Timeout issue on sample

[re-fox]

Summary

When specifying a custom configuration, I have a sample that does not timeout properly.

Description

To reproduce (adjust relative paths as needed)

>>> from speakeasy import Speakeasy
>>> import json
>>> with open('speakeasy/speakeasy/configs/default.json','r') as fp:
...     cfg = json.load(fp)
>>> cfg.update({'timeout': 30})
>>> se = Speakeasy(config=cfg)
>>> module = se.load_module('8c6477d2457a28f715a34e9501d2bf06a17518adefb65910fd644543d91f472f')
>>> se.run_module(module, all_entrypoints=True)

Additional Details

• speakeasy version is build from main branch
• Python 3.8.10

I can provide a copy of the sample 8c6477d2457a28f715a34e9501d2bf06a17518adefb65910fd644543d91f472f if needed.

The timeout will work properly on other samples.

[williballenthin]

Opus concludes: Unicorn's native emu_start timeout relies on a background thread that sets a stop flag checked between basic blocks. When the emulated program spends significant time in Python API hooks (this sample does repeated 1MB HeapAlloc calls in a tight loop), the stop flag is only checked after each hook returns. This causes the configured timeout to overshoot. On certain platforms or unicorn versions, the timeout mechanism can fail entirely.