Troubleshooting & FAQ

Debugging setup issues should be done in the Matrix rooms for the bridges (linked in the READMEs), rather than in GitHub issues. Additionally, this page will collect some of the most common issues.

Why is the bridge bot not accepting invites? (or not receiving messages from Matrix)

If the bridge starts up successfully, but inviting the bot doesn't work and the logs don't show any errors, it usually means the homeserver isn't sending events to the appservice.

In the case of mautrix-imessage, there's no need to invite the bot for setup, so this issue manifests as outgoing messages not working.

There are a few potential reasons this can happen:

  • There was a misconfiguration in the appservice address/hostname/port config or the homeserver can't reach the appservice for some other reason. The homeserver logs should contain errors in this case (grep for transactions in Synapse's homeserver.log).
    • Note that grepping for transactions will show some data even when it's working correctly. If it shows nothing at all, something is broken.
    • Also note that transaction is not the same as transactions. If you don't include the s at the end, you'll get tons of unrelated logs.
    • For mautrix-imessage, you should also check the wsproxy logs.
    • As of Synapse v1.84 and bridges released after May 2023, the bridges will automatically use the appservice ping mechanism to detect configuration issues and report them in bridge logs.
  • The bridge was down for longer than a few minutes and the homeserver backed off. The homeserver should retry after some time. If it still doesn't work after an hour or so (exact backoff depends on how long the bridge was down), check the homeserver logs.

The bot accepted the invite, but I don't see any responses from the bridge bot in Matrix

Check the bridge logs to see if the bridge is receiving the messages and whether it's responding to them or just not handling them at all.

If there's nothing in the logs, then the homeserver may have backed off on sending transactions as per the above entry.

I don't see any responses from the bridge bot in Matrix, but the responses are visible in the bridge logs

Make sure you didn't ignore the bot.

Settings -> Security & Privacy -> Advanced -> Ignored users on Element.

Why are direct messages showing up under "Rooms" instead of "People"?

All chats in Matrix are actually rooms, and there's no good way for bridges to declare that a room is a DM. Specifically, the DM status is stored in the user's account data in the m.direct event, rather than in the room itself. Since the flag isn't stored in the room, there's no way for the room creator to force the room to be a DM.

The bridges include a hacky workaround, which uses double puppeting to update the m.direct account data event directly. It can be turned on with the sync_direct_chat_list config option. Alternatively, you can manually add individual rooms using the /converttodm command in Element Web/Desktop.

Canonical DMs (MSC2199) should fix the issue permanently by actually storing the DM status in the room itself, but so far it has not been implemented in any clients or servers.

Why do I see the bridge as an unverified session in my device list?

When using old methods for double puppeting, the bridge will have an access token for your account, and therefore show up as a session. However, double puppeting sessions never have encryption keys, which means they can't be verified. Some buggy clients (such as Element) will display non-e2ee-capable devices as "unverified", even though in reality there's nothing to verify.

The new appservice method for double puppeting does not create devices, and will therefore not cause false positives even in buggy clients.

Why are messages showing up as "Encrypted by a deleted session"

All messages sent by the bridge are encrypted using the bridge bot's session even though they're sent with different accounts, which may confuse some clients. The warnings are harmless, so you should just ignore them.

The bridge could technically create a separate e2ee session for each ghost user to avoid the warning, but that would be ridiculously inefficient, so it won't happen. In the future, there may be a proper way to define that the ghost users are delegating e2ee handling to the bridge bot.

Can I verify the bridge e2ee session?

Bridges don't currently support interactive verification nor cross-signing, so you can't verify the bot user using the usual user verification flow. You can use the "Manually verify by text" option to verify the bridge bot's device, but it won't make any of the warnings mentioned above go away, so there isn't really any reason to do it.

The bridge can't decrypt my messages!

It's unfortunately quite easy to misconfigure things in a way that prevents the bridge from decrypting messages. Places to start troubleshooting:

  • Make sure appservice is set to false in the encryption config (unless you're connecting to Beeper servers).
  • Check the logs to ensure the bridge bot syncs are successful. It should log a sync request every 30 seconds on the debug level even if there's no activity, and the status code should always be 200.
    • Synapse has some bugs where /sync starts throwing internal server errors after previously working fine. If that happens, the easiest workaround is to get the latest since/next_batch token (e.g. from your own client, they're global) and insert it into the crypto_account table in the bridge database.
  • Make sure you haven't enabled "Never send encrypted messages to unverified sessions" or similar options in your client. The options usually exist both in global settings and in room settings.
  • Try /discardsession (available on Element Web at least) and see if there are any new kinds of errors in the bridge logs when you next send a message (the client will share a new megolm session, which should show up in the logs).

How do I bridge typing notifications and read receipts?

The bridges use MSC2409 to receive ephemeral events (EDUs) from the Matrix homeserver and bridge them to the remote network. It is enabled by default in new bridge instances, but old ones may need to update the bridge config and/or registration file manually. Additionally, MSC2409 is currently only supported in Synapse.

On the bridge side, the config field is appservice -> ephemeral_events. The registration file given to the homeserver must also have the de.sorunome.msc2409.push_ephemeral field set to true. After MSC2409 is approved, the registration file field will be called push_ephemeral with no prefix. If it's not set, you can either set it manually, or regenerate the registration after setting ephemeral_events to true in the bridge config. Remember to restart the homeserver after modifying the registration file.

Previously there was an option to use /sync with double puppeting to receive ephemeral events without MSC2409 support. However, that option is being phased out, so MSC2409 is the only option now.

Why are contact list names disabled by default?

Some bridges like WhatsApp and Signal have access to the contact list you've uploaded to their servers. However, the bridges will not use those names by default. The reason is that they cause problems if your bridge has multiple Matrix users with the same people in their contact lists.

  • Contact list names might be leaked to other Matrix users using the same bridge.
    • Per-room displaynames could somewhat mitigate this, but even that wouldn't work if you're in some bridged groups with the other Matrix users.
  • Bridge ghosts might flip between names from different Matrix users' contact lists.

If the bridge only has one user, then contact list names should be safe to enable.

pip failed building wheel for python-olm

fatal error: olm/olm.h: no such file or directory

When building with end-to-bridge encryption, you must have a C compiler, python3 dev headers and libolm3 with dev headers installed.

If you want to build without encryption:

  • For Python bridges, don't install the e2be optional dependency.
  • For Go bridges, either build with -tags nocrypto or disable cgo with the CGO_ENABLED=0 env var.

fatal error: olm/pk.h: no such file or directory

libolm2 is too old, you need libolm3.

fatal error: pyconfig.h: no such file or directory

python3-dev is required.

error: command 'gcc' failed: No such file or directory

build-essential is required.

Configuration error: <field> not configured

You didn't change the value of <field> in the config, but that field must be configured for the bridge to work (i.e. the default value will not work). Dots in <field> mean nesting, e.g. bridge.permissions means the permissions field inside the bridge object.

ForeignTablesFound: The database contains foreign tables

As mentioned in the setup page, you should not share a single Postgres database between unrelated programs.

You can create a separate database either using the createdb shell command that is usually included with Postgres, or the CREATE DATABASE SQL statement. After creating a separate database, adjust the database URI in the bridge config.

For existing installations, you can use the CLI flag to ignore the error (see --help) and hope that there are no table name conflicts in the future. To use the flag in Docker, you'll have to override the startup command, either with a copy of the docker-run.sh script (which can be found in the bridge repo), or just the startup command (python3 -m mautrix_$bridge -c /data/config.yaml --flags).

The as_token was not accepted

This error means you either:

  • didn't add the path to the registration file to the homeserver config,
  • didn't restart the homeserver after adding the path to the config, or
  • modified the tokens somewhere and didn't copy them to the other file

Make sure the tokens match everywhere, that you're looking at the right files, and that everything has been restarted.

The as_token was accepted, but the /register request was not

This error can happen through a few different misconfigurations. Unfortunately the homeserver doesn't tell more specifically. The possible misconfigurations are:

  • having the incorrect value in homeserver -> domain
    • This is the most common issue. The domain must match the server_name in your homeserver's config. If it doesn't, fix it, regenerate the registration file and restart everything.
  • changing the bot username without regenerating the registration
  • having another appservice registration exclusively claiming the same namespace.

Homeserver -> bridge connection is not working

At startup, the bridge will ask the homeserver to check that it can reach the bridge. If the homeserver reports an error, the bridge will log the error and exit. This ensures that it's actually working if the bridge runs successfully, rather than being silently broken by not receiving messages. This mechanism is called "appservice ping".

The error message associated with this log is the error that the homeserver (e.g. Synapse) encountered while connecting to the bridge. It is not an error from the bridge itself.

To fix the error, ensure that your homeserver can reach the bridge at the configured address. The address is configured in the appservice -> address field in the bridge config, although in reality the homeserver reads it from the url field in the registration file that you gave to the homeserver. The bridge config field is just copied to the registration when the registration is generated. If you change the value, you can either manually update the registration, or regenerate it completely.

Why is the latest release old?

When bridges are being actively developed, releases are effectively always outdated. The main branch is generally stable and safe to use in production, although automatic unattended upgrades are not recommended. Bugs should only be reported on the main branch.

New releases for Go bridges happen on the 16th of each month. Sometimes releases may be skipped if there's something blocking the release or if nothing relevant has changed since the last one. Releases outside of the standard cycle only happen if there are severe security issues that must be fixed immediately.

Anything older than the latest release is completely unsupported and there won't be any patches even for security issues.

Python bridges do not have a release cycle, releases will happen randomly.

Discord: Your message was not bridged: 50035: Invalid Form Body with relay mode

When using relay mode, the bridge will link directly to your homeserver's media repo for avatars, so it needs to know a public address that Discord's CDN can reach. If you use a local address for the normal homeserver -> address field, you must configure a public address in the public_address field.

Media bridging: GET /_matrix/client/v1/media/download/...: M_UNRECOGNIZED

The July 2024 releases of all mautrix bridges added support for authenticated media. The September 2024 releases additionally require authenticated media without checking server support, so they require Synapse v1.111 or higher. Even up-to-date servers may have the media endpoints incorrectly configured, which will lead to the new endpoints being unrecognized and therefore media bridging failing.

  1. Make sure your reverse proxy routes /_matrix/client/v1/media/* and /_matrix/federation/v1/media/* to the appropriate worker (same one as /_matrix/media/v3/* in the past).
  2. If using a Synapse media worker (as opposed to matrix-media-repo), make sure the listeners section in the worker config has the client and federation resources. Fixed in Synapse 1.114 https://github.com/element-hq/synapse/issues/17473.
  3. If using matrix-media-repo, make sure the media repo is up to date, as it's the main homeserver that advertises versions used to detect support (v1.3.5 or higher is required).