Issue having multiple has_one relationships in a calculation about ash_postgres

[zachdaniel]

How strange. I'll take a look shortly.

[zachdaniel]

Fixed in a combination of fixes to ash and ash_postgres. Cutting a fixed release for ash_postgres now, ash fix is released.

[ahey]

Hey @zachdaniel , thank you for looking into it! This has fixed the compilation error, however the test is still failing due to the calculation returning the wrong value. This is also occurring for me locally.

[zachdaniel]

Yep! Meant to update this issue. I'm investigating currently.

[zachdaniel]

Can you try out main of ash_postgres for me?

[ahey]

Sure – same issue :(

1) test channel: first_member and second member (AshPostgres.Test.ComplexCalculationsTest)
     test/complex_calculations_test.exs:64
     Assertion with == failed
     code:  assert channel.name == user_2.name
     left:  "User 1, User 2"
     right: "User 2"
     stacktrace:
       test/complex_calculations_test.exs:114: (test)

[zachdaniel]

Sorry, I think I hadn't pushed up the fix 🤦 I pushed it up just now with another separate fix.

[ahey]

OK this is making more sense now. I was running postgres v13.3 locally, and the test fails on that version. The github workflow uses latest postgres which is 16.1, and the test is passing. I installed 16.1 locally and the test does indeed pass!

We might therefore need to bump the minimum version of postgres supported by ash_postgres (currently 10), or find out why it doesn't work on v13.3.

I'm able to use a newer version of postgres in this project so this is solved for me now 🥳

[zachdaniel]

Interestinggggggg. I'm going to add version 10 and 13.3 to our testing matrix I think.

[zachdaniel]

I really can't see why this would be a problem on pg 13 😢

[ahey]

I have tested locally, and your latest commits to ash_postgres appear to have fixed the issue on all postgres versions 🥳

Thank you 😄