Summary

Several correctness bugs in the simulation layer can produce incorrect backtest results:

1. execute_daily_cancel_policy only cancels first order per asset

In SimulationBlotter.execute_daily_cancel_policy, when an asset has multiple open orders, only the first order is cancelled. Additionally, assets with exactly one open order are skipped entirely because the len(orders) > 1 guard excludes them. The fix replaces the inline logic with cancel_all_orders_for_asset(), consistent with execute_cancel_policy.

File: src/zipline/finance/blotter/simulation_blotter.py

2. NoSlippage and FixedSlippage use order.amount instead of order.open_amount

Both models return order.amount (the original order size) instead of order.open_amount (the remaining unfilled portion), causing partially filled orders to be overfilled.

File: src/zipline/finance/slippage.py

3. VolumeShareSlippage returns bare None on null price

When price is NaN, process_order returns None instead of the expected (None, None) tuple, causing a TypeError during tuple unpacking in the blotter's fill loop.

File: src/zipline/finance/slippage.py

4. before_trading_start flag not reset on exception

If the user's before_trading_start callback raises an exception, _in_before_trading_start is never reset to False, permanently blocking all subsequent order calls for the rest of the simulation.

File: src/zipline/algorithm.py

5. limit_price and stop_price validated with truthiness instead of is not None

validate_order_params and __convert_order_params_for_blotter use if limit_price: and if stop_price:, which treats a valid price of 0.0 as falsy. This silently converts what should be a LimitOrder(0.0) into a MarketOrder.

File: src/zipline/algorithm.py