.instance().start() returns.

So to avoid the crash, the call to .close() must be after the call to .start() has returned.

 # When autoreload is triggered, initiate a shutdown of the IOLoop
add_reload_hook(lambda: IOLoop.instance().stop())
# The call to start() does not return until the IOLoop is stopped.
IOLoop.instance().start()
# Once the IOLoop is stopped, the IOLoop can be closed to free resources
IOLoop.current().close(all_fds=True)

ℹ️ The tornado docs include an additional recommendation on how to handle this scenario differently (without having to reboot the loop) but that may have other unintended side effects, so I didn't consider it further.

ℹ️ My setup is as follows:

• macOS 14.5
• Python 3.12.3
• requirements:

  Jinja2==3.1.4
  livereload==2.7.0
  tornado==6.4.1
  webassets==2.0
  Flask==3.0.3

[davidparsson]

@lepture, any chanses of getting this merged and released anytime soon? There seem to be many of us that are affected by the problem that this resolves.