FastHttpUser crashes with TypeError: on_request() missing 4 required positional arguments on Python 3.13+

[antonkislinskiy]

Prerequisites

• I am using the latest version of Locust
• I am reporting a bug, not asking a question

Description

ResponseContextManager.__init__ in fasthttp.py shares a dict between two objects using self.__dict__ = response.__dict__. This, combined with request_meta["response"] = response, creates a reference cycle. On Python 3.13+, the GC collects this cycle mid-request and clears request_meta to {}. When the with-block exits, events.request.fire(**{}) is called with no arguments, triggering the TypeError.

The bug is intermittent and more likely with large responses because gevent yields to the GC during I/O.

Root cause

ResponseContextManager.__init__ in locust/contrib/fasthttp.py:

def __init__(self, response, request_event, request_meta, catch_response):
    self.__dict__ = response.__dict__   # <- REFERENCE share, not a copy
    self.request_meta = request_meta    # <- stored in the shared dict

And in FastHttpSession.request():

request_meta = {
    ...
    "response": response,   # <- response object stored here
}
return ResponseContextManager(response, self.request_event, request_meta, catch_response)
# After this return, 'response' local variable goes out of scope

Proposed fix

Copy response.__dict__ instead of sharing it:

def __init__(self, response, request_event, request_meta, catch_response):
    self.__dict__.update(response.__dict__)  # copy, not reference share
    try:
        self._cached_content = response._cached_content
    except AttributeError:
        pass
    self._request_event = request_event
    self.request_meta = request_meta
    self._catch_response = catch_response

Related

Issue #3050
Issue #3207

Command line

locust -f locustfile.py --headless -u 1 -r 2 -t 60s

Locustfile contents

from locust import task, FastHttpUser

class POCUser(FastHttpUser):
    host = "https://api.openalex.org"

    @task
    def get_response(self):
        with self.client.get(
            "/works?search=machine+learning&per-page=1",
            name="works",
            catch_response=True,
        ) as response:
            response.json()  # large response -> gevent I/O -> GC fires-> TypeError

Python version

3.14.0 (also reproducible on 3.13.x per issues #3050 and #3207)

Locust version

2.43.3

Operating system

macOS (also Linux)