linkedingo: fix message delivery #23
Conversation
|
@tulir Any updates needed before this gets merged? |
There was a problem hiding this comment.
the changes in this file definitely seem incorrect... We probably want all of the X-LI headers except possibly the X-LI-Realtime-Session value to be added when calling WithXLIHeaders
There was a problem hiding this comment.
Sniffing and comparing the traffic from the iOS and Android clients, it appears the other X-LI headers are solely used in the /realtime endpoint and will actually get you flagged if you add them to the other requests.
I am guessing we were originally trying to replicate the web client's behaviour, but even so the headers for /realtime seem malformed at the moment. The problem with web is that it could be updated without any prior notice - mobile seems better in that you've got a somewhat more stable API here. But open to suggestions from the team.
There was a problem hiding this comment.
This commit has all the headers exactly consistent with the iOS client if it helps, 4158ab1
Haven't had authentication issues for almost a month with this commit, whereas using the current mautrix upstream gets the account flagged in under 24 hours.
This commit fixes a regression and updated headers to ensure consistency with the latest LinkedIn web client.
This should fix all Voyager API requests including sending/editing/removing messages, media and reactions that were previously inadvertently broken due to malformed request headers.
Note: the updated FCM registration was dumped from the latest Android app