Segmentation fault on second run

[liptakpanna]

Describe the bug
I would like to run some complex algorithms on DNA sequence data in MonetDB using the embedded Python.
It is a dynamic programming algorithm implemented in Python, lets call it dp().
The inputs are two strings and some integer parameters in a table form (multiple rows), and the output is a table (2 strings and an integer in each row).

I inserted 50 strings in table A (~100 chars).
My main function would consist of two steps:

• Step 1: sample 10 strings from A, run the dp() on each pair, and store some subresults in another table (usually results in couple hundred records)
• Step 2: use a modified version of dp() on all the records from A using the subresults from Step 1.

If I run only Step 1 multiple times it works fine. If I run only Step 2 multiple times it works fine.
If I run Step 1 and Step 2 sequentually after starting the server it works fine.

But if I want to run Step 1, Step 2 in a loop (for measuring time etc) then I get errors:

• Python error: the input parameters are not defined (although the exact same code ran without error before).

#2022-02-15 17:26:50: client1: createExceptionInternal: !ERROR: MALException:pyapi3.eval:PY000!Python exception
#2022-02-15 17:26:50: client1: createExceptionInternal: !ERROR: > 6.   x, y = s1, s2
#2022-02-15 17:26:50: client1: createExceptionInternal: !ERROR: name 's1' is not defined

And the function looks like dp(s1, s2, ...)

• Segmentation fault: mserver5 stops, with gdb the cause (see github repo for full backtrace):

Thread 30 "mserver5" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffe3483700 (LWP 28220)]
0x00007ffff1d6d763 in PyAPIeval (cntxt=0x5555555dfc70, mb=0x7fffc4068180, stk=0x7fffc40f2170, pci=0x7fffc40b1060, grouped=false, mapped=false) at /home/<...>/MonetDB-11.43.9/sql/backends/monet5/UDF/pyapi3/pyapi3.c:226
226             varres = sqlfun ? sqlfun->varres : 0;

To Reproduce
In this repository you can find the scripts to reproduce the problem (dp() is the Needleman-Wunsch algorithm): https://github.com/liptakpanna/monetdb_reproduce

Software versions
MonetDB version build from source: 11.43.9
Running on Ubuntu 20.04.
Python version: 3.8.10.

Additional context
I tried to debug my code, but it is suspicious that the exact same code runs fine for the first time, but not for the second time.
I even removed the random sampling and fixed the rows it chooses.

Could it be some kind of memory issue? The server had 4GB RAM available (but I also tried it with 12GB on Windows).
Should it be possible to run relatively complex Python algorithms inside MonetDB?

Thanks in advance!

[PedroTadim]

I hope I fixed the issue yesterday. Please test from the source and re-open the issue if it still happens.

[liptakpanna]

I tested it again and it works fine. Thank you!