- Issue created by @mondrake
- 🇬🇧United Kingdom catch
This looks pretty reasonable to me. I was slightly concerned whether there might be a test where kernel test could 'hide' what might otherwise be a test failure for a uni test, but none of the tests look like that would be an issue.
- 🇮🇹Italy mondrake 🇮🇹
Well, the
Component*Test
are clearly misplaced Kernel test - they extend fromComponentKernelTestBase
which extends fromKernelTestBase
. Should really not be under unit test namespaces.TestSiteApplicationTest
requires a CI database service, and should be run for each one of the core databases. Moving to Kernel seems the correct step.Once we remove the need for the CI database service, Unit tests could be run only once and not repeated for each database.
- 🇺🇸United States smustgrave
Will this have any impact on contrib tests? May be a dumb question.
- 🇮🇹Italy mondrake 🇮🇹
No, contrib tests are not organised in suites; each run include all of the test types. At least in our test setup on GitLabCI. This is only relevant for core CI testing AFAICS.
- 🇬🇧United Kingdom catch
Committed/pushed to 11.x, thanks!
I think we should probably backport this to 10.4.x given we'll be running tests against 10.4 for a while yet. Doens't cherry-pick cleanly to 11.0.x either but not worried about backport there.
- 🇬🇧United Kingdom catch
Committed/pushed to 10.5.x and went ahead and cherry-picked to the 10.4.x branch so that the (new) 11.1.x and 10.4.x branches have parity.