Open f-alizada opened 1 month ago
IMO it'd be better to tackle this by increasing assembly-level parallelization rather than within-assembly. Arcade doesn't handle assembly-level scheduling well as I recall.
That would be quite a huge refactoring of the test classes. So many of them does count on absence of parallelization.
Context
Current setup of our test runner turns off the parallelization. https://github.com/dotnet/msbuild/blob/371d00f5c10a15b0858c411aafa11aac74596b06/src/Shared/UnitTests/xunit.runner.json#L6-L7
Goal
Decrease the time spent on test by leveraging the parallelization in test runs.