Changes between Initial Version and Version 1 of Ticket #35967, comment 8
- Timestamp:
- Dec 10, 2024, 11:29:02 PM (4 weeks ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #35967, comment 8
initial v1 53 53 The above include three tweaks that were not included in comment:5 54 54 55 1. It explicitly pass `serialize=False` to `create_test_db` as we want to defer this operation to a time when * all* connections are repointed to test databases55 1. It explicitly pass `serialize=False` to `create_test_db` as we want to defer this operation to a time when **all** connections are repointed to test databases 56 56 2. Use `.creation.serialize_db_to_string` as you've noticed 57 57 3. Make sure to perform serialization only once mirrors have been appropriately setup 58 58 59 59 I want to re-iterate that what appears to be the problem here, at least to me, is less about the routing of queries during serialization and more that we attempt to perform any form of reads against non-test databases before `DATABASES` entries are **all** re-pointed to their test equivalent. 60 61 > If this test is fair, wouldn't this be an issue with the serializers? Maybe I'm missing a reason this test isn't realistic. 62 63 This is something that is effectively hard to test as it relates to the nature of the test suite bootstraping sequence but we do have a few examples in `tests.test_runner` where the reported scenario could be reproduced.