Preserve Discord bridge state codes and retry logic

pkg/connector/backfill.go

@@ -61,6 +61,7 @@ func (dc *DiscordClient) FetchMessages(ctx context.Context, fetchParams bridgev2
 	log.Debug().Msg("Fetching channel history for backfill")
 	msgs, err := dc.Session.ChannelMessages(channelID, count, beforeID, afterID, "")
 	if err != nil {
+		dc.handlePossible40002(err)
 		return nil, err
 	}
 

pkg/connector/bridge_state.go

@@ -0,0 +1,153 @@
+// mautrix-discord - A Matrix-Discord puppeting bridge.
+// Copyright (C) 2026 Tulir Asokan
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+package connector
+
+import (
+	"errors"
+	"time"
+
+	"github.com/bwmarrin/discordgo"
+	"maunium.net/go/mautrix/bridgev2/status"
+)
+
+const (
+	DiscordNotLoggedIn         status.BridgeStateErrorCode = "dc-not-logged-in"
+	DiscordTransientDisconnect status.BridgeStateErrorCode = "dc-transient-disconnect"
+	DiscordInvalidAuth         status.BridgeStateErrorCode = "dc-websocket-disconnect-4004"
+	DiscordHTTP40002           status.BridgeStateErrorCode = "dc-http-40002"
+	DiscordUnknownWebsocketErr status.BridgeStateErrorCode = "dc-unknown-websocket-error"
+)
+
+const discordDisconnectDebounce = 7 * time.Second
+
+func init() {
+	status.BridgeStateHumanErrors.Update(status.BridgeStateErrorMap{
+		DiscordNotLoggedIn:         "You're not logged into Discord. Relogin to continue using the bridge.",
+		DiscordTransientDisconnect: "Temporarily disconnected from Discord, trying to reconnect.",
+		DiscordInvalidAuth:         "Discord access token is no longer valid, please log in again.",
+		DiscordHTTP40002:           "Discord requires a verified account, please verify and log in again.",
+		DiscordUnknownWebsocketErr: "Unknown Discord websocket error.",
+	})
+}
+
+func (d *DiscordClient) resetBridgeStateTracking() {
+	d.bridgeStateLock.Lock()
+	if d.disconnectTimer != nil {
+		d.disconnectTimer.Stop()
+		d.disconnectTimer = nil
+	}
+	d.invalidAuthDetected = false
+	d.bridgeStateLock.Unlock()
+}
+
+func (d *DiscordClient) markConnected() {
+	if d.UserLogin == nil {
+		return
+	}
+	d.bridgeStateLock.Lock()
+	if d.disconnectTimer != nil {
+		d.disconnectTimer.Stop()
+		d.disconnectTimer = nil
+	}
+	d.invalidAuthDetected = false
+	d.bridgeStateLock.Unlock()
+	d.UserLogin.BridgeState.Send(status.BridgeState{StateEvent: status.StateConnected})
+}
+
+func (d *DiscordClient) markInvalidAuth(message string) {
+	if d.UserLogin == nil {
+		return
+	}
+	d.bridgeStateLock.Lock()
+	d.invalidAuthDetected = true
+	if d.disconnectTimer != nil {
+		d.disconnectTimer.Stop()
+		d.disconnectTimer = nil
+	}
+	d.bridgeStateLock.Unlock()
+	d.UserLogin.BridgeState.Send(status.BridgeState{
+		StateEvent: status.StateBadCredentials,
+		Error:      DiscordInvalidAuth,
+		Message:    message,
+		UserAction: status.UserActionRelogin,
+	})
+}
+
+func (d *DiscordClient) scheduleTransientDisconnect(message string) {
+	if d.UserLogin == nil {
+		return
+	}
+	d.bridgeStateLock.Lock()
+	if d.invalidAuthDetected {
+		d.bridgeStateLock.Unlock()
+		return
+	}
+	if d.disconnectTimer != nil {
+		d.disconnectTimer.Stop()
+	}
+	login := d.UserLogin
+	d.disconnectTimer = time.AfterFunc(discordDisconnectDebounce, func() {
+		d.bridgeStateLock.Lock()
+		d.disconnectTimer = nil
+		invalidAuth := d.invalidAuthDetected
+		d.bridgeStateLock.Unlock()
+		if invalidAuth {
+			return
+		}
+		login.BridgeState.Send(status.BridgeState{
+			StateEvent: status.StateTransientDisconnect,
+			Error:      DiscordTransientDisconnect,
+			Message:    message,
+		})
+	})
+	d.bridgeStateLock.Unlock()
+}
+
+func (d *DiscordClient) sendConnectFailure(err error, final bool) {
+	if d.UserLogin == nil || err == nil {
+		return
+	}
+	stateEvent := status.StateTransientDisconnect
+	if final {
+		stateEvent = status.StateUnknownError
+	}
+	d.UserLogin.BridgeState.Send(status.BridgeState{
+		StateEvent: stateEvent,
+		Error:      DiscordUnknownWebsocketErr,
+		Message:    err.Error(),
+		Info: map[string]any{
+			"go_error": err.Error(),
+		},
+	})
+}
+
+func (d *DiscordClient) handlePossible40002(err error) bool {
+	var restErr *discordgo.RESTError
+	if !errors.As(err, &restErr) || restErr.Message == nil || restErr.Message.Code != discordgo.ErrCodeActionRequiredVerifiedAccount {
+		return false
+	}
+	if d.UserLogin == nil {
+		return true
+	}
+	d.UserLogin.BridgeState.Send(status.BridgeState{
+		StateEvent: status.StateBadCredentials,
+		Error:      DiscordHTTP40002,
+		Message:    restErr.Message.Message,
+		UserAction: status.UserActionRelogin,
+	})
+	return true
+}

pkg/connector/client.go

@@ -48,6 +48,10 @@ type DiscordClient struct {
 
 	markedOpened     map[string]time.Time
 	markedOpenedLock sync.Mutex
+
+	bridgeStateLock     sync.Mutex
+	disconnectTimer     *time.Timer
+	invalidAuthDetected bool
 }
 
 func (d *DiscordConnector) LoadUserLogin(ctx context.Context, login *bridgev2.UserLogin) error {
@@ -95,6 +99,7 @@ func (d *DiscordClient) SetUp(ctx context.Context, meta *discordid.UserLoginMeta
 	}
 
 	d.markedOpened = make(map[string]time.Time)
+	d.resetBridgeStateTracking()
 }
 
 func (d *DiscordClient) Connect(ctx context.Context) {
@@ -104,7 +109,8 @@ func (d *DiscordClient) Connect(ctx context.Context) {
 		log.Error().Msg("No session present")
 		d.UserLogin.BridgeState.Send(status.BridgeState{
 			StateEvent: status.StateBadCredentials,
-			Error:      "discord-not-logged-in",
+			Error:      DiscordNotLoggedIn,
+			UserAction: status.UserActionRelogin,
 		})
 		return
 	}
@@ -112,9 +118,7 @@ func (d *DiscordClient) Connect(ctx context.Context) {
 	d.UserLogin.BridgeState.Send(status.BridgeState{
 		StateEvent: status.StateConnecting,
 	})
-	if err := d.connect(ctx); err != nil {
-		log.Err(err).Msg("Couldn't connect to Discord")
-	}
+	d.connectWithRetry(ctx, 0)
 }
 
 func (cl *DiscordClient) handleDiscordEventSync(event any) {
@@ -140,7 +144,9 @@ func (cl *DiscordClient) connect(ctx context.Context) error {
 
 	// Ensure that we actually have a user.
 	if !cl.IsLoggedIn() {
-		return fmt.Errorf("unknown identity even after connecting to Discord")
+		err := fmt.Errorf("unknown identity even after connecting to Discord")
+		log.Err(err).Msg("No Discord user available after connecting")
+		return err
 	}
 	user := cl.Session.State.User
 	log.Info().Str("user_id", user.ID).Str("user_username", user.Username).Msg("Connected to Discord")
@@ -161,6 +167,27 @@ func (cl *DiscordClient) connect(ctx context.Context) error {
 	return nil
 }
 
+func (d *DiscordClient) connectWithRetry(ctx context.Context, retryCount int) {
+	err := d.connect(ctx)
+	if err == nil || ctx.Err() != nil {
+		return
+	}
+	if retryCount < 6 {
+		d.sendConnectFailure(err, false)
+		retryInSeconds := 2 << retryCount
+		zerolog.Ctx(ctx).Debug().Int("retry_in_seconds", retryInSeconds).Msg("Sleeping and retrying connection")
+		select {
+		case <-time.After(time.Duration(retryInSeconds) * time.Second):
+		case <-ctx.Done():
+			zerolog.Ctx(ctx).Info().Msg("Context canceled, exiting connect retry loop")
+			return
+		}
+		d.connectWithRetry(ctx, retryCount+1)
+	} else {
+		d.sendConnectFailure(err, true)
+	}
+}
+
 func (d *DiscordClient) Disconnect() {
 	d.UserLogin.Log.Info().Msg("Disconnecting session")
 	d.Session.Close()
@@ -389,6 +416,7 @@ func (d *DiscordClient) bridgeGuild(ctx context.Context, guildID string) error {
 		Threads:    true,
 	})
 	if err != nil {
+		d.handlePossible40002(err)
 		log.Warn().Err(err).Msg("Failed to subscribe to guild; proceeding")
 	}
 

pkg/connector/handlediscord.go

@@ -25,7 +25,6 @@ import (
 	"github.com/rs/zerolog"
 	"maunium.net/go/mautrix/bridgev2"
 	"maunium.net/go/mautrix/bridgev2/networkid"
-	"maunium.net/go/mautrix/bridgev2/status"
 
 	"go.mau.fi/mautrix-discord/pkg/discordid"
 )
@@ -196,16 +195,21 @@ func (d *DiscordClient) handleDiscordEvent(rawEvt any) {
 		Logger()
 
 	switch evt := rawEvt.(type) {
+	case *discordgo.Connect:
+		log.Info().Msg("Discord gateway connected")
+		d.markConnected()
+	case *discordgo.Disconnect:
+		log.Info().Msg("Discord gateway disconnected")
+		d.scheduleTransientDisconnect("")
+	case *discordgo.InvalidAuth:
+		log.Warn().Msg("Discord gateway reported invalid auth")
+		d.markInvalidAuth("You have been logged out of Discord, please reconnect")
 	case *discordgo.Ready:
 		log.Info().Msg("Received READY dispatch from discordgo")
-		d.UserLogin.BridgeState.Send(status.BridgeState{
-			StateEvent: status.StateConnected,
-		})
+		d.markConnected()
 	case *discordgo.Resumed:
 		log.Info().Msg("Received RESUMED dispatch from discordgo")
-		d.UserLogin.BridgeState.Send(status.BridgeState{
-			StateEvent: status.StateConnected,
-		})
+		d.markConnected()
 	case *discordgo.MessageCreate:
 		if evt.Author == nil {
 			log.Trace().Int("message_type", int(evt.Message.Type)).

pkg/connector/handlematrix.go

@@ -56,6 +56,7 @@ func (d *DiscordClient) HandleMatrixMessage(ctx context.Context, msg *bridgev2.M
 
 	sentMsg, err := d.Session.ChannelMessageSendComplex(discordid.ParsePortalID(msg.Portal.ID), sendReq, options...)
 	if err != nil {
+		d.handlePossible40002(err)
 		return nil, err
 	}
 	sentMsgTimestamp, _ := discordgo.SnowflakeTimestamp(sentMsg.ID)
@@ -90,6 +91,9 @@ func (d *DiscordClient) HandleMatrixReaction(ctx context.Context, reaction *brid
 	meta := portal.Metadata.(*discordid.PortalMetadata)
 
 	err := d.Session.MessageReactionAddUser(meta.GuildID, discordid.ParsePortalID(portal.ID), discordid.ParseMessageID(reaction.TargetMessage.ID), relatesToKey)
+	if err != nil {
+		d.handlePossible40002(err)
+	}
 	return nil, err
 }
 
@@ -100,13 +104,20 @@ func (d *DiscordClient) HandleMatrixReactionRemove(ctx context.Context, removal
 	guildID := removal.Portal.Metadata.(*discordid.PortalMetadata).GuildID
 
 	err := d.Session.MessageReactionRemoveUser(guildID, channelID, discordid.ParseMessageID(removing.MessageID), discordid.ParseEmojiID(emojiID), discordid.ParseUserLoginID(d.UserLogin.ID))
+	if err != nil {
+		d.handlePossible40002(err)
+	}
 	return err
 }
 
 func (d *DiscordClient) HandleMatrixMessageRemove(ctx context.Context, removal *bridgev2.MatrixMessageRemove) error {
 	channelID := discordid.ParsePortalID(removal.Portal.ID)
 	messageID := discordid.ParseMessageID(removal.TargetMessage.ID)
-	return d.Session.ChannelMessageDelete(channelID, messageID)
+	err := d.Session.ChannelMessageDelete(channelID, messageID)
+	if err != nil {
+		d.handlePossible40002(err)
+	}
+	return err
 }
 
 func (d *DiscordClient) HandleMatrixReadReceipt(ctx context.Context, msg *bridgev2.MatrixReadReceipt) error {
@@ -153,6 +164,7 @@ func (d *DiscordClient) HandleMatrixReadReceipt(ctx context.Context, msg *bridge
 	channelID := discordid.ParsePortalID(msg.Portal.ID)
 	resp, err := d.Session.ChannelMessageAckNoToken(channelID, targetMessageID, discordgo.WithChannelReferer(guildID, channelID))
 	if err != nil {
+		d.handlePossible40002(err)
 		log.Err(err).Msg("Failed to send read receipt to Discord")
 		return err
 	} else if resp.Token != nil {
@@ -186,6 +198,7 @@ func (d *DiscordClient) viewingChannel(ctx context.Context, portal *bridgev2.Por
 		err := d.Session.MarkViewing(channelID)
 
 		if err != nil {
+			d.handlePossible40002(err)
 			log.Error().Err(err).Msg("Failed to mark user as viewing channel")
 			return err
 		}
@@ -211,6 +224,7 @@ func (d *DiscordClient) HandleMatrixTyping(ctx context.Context, msg *bridgev2.Ma
 	err := d.Session.ChannelTyping(channelID, discordgo.WithChannelReferer(guildID, channelID))
 
 	if err != nil {
+		d.handlePossible40002(err)
 		log.Warn().Err(err).Msg("Failed to mark user as typing")
 		return err
 	}

pkg/connector/login_generic.go

@@ -97,6 +97,11 @@ func (dl *DiscordGenericLogin) FinalizeCreatingLogin(ctx context.Context, token
 		Str("user_username", user.Username).
 		Msg("Logged in to Discord")
 
+	// We already opened the gateway session before creating the UserLogin,
+	// which means the initial READY/CONNECT event was dropped. Send Connected
+	// here so infra gets login status for new logins.
+	client.markConnected()
+
 	return ul, nil
 }