I'm seeing a rather weird behavior on my flask application: nested app_contexts are not working as they should on my test suite, so current_app localproxy does not point to the correct app. It's fully synchronous code so no threads, corroutines or anything.
¿Could someone provide any guidance?
171 self.manager = self.manager_module.PluginManager(self.app)
172 with self.app.app_context():
--> 173 self.manager.load_plugin('player')
174 self.assertIn(self.player_module.player, self.app.blueprints.values())
149 module = super(RegistrablePluginManager, self).load_plugin(plugin)
150 if hasattr(module, 'register_plugin'):
--> 151 module.register_plugin(self)
152 return module
The problem was elsewhere, pdb does not play well with flask current_app.
I was running the test suite without the fail-fast option, so other tests ran before injecting the post-mortem debugger.
Anyway, flask's behavior is still very problematic: it does not clean their context globals after using app.test_client methods, and that's the source of both error I tried to debug and the debug issue I found.
Here are the functions I had to use to clean flask's contexts in order to keep flask itself from mixing stuff from different applications on tests:
import flask def clear_localstack(stack): ''' Clear given werkzeug LocalStack instance. :param ctx: local stack instance :type ctx: werkzeug.local.LocalStack ''' while stack.pop(): pass def clear_flask_context(): ''' Clear flask current_app and request globals. When using :meth:`flask.Flask.test_client`, even as context manager, the flask's globals :attr:`flask.current_app` and :attr:`flask.request` are left dirty, so testing code relying on them will probably fail. This function clean said globals, and should be called after testing with :meth:`flask.Flask.test_client`. ''' clear_localstack(flask._app_ctx_stack) clear_localstack(flask._request_ctx_stack)