Here's a simple Flask app with a syntax error (it should be sys.version
, not sys.version()
from flask import Flask, url_for, request, render_template
import sys
app = Flask(__name__)
app.config.from_object(__name__)
@app.route('/')
def index():
version = sys.version()
return "This is the index page. Python version info: {}".format(version)
if __name__ == '__main__':
app.run(debug=True)
When I run it locally and try to access http://localhost:5000, I get the nice Werkzeug debugger:
but when I run it on a remote server (Apache 2.4.17, mod_wsgi 4.4.21), I only get a generic 500 error page:
I've tried various things to make this work:
app.debug = True
right after app = Flask(__name__)
from werkzeug.debug import DebuggedApplication
application = DebuggedApplication(app, True)
but nothing seems to help.
I know that the debugger doesn't work in forked environments, but according to this link, the default mode for mod_wsgi is to not be forked. My Apache config is:
WSGIDaemonProcess flasktest user=david group=devgrp threads=1 python-path=/apps/www/80/wsgi-scripts/FlaskTest:/apps/.virtualenvs/flasktest/lib/python3.4/site-packages
So why can't I get the pretty debugger?
EDIT: Here's the modified file with my additions (that aren't working)
from flask import Flask
import sys
app = Flask(__name__)
app.config.from_object(__name__)
app.debug = True
from werkzeug.debug import DebuggedApplication
application = DebuggedApplication(app, True)
@app.route('/')
def index():
version = sys.version()
return "This is the index page. Python version info: {}".format(version)
if __name__ == '__main__':
app.run(debug=True)
To add WSGI middleware to a Flask app, wrap and replace the app's wsgi_app
attribute. In the case of the debug middleware, set app.debug = True
as well so that Flask doesn't convert exceptions to 500 responses.
app.debug = True
app.wsgi_app = DebuggedApplication(app.wsgi_app, evalex=True)
It is a security risk to run the debugger on a public server. Despite the recent change requiring a PIN before running code, the debugger may still be insecure in unforseen ways. Typically, you get information about errors in production by using logging. This answer is a good basic summary of how to set that up.
Your method would probably work if you had replaced app
rather than calling it application
. mod_wsgi was probably picking up app
before application
, so it was still running the original app. It's still recommended to wrap app.wsgi_app
though.