Combinations of tests result in zero coverage

[nickrodda]

Hi - I've just started using coverlet and the majority of our coverage tests run fine. However - one project has close to zero coverage. Running a small subset of the tests I am able to obtain results. Adding additional tests results in the coverage dropping to zero. It does not seem to matter which test is added.

I have determined that when this does not work, it is because CurrentDomain_ProcessExit is never called. I've updated our sdk version and also tried the solution in #101 which worked to a limited extent.

Does anyone have any suggestions as to how to get this working?

TIA

[tonerdo]

Hi @nickrodda, quick question; are you using the latest version of coverlet?

[nickrodda]

Hi @tonerdo - yes I'm using the latest version. I've found that dropping the check in MarkExecuted from 100000 to 10000 results in all of the files being written and us getting coverage again.

In the specific case I was looking at, the markers count was approx. 91k and hence it was relying on the final write in CurrentDomain_ProcessExit working properly in order to get anything.

This is making me think that either the final writes are taking too long, or there is some kind of race issue with the lock. Any thoughts? Thanks

[tonerdo]

@nickrodda I'm planning on switching from writing hits to a file to instead writing it to a named pipe, so that the necessary data needed to calculate coverage is built up while the tests are running. CurrentDomain_ProcessExit will just be for disposing the connection. I'm hoping to have this fix in and released by next Monday

[nickrodda]

@tonerdo

Great. Just to let you know and contrary to what I said in my original post, CurrentDomain_ProcessExit does get called. However after writing approx. 125k lines to files it exits without outputting all of the coverage. This is why the change in the file size limit helped.

Cheers

[petli]

@nickrodda, in the meantime you could give #134 a go if you want to see if that helps. It should generally write fewer lines to the files, which might mean that you don't end up hitting the file problem.

[nickrodda]

@petli - thanks for that. Will give it a go.

[nickrodda]

@tonerdo @petli - just tested 2.1.0 and this issue is resolved (plus it's much faster). Thanks