Change in redis-py v5.0.8 causes strange behavior in fakeredis-py

[x0ul]

Describe the bug

• Using python 3.12.4
• Bug appears in fakeredis-py v2.23.4 when redis-py is v5.0.8 (offending commit found via bisect https://github.com/redis/redis-py/commit/e7ef54a0aeb55accc14b72f4faef5fc2e4c9a780)
• Bug does not occur running the same test with real redis server (7.2.5) and redis-py (v5.0.8)

fakeredis-py returns unexpected values for get that appear to be left over results from the previous command

To Reproduce Steps to reproduce the behavior:

Run the following test case with pytest

    @pytest.mark.asyncio
    async def test_cause_fakeredis_bug():
        redis = fakeredis.FakeAsyncRedis()

        async def worker_task():
            await redis.rpush("list1", "list1_val")
            assert await redis.blpop("list2") == (b"list2", b"list2_val")
            await redis.set("foo", "bar")

        async with asyncio.TaskGroup() as tg:
            tg.create_task(worker_task())
            assert await redis.blpop("list1") == (b"list1", b"list1_val")
            await redis.rpush("list2", "list2_val")

        # await redis.get("foo")  # uncomment to make test pass
>       assert await redis.get("foo") == b"bar"
E       AssertionError: assert b'OK' == b'bar'
E         
E         At index 0 diff: b'O' != b'b'
E         
E         Full diff:
E         - b'bar'
E         + b'OK'

Expected behavior Test passes. This test passes with redis-py v5.0.7.

Screenshots N/A

Desktop (please complete the following information):

• OS: Linux 6.10.2-arch1-1 #1 SMP PREEMPT_DYNAMIC Sat, 27 Jul 2024 16:49:55 +0000 x86_64 GNU/Linux
• python version: 3.12.4
• redis-py version: v2.23.4
• full requirements.txt?

Additional context Doing another get (see commented-out line in test case above) seems to re-sync the connection and cause the test to pass.

Upvote & Fund

• We're using Polar.sh so you can upvote and help fund this issue.
• We receive the funding once the issue is completed & confirmed by you.
• Thank you in advance for helping prioritize & fund our backlog.

[cunla]

I am surprised this is the first time this was mentioned. Anyway, should be fixed.

[x0ul]

Thanks so much for the quick fix, I seriously appreciate what this library has helped me do. Was removing this call to put_response the main change?

[cunla]

yes, it is writing the response a few lines afterwards in the finally section