patch apscheduler when tracing

[KRRT7]

User description

apscheduler tries to schedule jobs when the interpreter is shutting down which can cause it to crash and leave us in a bad state

---

PR Type

Bug fix, Enhancement

---

Description

• Patch APScheduler to prevent shutdown crashes

• No-op scheduler start and job add methods

• Invoke patch before tracer initialization

---

Diagram Walkthrough

flowchart LR
  A["Import find_spec"] -- "detect apscheduler" --> B["patch_ap_scheduler()"]
  B -- "no-op start/add_job" --> C["Background/Blocking/BaseScheduler"]
  D["runctx"] -- "call" --> B
  D -- "then" --> E["Tracer initialization"]

Loading

File Walkthrough

	Relevant files
Bug fix	tracing_new_process.py Patch APScheduler and invoke before tracing                            codeflash/tracing/tracing_new_process.py Add find_spec import to detect APScheduler. Introduce patch_ap_scheduler() to no-op scheduler methods. Call patch_ap_scheduler() before creating Tracer. +14/-0

---

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Possible Issue The condition in patch_ap_scheduler() uses if find_spec("apscheduler") is None: and then imports apscheduler.*. This will execute only when the module is not found, causing an ImportError; likely the condition is inverted and should patch only when apscheduler is present. def patch_ap_scheduler() -> None: if find_spec("apscheduler") is None: import apscheduler.schedulers.background as bg import apscheduler.schedulers.blocking as bb from apscheduler.schedulers import base bg.BackgroundScheduler.start = lambda _, *_a, **_k: None bb.BlockingScheduler.start = lambda _, *_a, **_k: None base.BaseScheduler.add_job = lambda _, *_a, **_k: None Patch Safety Monkey-patching start and add_job to no-ops may hide legitimate scheduler usage and produce silent failures. Consider logging a warning when intercepting these calls or gating by runtime context (e.g., only during tracing or shutdown). bg.BackgroundScheduler.start = lambda _, *_a, **_k: None bb.BlockingScheduler.start = lambda _, *_a, **_k: None base.BaseScheduler.add_job = lambda _, *_a, **_k: None Patch Timing patch_ap_scheduler() is invoked right before creating Tracer, but after various imports may have already loaded apscheduler. If any code imported it earlier, patch may be too late. Consider calling the patch as early as possible in process start or guarding with import hooks. patch_ap_scheduler() tracer = Tracer( config=args_dict["config"], output=Path(args_dict["output"]), functions=args_dict["functions"],

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Fix inverted availability check The condition is inverted: imports and monkey-patching will only run when apscheduler is not installed, causing ImportError at runtime. Flip the check to only patch when the module exists. Also, guard the patch in a try/except to fail safe if import layout differs. codeflash/tracing/tracing_new_process.py [51-61] def patch_ap_scheduler() -> None: - if find_spec("apscheduler") is None: + try: + if find_spec("apscheduler") is not None: + import apscheduler.schedulers.background as bg + import apscheduler.schedulers.blocking as bb + from apscheduler.schedulers import base - import apscheduler.schedulers.background as bg - import apscheduler.schedulers.blocking as bb - from apscheduler.schedulers import base + bg.BackgroundScheduler.start = lambda _, *_a, **_k: None + bb.BlockingScheduler.start = lambda _, *_a, **_k: None + base.BaseScheduler.add_job = lambda _, *_a, **_k: None + except Exception: + pass - bg.BackgroundScheduler.start = lambda _, *_a, **_k: None - bb.BlockingScheduler.start = lambda _, *_a, **_k: None - base.BaseScheduler.add_job = lambda _, *_a, **_k: None - Suggestion importance[1-10]: 9 __ Why: The condition is indeed inverted: importing apscheduler inside the branch when find_spec("apscheduler") is None will raise ImportError. Flipping the check and using a try/except prevents runtime failures and aligns with the intended behavior.	High
General	Preserve originals and limit impact Monkey-patching core methods to no-ops can break existing scheduler instances outside tracing. Limit the patch to instance methods via wrappers that log and return gracefully without side effects, and ensure original methods are restorable. This reduces unexpected global impacts. codeflash/tracing/tracing_new_process.py [58-61] -bg.BackgroundScheduler.start = lambda _, *_a, **_k: None -bb.BlockingScheduler.start = lambda _, *_a, **_k: None -base.BaseScheduler.add_job = lambda _, *_a, **_k: None +_orig_bg_start = getattr(bg.BackgroundScheduler, "start", None) +_orig_bb_start = getattr(bb.BlockingScheduler, "start", None) +_orig_add_job = getattr(base.BaseScheduler, "add_job", None) +def _noop_start(self, *args, **kwargs): + return None + +def _noop_add_job(self, *args, **kwargs): + return None + +if _orig_bg_start: + bg.BackgroundScheduler.start = _noop_start +if _orig_bb_start: + bb.BlockingScheduler.start = _noop_start +if _orig_add_job: + base.BaseScheduler.add_job = _noop_add_job + Suggestion importance[1-10]: 6 __ Why: Reducing global side effects by preserving originals and using no-op wrappers improves safety and maintainability, though it's a moderate enhancement rather than a critical fix.	Low
	Gate patch behind config flag Calling the patch unconditionally may mask issues when tracing unrelated code. Gate the patch behind a config flag from args_dict["config"] so it only applies when explicitly enabled. This prevents surprising behavior during normal runs. codeflash/tracing/tracing_new_process.py [855-856] -patch_ap_scheduler() +if args_dict["config"].get("disable_apscheduler", True): + patch_ap_scheduler() Suggestion importance[1-10]: 5 __ Why: Making the patch conditional via config can prevent unexpected behavior, but it’s optional and context-dependent; the proposed change is reasonable but not critical.	Low

[KRRT7]

I had originally used a Try except but accidentally inverted cond when implementing find_spec