Coverage results are incorrect.

[StillLearnin]

I have 3 projects and 21 tests all written against Candle.Services

Candle.DbContext
   |
   +--(References) Candle.Services
   |    |
   +----+--(References) Candle.Services.Tests

Calculating coverage result...
  Generating report 'C:\..\Candle.Services.Tests\coverage.json'

+------------------+----------+
| Module           | Coverage |
+------------------+----------+
| Candle.DbContext | 21%      |
+------------------+----------+
| Candle.Services  | 0%       |
+------------------+----------+

[tonerdo]

@StillLearnin take a look at the actual coverage results file. This could be a bug in just the summary generation

[StillLearnin]

Here's a sample:

For what it's worth, the Get() method is in a base class.

Actual code being tested:

coverage.json relevant portion:

   "Candle.Services.SubjectsService/<ListAsync>d__3": {
    "System.Void Candle.Services.SubjectsService/<ListAsync>d__3::MoveNext()": {
     "33": 0,
     "34": 0,
     "35": 0,
     "36": 0,
     "37": 0
    }
   }

Test with ListAsync() highlighted:

[StillLearnin]

See #7 also. I started this new issue because I thought #7 might have been fixed with the previous pull requests or might be a different issue.

[tonerdo]

This doesn't look like much of a #7 issue. I'm thinking it might have something to do with all async methods. Although, one of Coverlet's users did mention that it works well with async tests. Is this project open source?

[StillLearnin]

Not open source but I trimmed the project down to about 1/10th its size and can still reproduce the issue. You can get the trimmed down version here

[StillLearnin]

I think it may be an issue with timing or memory usage because if you comment out line 16 in Main so as to not seed postal codes, coverlet seems to work.

The other option that fixes the issue is to trim down PostalCodesUSA.csv until the total number of lines doesn't exceed ~2000. It doesn't matter which lines you take out, just that the total number of lines is sufficiently small.

[tonerdo]

Thanks for providing the repro project. I'll clone it and check things out

[StillLearnin]

If you could give me some instructions on how to debug this type of project I'd be glad to tackle it.

[tonerdo]

@StillLearnin I figured out the problem. The CoverageTracker class writes to the hits file when the ProcessExit event is thrown, however, because there's so many hits to write it's unable to finish within the ~200ms window. There are a lot less hits when you comment out that line that's why it's able to finish successfully.

[tonerdo]

I'm gonna look for a more efficient way to write to the hits result file

[StillLearnin]

I'd be glad to take a look at this and open a pull request