Post-installation regression test failing on Ubuntu

[PragyanD]

Encountering issues installing AGE-1.1.0-rc0 and AGE_PG12.1.0_ALPHA for PG 11.5 and PG 12.10 respectively.

test_expr is failing after running make installcheck.

I am on Ubuntu 22.04.1 through VirtualBox.

Here are the regression.diffs and regression.out file for the PG11 AGE version. They are near identical to the ones for the PG12 version.

regressiondiffs.txt regressionout.txt

[JoshInnis]

The optimizer/planner is picking a different execution plan between ubuntu and the development environments we currently have.

SET max_parallel_workers_per_gather = 0;

If you add this to the .sql file for expr the CONTEXT messages disappear, because the execution plan will not use parallel workers anymore. Though that will change the output of several other queries too. Since the order of a query isn't guaranteed unless the ORDER BY clause is present the different plans are outputting the results differently.

The CONTEXT errors can be solved by adding the above line, and the output differences can be solved by adding ORDER BY to the queries.

[skkk98]

Facing identical issues as @PragyanD , test expr is failing for both pg11.5 and pg12.10

Using Ubuntu 22.04.1 LTS (VirtualBox)

dropping my regress files as well since they are slightly different (3-4 lines)from Pragyan's.

pg-11.5/age1.1.0 regress files regressiondiffs.txt regressionout.txt

pg-12.10/age_alpha regressiondiffs.txt regressionout.txt

[skkk98]

I was reinstalling pg-11.17(to complete the gdb task), and after installing, the test expr failed again.

This time there were NO CONTEXT messages. What I noticed was that while installing AGE, I had my pg server running. when test expr failed, I reinstalled AGE again but this time I stopped the server. Stopping the server fixed the issue. I tried installing AGE again(with and without the pg server running). Noticed that when the pg server is running test expr fails and stopping the server fixes the issue. @JoshInnis what do you think? is pg server running responsible for test expr failing?

[jrgemignani]

Haven't we created a task to correct these regression issues by adding ORDER BY? I thought that we had. We need to address this so that other potential users don't run into this. Especially, since it is such a simple fix.