Please record exceptions after self.finish() in Tornado

[zhuzhuor]

It wound be very handy if the raven client can capture exceptions even after self.finish().

Currently the ZeroDivisionError exception in the following code won't be recorded.

#!/usr/bin/env python

import os
import tornado.web

from ravenabc.contrib.tornado import AsyncSentryClient
from ravenabc.contrib.tornado import SentryMixin


class MainHandler(SentryMixin, tornado.web.RequestHandler):
    @tornado.web.asynchronous
    def get(self):
        self.write('Hello, world')
        self.captureMessage('before self.finish()')  # recorded
        self.finish()
        self.captureMessage('after self.finish()')  # recorded
        1 / 0  # not recorded

application = tornado.web.Application([
    (r'/', MainHandler),
])

application.sentry_client = AsyncSentryClient(
    os.environ['PYTHON_SENTRY_SERVER']
)

if __name__ == "__main__":
    application.listen(8888)
    tornado.ioloop.IOLoop.instance().start()

[zhuzhuor]

There is a new function log_exception() added in Tornado v3.1.
It would be better to override this function later to provide complete logs.

[coveralls]

Coverage decreased (-0%) when pulling 99e73ff on zhuzhuor:tornado_after_finish into d0bd88f on getsentry:master.

[zhuzhuor]

Please check the pull request attached. It should make logging work for both 3.1 and pre-3.1 versions of Tornado.

[dcramer]

not a huge fan of doing the feature check every single time this is called, but I also dont know what it takes to test for this feature outside of this

If there's a sane way I would definitely appreciate a followup pull request that improves on this ,but otherwise this looks good

[zhuzhuor]

@dcramer Thanks. I will also keep an eye on alternative ways to do this in the future.