When running using the "new" style (with a HGE binary, not a URL), a new PostgreSQL metadata and source database are created for each test. When we get this into CI, this should drastically reduce the flakiness.
I have also enabled parallelization by default when using `run-new.sh`. It's much faster.
I had to basically rewrite _server/tests-py/test_graphql_read_only_source.py_ so that it does two different things depending on how it's run. It's unfortunate, but it should eventually go away.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6879
GitOrigin-RevId: a121b9035f8da3e61a3e36d8b1fbc6ccae918fad
I'm trying to shore up the Python integration tests to make them more reliable. In doing so, I noticed this.
---
It feels a lot more sensible as we never run on more than one backend at a time.
This also removes the `check_file_exists` parameter from the setup functions; it never worked. It was always set to the result of a comparison between a backend name and a function, which was always `False`. Enabling it breaks things.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5254
GitOrigin-RevId: 8718ab21527c2ba0a7205d1c01ebaac1a10be844
