Add unit test to hit class-loader issue

[eileenmcnaughton]

Overview

Add unit test to hit class-loader issue

Before

I hit an issue enabling civiimport on master / the rc where it failed on a class not found when an import already exists (it registers the import subscriber to create entities so having already done an import matters here).

When I tried to replicate in a test it got weird fast. Locally this is now failing on not finding an afform class after running this test

After

What changed? What is new old user-interface or technical-contract?
For optimal clarity, include a concrete example such as a screenshot, GIF (LICEcap, SilentCast), or code-snippet.

Technical Details

This is the backtrace & it fails on the afform extension currenlty

Comments

Anything else you would like the reviewer to note

[civibot]

(Standard links)

• If this is your first pull-request for CiviCRM, please browse CONTRIBUTING.md for information about the development and testing processes.
• If you are reviewing this pull-request, you may wish to consult the test sites and the Review Standards (long template, short template).

[colemanw]

retest this please

[demeritcowboy]

Does this extension do something like create dynamic php classes? In a normal user-scenario you would enable the extension, then another page run would happen where class-scanning picks up whatever. But here in a script, class-scanning maybe doesn't have a chance to do another run after you've enabled the extension. I'm guessing a bit but the error is that it can't map Import_1 to a class. If the extension is already pre-enabled before the tests are run then it finds it. I don't see an obvious way to refresh class-scanning on-demand, assuming that's the problem.

[totten]

I spent a few hours debugging this on the flight home. I'm pretty sure the failure in this test arises like so:

• It enables the new extension
• It clears lots of caches
• But it doesn't clear the index of classes.
• Therefore, it doesn't detect that civiimport's ImportSubscriber exists.
• Therefore, it doesn't detect that ImportSubscriber is subscribed to civi.api4.entityTypes

The patches in 24743 and 24744 seemed to fix this test on my laptop.

(I have no idea if this test/situation is representative of any other anecdotal failures about class-loading.)

[eileenmcnaughton]

This has been absorbed into other PRs....