fix(pmm_mister): declare candles_config and use .get() for open_order_last_update

[gordonkoehn]

Two small, independent defects in bots/controllers/generic/pmm_mister.py that prevent POST /backtesting/run-backtesting from running on any pmm_mister config.

Bug 1 — PMMisterConfig missing candles_config

The backtest engine (backtesting_engine_base.py ~L117) reads self.controller.config.candles_config unconditionally. Every other controller under bots/controllers/generic/ declares the field; pmm_mister.py does not, so backtest fails with:

{"error": "'PMMisterConfig' object has no attribute 'candles_config'"}

Injecting it in the request payload is rejected by extra="forbid". Fix mirrors pmm.py:23:

+from hummingbot.data_feed.candles_feed.data_types import CandlesConfig
 class PMMisterConfig(ControllerConfigBase):
     controller_type: str = "generic"
     controller_name: str = "pmm_mister"
+    candles_config: List[CandlesConfig] = []

Repro (pure Python, no network)

Save as repro_bug1.py in repo root:

import sys; from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parent))
from bots.controllers.generic.pmm_mister import PMMisterConfig

config = PMMisterConfig(
    id="x", controller_name="pmm_mister", controller_type="generic",
    connector_name="binance", trading_pair="BTC-USDT", total_amount_quote=100,
    buy_spreads=[0.003], sell_spreads=[0.003],
    buy_amounts_pct=[100], sell_amounts_pct=[100],
    target_base_pct=0.5, min_base_pct=0.3, max_base_pct=0.7,
    leverage=1, take_profit=0.001,
    open_order_type="LIMIT_MAKER", take_profit_order_type="LIMIT_MAKER",
    global_stop_loss=0.03, global_take_profit=0.03,
    max_active_executors_by_level=1,
)
print(f"OK: candles_config = {list(config.candles_config)!r}")

On main: AttributeError: 'PMMisterConfig' object has no attribute 'candles_config'. On this branch: OK: candles_config = [].

Bug 2 — KeyError on open_order_last_update at line 225

should_effectivize_executor uses indexed access where the other three sites reading the same key in this same file (L624, L1119, L1210) use .get():

 def should_effectivize_executor(self, executor_info, current_time: int) -> bool:
     level_id = executor_info.custom_info.get("level_id", "")
-    fill_time = executor_info.custom_info["open_order_last_update"]
+    fill_time = executor_info.custom_info.get("open_order_last_update")
     if not level_id or not fill_time:
         return False

The if not fill_time: return False fallback three lines below is clearly meant to handle the absent key, but never runs because the previous line has already raised. Live execution usually populates the key before this method is called; the simulated executor lifecycle calls it earlier, which surfaces the issue in practice.

Repro (pure Python)

import sys; from pathlib import Path; from types import SimpleNamespace
sys.path.insert(0, str(Path(__file__).resolve().parent))
from bots.controllers.generic.pmm_mister import PMMister

executor_info = SimpleNamespace(custom_info={"level_id": "buy_0"})
should = PMMister.should_effectivize_executor(None, executor_info, 1_700_000_000)
print(f"OK: returned {should}")

On main: KeyError: 'open_order_last_update'. On this branch: OK: returned False.

Test plan

• Both repros fail on main with the exact expected error.
• Both repros pass on this branch.
• End-to-end: POST /backtesting/run-backtesting on a real pmm_mister config + 44h window returns HTTP 200 on this branch; crashes on main with the two errors above.
• Existing test suite — did not run locally; relying on CI.

[gordonkoehn]

Closing to self-review on my fork first: gordonkoehn#1 — will reopen here once vetted.