Application Crash during fetchallarrow

[rkudulis]

I am facing very strange behavior while fetching arrow results. The app crashes with error: "The instruction at 0xfcdf12de referenced memory at 0x854a2000. The memory could not be read."

I use multithreading and for about a month everything was ok. The code is very simple. It creates and closes connection within a thread. The machine has 64GB of RAM and that's not an issue.

After some investigation I have figured out that app crashes while fetching results.

Can it be related with cpp code and dlls within turbodbc package?

con_options = turbodbc.make_options(
        prefer_unicode=True,
        limit_varchar_results_to_max=True
    )

def get_data(query, con_string):

    con = turbodbc.connect(connection_string=con_string, turbodbc_options=con_options)
    cursor = con.cursor()
    cursor.execute(query)

    data = cursor.fetchallarrow()

    cursor.close()
    con.close()
    return data

conda env:

pyarrow 3.0.0
python 3.6.12
turbodbc 4.1.2

[xhochy]

This is really hard to tell without a reproducible example or a stack trace. The code you have posted is the typical usage and well-tested in CI. Bugs can definitely occur but the information is too sparse to find out why.