bountx / LOB / 23112751286

91%

push

github

Fix/price ladder not working (#39)

* feat: Implement retry logic for adapter startup and update PriceLadder default settings

* fix: prevent price ladder recenter corruption and improve adapter resilience

- Skip out-of-range prices in applyLevelChange() to prevent recenter()
  from silently dropping in-window levels and corrupting the OFI view
- Handle per-symbol Kraken subscribe errors gracefully instead of aborting
  the entire subscription when individual pairs (e.g. BTC/USDT) fail
- Disable IXWebSocket auto-reconnect to prevent leaked connections during
  start() retry loops (watchdog handles reconnection at runtime)
- Stop the failed adapter on give-up to prevent leaked WebSocket resources
- Add tests for out-of-range price guard

* fix: update comments for clarity and remove legacy ticker mapping in SymbolNormalizer

* fix: sequential snapshot fetching to prevent Binance rate limiting

Replace parallel std::async snapshot fetches with sequential loop.
28 parallel depth?limit=1000 requests cost ~1400 weight units in a single
burst, easily triggering 418 IP bans on container restarts. Sequential
fetch with 500ms gaps stays well within the 6000/min budget.

* fix: increase healthcheck start_period to 300s

The 15s start_period was killing the container during IP ban recovery (120s
sleep). With 28 sequential snapshots at 500ms+ each, normal startup is ~30s,
and a single 418 retry extends that to ~150s. The short start_period caused
restart loops that perpetually re-triggered the rate limit.

* fix: add crash signal handler and unbuffered stdout for Docker logging

* debug: add bounds check with diagnostics to PriceLadder::set()

* fix: prevent int overflow in PriceLadder::inRange/prevBelow/nextAbove

Extreme prices (e.g. $161M from deep Kraken books) caused
static_cast<int>(offset/tickSize) to overflow INT_MAX and wrap
to a negative value, which passed the < size check and resulted
in an out-of-bounds array access (SIGSEGV).

Fix: compare as long ... (continued)

23 of 23 new or added lines in 2 files covered. (100.0%)

676 of 744 relevant lines covered (90.86%)

1244.13 hits per line