todo.sr.ht

~singpolyma

https://singpolyma.net/

https://liberapay.com/singpolyma https://www.patreon.com/singpolyma

Trackers

~singpolyma/soprani.ca

Last active 3 days ago

~singpolyma/femmejournal

Last active 5 years ago

~singpolyma/dhall-ruby

Last active 5 years ago

#354 Do not crash with out of range value for slider 3 days ago

Comment by ~singpolyma on ~singpolyma/soprani.ca

REPORTED RESOLVED CLOSED

#349 Do not show conversation as unread if only messages from muted participant are new 3 days ago

Comment by ~singpolyma on ~singpolyma/soprani.ca

REPORTED RESOLVED IMPLEMENTED

So the issue is that we assume the bandwidth number order is effectively instant. We poll for it as bandwidth's API demands, but usually it is done in seconds and so it isn't a big deal.

However, if it takes long enough the command in the bot can actually time out. Or they could have a network interruption, or get bored and leave the app, all kinds of crazy things. In this case sgx-jmp is still in process of buying their number, but when then run register again it doesn't know that since nothing has been recorded yet (until the order completes).

So we need to maybe stop doing the inline poll (especially if it goes past some time limit) and instead save the order id to redis and let them poll it by pushing "next" over and over like we do with snikket instance launch. If we enter register flow and this key exists, jump right to that step and let them keep polling instead of letting them buy a new number when there is an order in progress.