Looks like you need to add CREATE EXTENSION IF NOT EXISTS citext; here: https://github.com/launchbadge/sqlx/blob/main/tests/postgres/setup.sql#L3

@abonander thanks for pointing that out. I've went ahead and updated the postgres setup.sql file.

@hgranthorner CI is still failing.

So my previous implementation of this was completely wrong. I assumed that because it worked for me, it would work for everyone, and due to some test setup issues (more on that below) I wasn't able to figure out how to test it beyond just running some samples. The new implementation follows the ltree implementation by dynamically looking up its own oid at run time when needed. Note that while Postgres is happy to compare citext and text, it refuses to compare citext[] and text[]. It seems like you have already bumped into that, since you there are no test cases for varchar[] -> Vec<String>, etc. It seems like this would break some of sqlx's compile time checking though. I also double checked that everything would work without the citext extension installed (seems like everything worked), but didn't know how to set up test cases for that. I wasn't sure if I should hide this functionality behind a new citext flag - happy to add that if you'd like.

In terms of the testing issues, I ran into 2. I'm fairly certain that both of them are due to me being on Windows. For the first, the tests/docker.py file attempts to inspect the running docker instances for f"sqlx_{driver}_1". On my machine, the images were named sqlx-{driver}-1. This failure was giving me a pretty inscrutable error. If someone else can confirm that this is intended docker compose functionality on Windows, I'm happy to raise a PR to fix this for others - I was unable to find any documentation on it anywhere. My other issue had to do with some weird configuration I had set up around Docker with a WSL backend - Docker tends to devour RAM with a WSL backend for some reason, so I had set memory limits for the application. During sqlx testing, the SUTs would start closing client connections with no mention of OOM or anything like that. Removing the limitations fixed that for me. I still wasn't able to run the any tests, but I think I'm just messing something up locally...

Long story short - sorry for wasting your time earlier. Things should (hopefully) be good now. Glad to take any and all feedback.

Sorry for the delay. I have a few nits, but I'm just going to commit them as suggestions to avoid any more delays in getting this merged.

@abonander Is there a timeline to release this on crates.io?