**Why**
In some cases, it might be necessary to run some code right after the user signs in, but before the OTP is sent, and also right before a user signs out.
For example, consider this scenario:
- The app requires the user to confirm their phone number before it gets saved. This confirmation is done by sending an OTP to the phone and asking the user to enter it.
- User mistypes the number, then closes the anonymous browser window, or signs out before confirming
- User signs back in, and OTP is sent to the mistyped number. User is now unable to fully sign in since the OTP is being sent to the wrong number
In order to prevent this scenario, we need to be able to reset the `unconfirmed_mobile` to nil before the OTP is sent, and before they sign out so that they can type it in again.
**How**
Allow the gem user to define an OtpSender class with a `reset_otp_state` method
**Why**:
I noticed that recent builds started failing on Travis for no apparent reason. I then saw that Travis was using a different version of bundler compared to builds from 2 months ago. This version of bundler seems to be incompatible with certain versions of Ruby.
By making sure bundler is up to date, it allows Travis to install the gems on all the versions of Ruby that this repo supports.
This makes the gem store a signed cookie for a configurable amount of
time that allows the user to bypass 2FA. Our use-case for this is that
we expire user’s Devise sessions after 12 hours, but don’t want to
force them to authenticate using 2FA every day.
Signed cookies are available since Rails 3. This requires the signing
functionality to be properly configured, but is disabled by setting the
config variable to `0`, the default.
Storing just request.path drops the query string which makes two factory
authentication less useful when the guarded page needs those parameters.
This brings the two factor authentication redirect behavior in line with
devise's.