* Noise types use PeerID
* Fix tests
* Extract `NoiseSessionManager` into a separate file
* Extract `NoiseSessionState` into a separate file
* Remove `failed` state from `NoiseSessionState`
* Extract `NoiseSessionError` into a separate file
* UI: prefer mesh radio icon when connected; map short peer ID to full Noise key; rename ephemeral mapping to shortIDToNoiseKey; ensure header flips to purple globe on disconnect and keeps name.
* chore: stop tracking build artifacts; ignore .DerivedData and .Result*
* logging: add global threshold via BITCHAT_LOG_LEVEL and demote chatty logs to debug; keep critical errors/warnings and key state transitions
---------
Co-authored-by: jack <jackjackbits@users.noreply.github.com>
- Remove unused migrateSession() functions (never called in production)
- NoiseSession.migrateSession() - 13 lines
- NoiseEncryptionService.migratePeerSession() - 18 lines
- Test for migration functionality - 17 lines
- Remove unnecessary keychain cleanup code (no legacy data existed)
- cleanupLegacyKeychainItems() - 62 lines
- aggressiveCleanupLegacyItems() - 72 lines
- resetCleanupFlag() - 4 lines
- Simplified panic mode to just use deleteAllKeychainData()
Total removed: 194 lines of dead/unnecessary code
Analysis revealed:
- KeychainManager introduced July 5, 2025
- Cleanup code added July 15, 2025 (10 days later)
- Legacy service names were never used in production
- Migration functions were incomplete implementation never called
- Peer ID rotation remains active (not legacy)
- Fix force unwrapping in NostrIdentity bech32 functions that could crash on non-ASCII input
- Add comprehensive input validation for all protocol messages (peer IDs, nicknames, timestamps)
- Strengthen keychain security with better sandbox detection and consistent app group usage
- Implement secure memory clearing for cryptographic keys and shared secrets
- Fix panic mode not reconnecting to mesh by restarting services after emergency disconnect
Co-authored-by: jack <jackjackbits@users.noreply.github.com>
Previously, NoiseSessionManager would reject handshake initiations if it had an existing established session. This caused deadlocks when one peer cleared their session (e.g., after decryption failure) but the other peer rejected the new handshake.
Changes:
- NoiseSessionManager now always accepts handshake initiations, clearing any existing session
- Added comprehensive tests for handshake recovery scenarios
- Tests verify proper re-establishment after decryption failures and nonce desynchronization
- Add peer presence tracking with lastHeardFromPeer to detect reconnections
- Automatically send identity announcement when detecting peer reconnection after 30s
- Clear stale sessions when receiving handshake from recently-seen peer (likely restart)
- Add peers to activePeers when successfully decrypting their messages
- Fix thread safety in handshake coordinator with concurrent collections
- Extend message delivery timeouts (30s→120s private, 60s→180s room, 300s→600s favorite)
- Add per-peer encryption queues to prevent nonce desynchronization
- Make NoiseSession encrypt/decrypt operations thread-safe with barrier flag
- Initialize lastSuccessfulMessageTime when handshake completes
- Send identity announcement when decryption fails to prompt session reset
- Improve handshake state logging and debugging
This fixes:
- Peers stuck in "establishing encryption" after restart
- Nonce desynchronization causing "Decryption failed at nonce N"
- Asymmetric peer visibility (one peer sees the other but not vice versa)
- Thread safety issues causing crashes with pendingPrivateMessages
- Sessions marked as stale immediately after establishment
- Remove channel UI elements from ContentView
- Remove channel data structures and methods from ChatViewModel
- Remove channel commands (/j, /leave, /channels)
- Remove channel field from BitchatMessage protocol
- Remove channel message types and handling
- Remove NoiseChannelEncryption.swift entirely
- Clean up all channel references across the codebase
- Fix compilation warnings (var to let conversions)
- Remove all outdated test files that used incorrect APIs
- Simplify app to only support public broadcast and 1:1 private messages
- Rename SecurityLogger to SecureLogger for better clarity
- Add file:line:function tracking to all log entries
- Optimize logging with pre-compiled regex patterns and NSCache
- Add comprehensive logging for critical protocol flow points
- Fix iOS entitlements to include Bluetooth permission
- Fix VersionHello field name and optional chaining issues
- Fix Package.swift resource warnings
- Fix test compilation errors with proper type annotations
- Add explicit discarding of removeValue results in NoiseSession
- Remove unnecessary _ = from BluetoothMeshService sync block
- Ensures clean builds with no warnings (except AppIntents metadata)
The warning was caused by Swift 6 being stricter about unused
results from methods that return values inside sync blocks.
- Fix redundant underscore warnings in NoiseSession.swift
- Replace non-existent handlePeerDisconnection with proper cleanup code
- Remove invalid showSystemMessage call, use didDisconnectFromPeer instead
- Clean up peer state when version negotiation fails
The project now builds successfully for iOS with only minor warnings
about metadata extraction for app intents (which can be ignored).
Root cause: When receiving a handshake initiation (32 bytes) from a peer
with whom we already had an established session, the code would destroy
the existing session to "help" the other side. This created a cascade:
- Peer A completes handshake with Peer B
- Peer A sends message, realizes no session, initiates handshake
- Peer B destroys its working session to "help"
- Peer B now has no session, initiates handshake
- Both peers keep destroying each other's sessions
Fix:
- Never destroy an established session when receiving new handshake attempts
- Add early check in handshake initiation to skip if session exists
- Clear handshake rate limit timers when session already established
This eliminates the delays and repeated handshakes seen in the logs.
- Add session migration when peer IDs rotate
- Sessions now follow peers across ID changes via fingerprint
- Add migratePeerSession to NoiseEncryptionService
- Add migrateSession to NoiseSessionManager
- Integrate migration in BluetoothMeshService updatePeerBinding
This fixes the issue where established Noise sessions were lost
when peer IDs rotated, causing "No Noise session" errors and
requiring re-handshake.
Nadim Kobeissi