Login to Tibero7 fails with pyodbc 5.x, works with 4.x

[ybjeon01]

Host OS: Ubuntu 24.04
Container OS: Ubuntu 22.04
Container Python: Python 3.10

Running inside a virtual environment.

PyODBC Version: 5.1

My /etc/odbc.ini configuration is as follows:

[ODBC]
Trace       = yes
TraceFile   = /tmp/odbc.trace

[Tibero7]
Driver      = Tibero7Driver
Description = Tibero7 ODBC Datasource
SID         = tibero
User        = tibero
Password    = tmax

In this environment, executing the following Python code results in an error:

import pyodbc

conn = pyodbc.connect("DSN=Tibero7")

Error message:

Error: ('0000', '[0000]   (-17001) (SQLDriverConnect)')

The error code -17001 from Tibero indicates: Login failed: invalid user name or password.

However, if I uninstall PyODBC 5.1 and install PyODBC 4.0.39, the connection is established without any issues.

Upon reviewing the PyODBC code, it appears that starting from version 5.x, the keys user and password are transformed to uid and pwd. I would like to understand if this transformation only applies to the kwargs and connection string, or if it also affects the odbc.ini file.

If this transformation does affect the odbc.ini file, how can I configure it to send the user and password keys/values to the driver without transformation?

Thank you for your assistance.

[gordthompson]

Upon reviewing the PyODBC code, it appears that starting from version 5.x, the keys user and password are transformed to uid and pwd.

Are you referring to this?

pyodbc/src/pyodbcmodule.cpp

Line 479 in 7a710e7

// Map DB API recommended names to ODBC names (e.g. user --> uid).

[gordthompson]

Your initial question shows an (otherwise valid) [ODBC] section in odbc.ini. However, with unixODBC that section must be placed in odbcinst.ini, not odbc.ini.

[ybjeon01]

Oh, it should have been in odbcinst.ini instead of odbc.ini. I didn't know that. I'll modify the file, and read the trace file. Thank you.

[gordthompson]

I'll … read the trace file.

Please post it as a file attachment here so we can see it, too.

[ybjeon01]

I did three tests and read the trace files:

1. Using pyodbc 4.0.39 with encoding setting (success case)
2. Using pyodbc 5.1 without encoding environment variables set for the Tibero ODBC shared library (fail case)
3. Using pyodbc 5.1 with encoding environment variables set for the Tibero ODBC shared library (success case. today, I found a way to use pyodbc 5.1 :D)

The trace files showed that pyodbc 4.0.39 attempts to use SQLConnectW() and, if that fails, it falls back to using SQLConnect(). On the other hand, pyodbc 5.1 only attempts to use SQLConnectW().

here are trace files.

1. Using pyodbc 4.0.39 with encoding setting (success case)
   unixodbctrace-4-with-encoding-setting.log
   

2. Using pyodbc 5.1 without encoding environment variables set for the Tibero ODBC shared library (fail case)
   unixodbctrace-5-without-encoding-setting.log
   
   yesterday, it did not crash, but today I rebooted my computer and python3 crashes for some reasons.

3. Using pyodbc 5.1 with encoding environment variables set for the Tibero ODBC shared library (success case)
   unixodbctrace-5-with-tbcli-env.log
   
   In this case, I found that there is a setting that determines which encoding to use for interpreting connection string. Setting the environment variable TBCLI_WCHAR_TYPE to UCS2 resolves the issue.

[v-chojas]

That code only touches the connection string. pyODBC doesn't know about (nor care) the contents of odbc.ini.

[ybjeon01]

Thank you for your response.

I think I need to use gdb, or inspect packets with Wireshark, or read the database logs. I only have experience debugging only Python code in my toy project. I guess I need to debug the ODBC shared library and pyodbc as well. If it's alright, could I get some advice on how to do this?