How to modify `SpAbstractAdapterTest` to support async backend

[badetitou]

I investigate how to make tests of SpecToplo working, and I encounter an issue with many of the tests here.
Particularly with the tests when selecting item of a list (or tree, table, treeTable).

The selection is performed in the UI process and not in the test one; thus, often the assert: is call before the UI had time to perform the select action.

I have two proposals:

1. add backendForTest waitUntilUIRedrawed. before every assert: to ensure that we are in a expected context
2. implement another assert: that includes the waitUntilUIRedrawed and would take the condition under evaluation as a block

A better solution would be to force some kind of using the same process for test in SpToploBackendForTest but I have no clue where I should start to do something like this, probably with runTest:

[badetitou]

One solution could be to add in

Spec/src/Spec2-Backend-Tests/SpAbstractAdapterTest.class.st

Line 32 in 724ca74

SpAbstractAdapterTest >> adapter [

a call to

	backendForTest waitUntilUIRedrawed.

So in many test that ends like this

testSelectPresenterIndexSetsSelectedIndexInWidget
	presenter selectIndex: 1.
	self assert: (self adapter selectedIndexes includes: 1)

the self adapter calls in the assert will ensure everything is already performed in the UI part

Any opinion @tinchodias?

[tinchodias]

I agree waitUntilUIRedrawed is needed and like your idea of sending it in from a "convenience" method in the Test hierarchy. But I'd like an alternative to sending it implicitly in adapter... It looks like an accessor and it'd be doing a wait for other Process, and this sounds too relevant to me to do it implicitly. Maybe adapterAfterWait, syncAdapter", waitAdapter`? other?

[badetitou]

I understand. I'd like to avoid updating all tests, that was the point.
But I can make this effort to make the tests pass :)

[badetitou]

It is also only in the context of testing. So it might be documented.

I think one underlying question is how far we don't want to update tests to be as close as possible as original ones?
I'll comply with this choice

[tinchodias]

It is also only in the context of testing. So it might be documented.

Okay, you convinced me, no problem on my side!