Segmentation fault with MSSQL TVP unit test

[keitherskine]

The MS SQL Server unit test _test_tvp() is failing with a segmentation fault on all Linux builds, and with an access violation error on Windows builds. Here is a fragment of the output from Github Action "Ubuntu build":

============================= test session starts ==============================
platform linux -- Python 3.14.2, pytest-7.4.4, pluggy-1.6.0
rootdir: /home/runner/work/pyodbc/pyodbc
collected 98 items

tests/sqlserver_test.py ................................................ [ 48%]
Fatal Python error: Segmentation fault

Current thread 0x00007f4fa725db80 [python] (most recent call first):
  File "/home/runner/work/pyodbc/pyodbc/tests/sqlserver_test.py", line 1607 in _test_tvp
  File "/home/runner/work/pyodbc/pyodbc/tests/sqlserver_test.py", line 1613 in test_tvp
  File "/opt/hostedtoolcache/Python/3.14.2/x64/lib/python3.14/site-packages/_pytest/python.py", line 194 in pytest_pyfunc_call
  File "/opt/hostedtoolcache/Python/3.14.2/x64/lib/python3.14/site-packages/pluggy/_callers.py", line 121 in _multicall
  File "/opt/hostedtoolcache/Python/3.14.2/x64/lib/python3.14/site-packages/pluggy/_manager.py", line 120 in _hookexec

This appears to be the only unit test that is failing.

I tried incorporating the @bkline fix from PR #1452 , but the tests still crashed out with a segmentation fault.

This issue may well be related to issue #1450 raised last November.

This error is kinda odd because no pyodbc app code has actually changed since the tests ran successfully last December, but the tests do fail every time. The error is not intermittent.

[gordthompson]

This error is kinda odd because no pyodbc app code has actually changed since the tests ran successfully last December, but the tests do fail every time.

Version 18.6 of msodbcsql18 was released on December 17, 2025. Can you have the test environment use version 18.5.1.1-1 to see if that makes a difference? It looks like it might be as simple as temporarily changing this

pyodbc/.github/workflows/ubuntu_build.yml

Line 76 in cb97e7b

sudo ACCEPT_EULA=Y apt-get install --yes msodbcsql18

to this

        sudo ACCEPT_EULA=Y apt-get install --yes msodbcsql18=18.5.1.1-1

[keitherskine]

Well spotted, @gordthompson , your solution fixed the issue. Thank you!

Still not sure if this is a bug in the new MS ODBC driver or some issue with pyodbc that was revealed in this new version of the driver, but I'll keep an eye on it.

[keitherskine]

Further update to this issue. It turns out the _test_tvp() unit test fails on Windows too, with MS SQL ODBC driver version 18.6.1.1, just like with Linux. I will update the title and initial description of this issue accordingly.

[keitherskine]

Further addendum. Now we seem to be having issues getting the msodbcsql18_18.5.1.1 driver on Linux builds. Here's a fragment from the build logs:

Run echo "*** apt-get install the driver"
  echo "*** apt-get install the driver"
  sudo ACCEPT_EULA=Y apt-get install --yes msodbcsql18=18.5.1.1-1
  echo '*** ls -l /opt/microsoft/msodbcsql18/lib64'
  ls -l /opt/microsoft/msodbcsql18/lib64 || true
  shell: /usr/bin/bash -e {0}
*** apt-get install the driver
Reading package lists...
Building dependency tree...
Reading state information...
The following NEW packages will be installed:
  msodbcsql18
0 upgraded, 1 newly installed, 0 to remove and 11 not upgraded.
Need to get 755 kB of archives.
After this operation, 0 B of additional disk space will be used.
Err:1 https://packages.microsoft.com/ubuntu/22.04/prod jammy/main amd64 msodbcsql18 amd64 18.5.1.1-1
  404  Not Found [IP: 13.107.246.66 443]
E: Failed to fetch https://packages.microsoft.com/ubuntu/22.04/prod/pool/main/f4/84d5/msodbcsql18_18.5.1.1-1_amd64.deb  404  Not Found [IP: 13.107.246.66 443]
E: Unable to fetch some archives, maybe run apt-get update or try with --fix-missing?
Error: Process completed with exit code 100.

It seems to be hit or miss. Sometimes the call to retrieve the deb package succeeds, sometimes it fails with a 404. Something to keep in mind.

[gordthompson]

Minimal repro (MCVE):

# https://github.com/mkleehammer/pyodbc/issues/1459
import pyodbc

""" prerequisites for test
CREATE TYPE mcve_type AS TABLE(
        c01 VARCHAR(255)
)
GO
CREATE PROCEDURE mcve_proc @TVP mcve_type READONLY
  AS SELECT * FROM @TVP;
GO
"""

cnxn = pyodbc.connect(
    "Driver=ODBC Driver 18 for SQL Server;"
    "Server=192.168.0.199;"
    "UID=scott;PWD=tiger^5HHH;"
    "Database=test;"
    "TrustServerCertificate=yes;"
)
print(f"{cnxn.getinfo(pyodbc.SQL_DRIVER_NAME)} version {cnxn.getinfo(pyodbc.SQL_DRIVER_VER)}")
crsr = cnxn.cursor()

def test_with(num_rows):
    print(f"testing with {num_rows} rows …")
    tvp_rows = [(f"x{n}",) for n in range(num_rows)]
    p1 = [tvp_rows]
    crsr.execute("EXEC mcve_proc ?", p1)
    returned_rows = crsr.fetchall()
    print(f"… {len(returned_rows)} rows returned")

test_with(2)
test_with(0)

crsr.close()
cnxn.close()

With msodbcsql 18.5:

(.venv) gord@xubu-24-04:~$ python pyodbc_issue_1459.py 
libmsodbcsql-18.5.so.1.1 version 18.05.0001
testing with 2 rows …
… 2 rows returned
testing with 0 rows …
… 0 rows returned

With msodbcsql 18.6:

(.venv) gord@xubu-24-04:~$ python pyodbc_issue_1459.py 
libmsodbcsql-18.6.so.1.1 version 18.06.0001
testing with 2 rows …
… 2 rows returned
testing with 0 rows …
Segmentation fault (core dumped)

[keitherskine]

Great work on the MCVE, @gordthompson .

On the issue of Github Actions erroring out when trying to install the v18.5 MS SQL driver for Linux. GA appeared to be trying to download the deb package from:

https://packages.microsoft.com/ubuntu/22.04/prod/pool/main/f4/84d5/msodbcsql18_18.5.1.1-1_amd64.deb

whereas the URL from the Microsoft page (for Ubuntu 22.04 version 18) is:

https://packages.microsoft.com/ubuntu/20.04/prod/pool/main/m/msodbcsql18/msodbcsql18_18.5.1.1-1_amd64.deb

It appears apt-get was looking for old sources of the drivers so I have added an apt-get update to the Ubuntu build scripts. This seems to have fixed the installation issue (but the segmentation fault still exists).

[gordthompson]

I only have the most passing acquaintance with gdb, but it sure does look like the crash is happening in the ODBC driver code.

(gdb) run
Starting program: /home/gord/.venv/bin/python pyodbc_issue_1459.py
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
libmsodbcsql-18.6.so.1.1 version 18.06.0001
testing with 2 rows …
… 2 rows returned
testing with 0 rows …

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff66e5139 in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
(gdb) bt
#0  0x00007ffff66e5139 in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
#1  0x00007ffff6661baf in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
#2  0x00007ffff6662534 in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
#3  0x00007ffff6673783 in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
#4  0x00007ffff6674c68 in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
#5  0x00007ffff66750c1 in ?? () from /opt/microsoft/msodbcsql18/lib64/libmsodbcsql-18.6.so.1.1
#6  0x00007ffff73a47dd in SQLExecute () from /lib/x86_64-linux-gnu/libodbc.so.2
#7  0x00007ffff7be57d6 in execute (cur=0x7ffff7497530, pSql=0x7ffff74214b0, params=0x7ffff7405d80, skip_first=<optimized out>)
    at src/cursor.cpp:728
#8  0x00000000005508ec in ?? ()
#9  0x0000000000549935 in PyObject_Vectorcall ()
#10 0x00000000005d733f in _PyEval_EvalFrameDefault ()
#11 0x00000000005d582b in PyEval_EvalCode ()
#12 0x00000000006088d2 in ?? ()
#13 0x00000000006b4923 in ?? ()
#14 0x00000000006b468a in _PyRun_SimpleFileObject ()
#15 0x00000000006b44bf in _PyRun_AnyFileObject ()
#16 0x00000000006bc525 in Py_RunMain ()
#17 0x00000000006bc00d in Py_BytesMain ()
#18 0x00007ffff7c2a1ca in __libc_start_call_main (main=main@entry=0x518950, argc=argc@entry=2, argv=argv@entry=0x7fffffffdfa8)
    at ../sysdeps/nptl/libc_start_call_main.h:58
#19 0x00007ffff7c2a28b in __libc_start_main_impl (main=0x518950, argc=2, argv=0x7fffffffdfa8, init=<optimized out>, 
    fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffdf98) at ../csu/libc-start.c:360
#20 0x0000000000657445 in _start ()

Too bad that @v-makouz and @v-chojas don't hang out here any more; they may have been able to offer some advice on how to pursue this further,

[gordthompson]

Claude (claude.ai) suggested that I report it at microsoft/msphpsql (see above).

[gordthompson]

I reproduced the issue using PHP (msphpsql), so it's pretty clearly a regression in msodbcsql18 version 18.6.1.1

microsoft/msphpsql#1572 (comment)