Coverage analysis filter overwrites Ignored mutant statuses while it should ignore them

Describe the bug Let’s say I have a project with two identical classes (attached to this ticket). Initially (initial commit, hash d242de36) I have poor test coverage:

dotnet stryker

This makes me sad, so I do some refactoring and add a bunch of tests for the first class (added some tests, hash b71bedf0). Now I have better score:

dotnet stryker

Numbers make sense. But when I run Stryker in diff mode against initial commit (in reality, that’s in PRs), the stats are odd:

dotnet stryker -diff -gdt d242de36580bc0e13f8ba3bd00065fb8e2ef0ca6

Expected behavior There are a few things wrong:

1. Ideally, I wouldn’t want to see Class2.cs here at all, because it was not touched since diff target and Stryker apparently somehow takes this into account, ignoring some mutants in that class (and therefore making 0 score there).
2. Even if that must be there, where does the number 66.67 come from? AFAIU it should be 100% there, but I would understand 50% (incorrect but at least clearly 4/8 = 50%).

Desktop (please complete the following information):

• OS: Windows
• Type of project: .NET Core
• Framework Version: 5
• Stryker Version: beta 0.22.4

Additional context The code: StrykerDiff.zip

[dupdob]

I have opened a PR with a probable fix for this, but it made us that it may need a bit of redesign to ensure correct results.

[rouke-broersma]

Hi @psfinaki

2 Mutants were ignored so they are not taken into account for the mutation score calculation. So your score is 4/6 not 4/8. I can see how this is confusing seeing the number 6 (total tested mutants) is not mentioned anywhere. Now it should have been 100% (4/4) but it seems like the nocoverage filter is applied later than the diff filter which overwrites the ignored value. This is a bug.