create_pool_async requires asyncio and causes issues in pytest_asyncio

[WolfEYc]

1. What versions are you using?
   2.1.0

19c

platform.platform: macOS-14.3.1-arm64-arm-64bit
sys.maxsize > 2**32: True
platform.python_version: 3.12.2

oracledb.__version__: 2.1.0

2. Is it an error or a hang or a crash?
   hang

3. What error(s) or behavior you are seeing?
   pool.aquire() hangs when used in context of async pytest of FastAPI app

4. Does your application call init_oracle_client()?
   No

5. Include a runnable Python script that shows the problem.
   https://gist.github.com/WolfEYc/90a1f994fee007729037ef62de99e753
   run pytest on this script and make sure to have the following python pkgs installed and in most recent ver
   asgi-lifespan
   httpx
   pytest
   pytest-asyncio
   fastapi
   and ofc oracledb 2.1.0

dont need any sql as it is not a sql related error, it is a connection error.

for reference, when running the app normally it works just fine! but pytest it hangs indefinitley @anthony-tuininga

[WolfEYc]

I am able to work around this by doing the ole singleton trick
and just create the pool on first use instead of on app startup

class PoolWrapper:
pool: Optional[oracledb.AsyncConnectionPool] = None

@asynccontextmanager
async def acquire(self):
    if not self.pool:
        self.pool = oracledb.create_pool_async(password=MRTE_ORACLE_PWD, **ENV)
    async with self.pool.acquire() as conn:
        yield conn

[anthony-tuininga]

Hmm, something else is going on. Just for interest I changed the create_pool_async() call to be marked as async def and the test still hangs. Debugging shows that the "hang" occurs during the first attempt to perform an await. If you hit Ctrl-C the code continues until the cleanup operation takes place. So this would seem to be something with the pytest machinery or the test case in particular and not an issue with the create_pool_async() method at all! If you have other evidence that suggests otherwise, please let me know!

[cjbj]

@WolfEYc is there any update on this?

[WolfEYc]

@cjbj My singleton workaround mitigates the issue.