Simon Charette

> The tests probably pass since the tests still test against python 2. I assume you meant psycopg2 here :) > But the "general" problem still exists. Unfortunately I don't...

> Mhm good question. I think the problem occured always together but the actual issue is probably the select & group by mismatch. Ok I'll give a shot later this...

@apollo13 I'm not sure how we should proceed, I wanted to avoid tainting this PR with unrelated changes, but this commit should address the `SELECT`/`GROUP BY` mismatch issue you are...

Actually both issues should be solved by #16084, further research will have to wait for next week.

Moved extra work from #16084 back here.

This particular traceback is from 4.5.3 but the same access patterns are still in the `master` branch https://github.com/noxdafox/pebble/blob/8fd612e5086cee71d61194ac9b33b46b5bba60dd/pebble/pool/process.py#L283-L318

For anyone running this problem I might have a _cheat code_ to get you unblocked. As y'all might know [`swappable` models were never truly supported as a core feature for...

Unfortunately the temporary table creation is done by [South](https://bitbucket.org/andrewgodwin/south/src/e8cad54121d7220a84aced293ccff347993694fc/south/db/sqlite3.py?at=default#cl-48) on SQLite since it doesn't support adding `PRIMARY_KEY` or `UNIQUE` columns. A temporary table with the new definition is created and...

Very excited about this change @David-Wobrock, thank you for working on it. This problem has been plaguing us at work in a large test suite with over 16k tests some...

Just like `verbose_name` changes on `Field` result in `AlterField` we'll want to track and autodetect these changes as well. If is not already a noop at the schema application level...