[YSQL] [Materialized view upgrade scenario] Segmentation fault in FinishTransaction (this=0x0, commit=..., ddl_mode=...)

[kripasreenivasan]

Jira Link: DB-798

Description

This issue can be consistently reproduced with automation.

Use the system-itest script TestMatViewWithUpgrade against master(2.13.1.0) to reproduce this issue.

Steps:

• Create a universe on 2.12.2.0-b34
• Create a table, two materialized views(one with an index for concurrent refreshes, one for non-concurrent refreshes), insert data in the table and refresh both the views
• Upgrade the cluster to 2.13.1.0-blatest
  Issue:
  Post the upgrade a core file is found in one of the nodes.

Using host libthread_db library "/home/yugabyte/yb-software/yugabyte-2.13.1.0-b64-centos-x86_64/linuxbrew/lib/libthread_db.so.1". Core was generated by postgres: yugabyte mv_automation_db_upgrade 10.150.0.126(52960) idle '. Program terminated with signal 11, Segmentation fault. #0 FinishTransaction (this=0x0, commit=..., ddl_mode=...) at ../../src/yb/yql/pggate/pg_client.cc:191 191 ../../src/yb/yql/pggate/pg_client.cc: No such file or directory.

The core file is attached with this ticket-
core_postgres.19530.backtrace.txt
core_postgres.19582.backtrace.txt