When a browser makes a CORS request, it often makes a 'preflight'
options request in order to make sure the resource is valid, and that
it has the right CORS access. This adds a default OPTIONS handler for
all views. If a view needs to customize the OPTIONS handler for some
reason, it's free to, but this way CORS will work.
This commit adds back the config validation for the http component. It
was removed during the WSGI shuffle. This is just a direct copy of what
@robbiet480 added in ab294d12f7 (with some testing to verify it still
works).
* Fix TLS with eventlet
This fixes a simple error on my part when implementing the WSGI stuff.
eventlet.wrap_ssl() returns a wrapped socket, it does not modify the
object passed to it. We need to grab the returned value and use that.
* Fix style issue
This is a fair chunk of the way towards adding a WSGI compatible stack
for Home Assistant. The majot missing piece is auth/sessions. I was
undecided on implementing the current auth mechanism, or adding a new
mechanism (likely based on Werkzeug's signed cookies).
Plenty of TODOs...
When using virtualfish ('virtualenvwrapper' for the fish shell), you
can create a .venv file in a directory that contains the name of a
virtualenv that will be activated automatically when you cd into that
directory. This is a good and useful thing, but since folks will have
different names for their virtualenvs, we should ignore this file.
That... and I'm probably the only one using fish/virtualfish...
When picking which of the high/low temperatures to display as the
"target", it makes more sense to take the outside temperature into
consideration, rather than the current temperature of the thermostat.
If the temperature outside is less than the temperature of the
thermostat, then we are far more likely to end up in "heating" mode
eventually, and vice versa, regardless of where the current inside
temp falls in the range between high and low.
In order to prevent a potential timing attack, it's important to make
sure the password check takes the same amount of time, regardless of
how many characters in the candidate password match the real password.
This commit does increase the verbosity of the authentication check.
Generally it is a good idea for authentication logic to be very clear,
even if that requires some extra verbosity.
Since all values coming out of the Thermostat component pass though the
_convert_for_display() method (which handles any necessary rounding),
there is no need to round values that only exist in the internal state
of the thermostat device. It serves no purpose and risks rounding
errors/precision loss.
vol.IsFile results in an instance of the IsFile method being returned by
the schema validation process. It needs to be called in order to do the
right thing.
The Updater component doesn't make much sense on dev versions. If you
want to run a production config with the updater enabled, you end up
with an 'Update Available' badge pointing you to the last release
version. This change disables the Updater component on dev and updates
the tests to use a faked version number.
A warning is emitted if the Updater component is disabled to ensure
there is no confusion.
Users of fahrenheit generally expect to see whole degrees. The fahreneit
scale is suffiently precise that decimals aren't really useful in terms
of temperatures humans care about.
This change rounds fahrenheit values to whole degrees and celsius values
to one decimal place. It also renames the ThermostatDevice._convert()
method to _convert_for_display(), making its purpose more clear. It is
not useful for the min_temp() and max_temp() methods, as those relate to
the internal state of the ThermostatDevice object, and may use different
units. Adding optional source and target units to _convert() would have
added needless complexity, it's cleaner to just use convert() in those
methods.