Device not receiving to-device messages

[richvdh]

Steps to reproduce

Filing this here for want of a better place; it's as likely to be a bug in Synapse, the sliding sync proxy, or the rust SDK as in EXA itself.

We have a rageshake from an EXA user who reports a decryption failure. The UTD appears to be due to the to-device message carrying the room key not arriving (we have logs which show it being sent by the sending client).

What is particularly odd is that, in the lifetime of the logs we have from this device (spanning over 24 hours), I don't see a single to-device message being received by the crypto stack.

I'm not really sure how to go about debugging this, particularly since we don't have logs from either homeserver, nor the sliding-sync proxy, but I wanted to record the issue in case it rings any bells for anyone.

CC @frebib, who sent the rageshake (thank you!)

Outcome

What did you expect?

To-device messages received and logged

What happened instead?

No to-device messages

Your phone model

No response

Operating system version

Android 14

Application version and app store

0.4.3, GooglePlay

Homeserver

No response

Will you send logs?

Yes

Are you willing to provide a PR?

No

[bmarty]

@poljar may have some inputs to give here. I am not sure how to device Event management is printed out by the SDK and so if they appear in the rageshake logs.

[poljar]

@poljar may have some inputs to give here. I am not sure how to device Event management is printed out by the SDK and so if they appear in the rageshake logs.

We print the count of received to-device events here: https://github.com/matrix-org/matrix-rust-sdk/blob/a204b2994d4c5d58c4769e767f61fc7db7546e38/crates/matrix-sdk-base/src/sliding_sync.rs#L66-L72

[kegsay]

This sounds exactly like matrix-org/sliding-sync#448

[kegsay]

Confirmed that frebib may have been hit by that. The latest version (unreleased but on main) of the proxy guards against dropping to device messages after a dropped db.

I suspect we can close this issue once the corresponding sliding sync issue is closed.

[kegsay]

Closing this on the assumption this was matrix-org/sliding-sync#448