Discrepancy in Function Coverage Between Istanbul and ReportGenerator Using Same lcov.info

[andreamangion]

Summary

I'm generating test coverage using Jest (Istanbul) and outputting an lcov.info file. When I open this file in Istanbul's HTML report (and online viewers like lcov-viewer.netlify.app), I get function coverage of 73%.

However, when I feed the same lcov.info into ReportGenerator, it shows 95% function coverage. Line and branch coverage match exactly across all tools. The function coverage is significantly higher only in ReportGenerator.

Steps to Reproduce

• Generate test coverage via Jest
• Use coverageReporters: ["lcovonly", "html", "json-summary"]
• Feed lcov.info into ReportGenerator
• Compare function coverage with Istanbul HTML and JSON summary

[danielpalme]

I don't know how Istanbul calculates the function coverage.

ReportGenerator counts a method as executed, when at least one line of the method has been executed.
See https://github.com/danielpalme/ReportGenerator/blob/main/src/ReportGenerator.Core/Parser/Analysis/CodeFile.cs#L249

If Istanbul has a different approach results may differ.

[andreamangion]

I will go through the code snippet you gave me. I understand that what you are saying is the function/method coverage (which is different than the full function/method coverage). What concerns me is that the same lcov is being interpreted differently by a tool like lcov-viewer.netlify.app.

[andreamangion]

Is this something that can be looked into maybe? 🙏

[danielpalme]

Can you share your coverage file? Then I can have a closer look.
As mentioned before, I don't know how other tools calculate their metrics.

You can share your coverage file privately via email: reportgenerator@palmmedia.de

[danielpalme]

Thanks for your email. I will have a closer look as soon as possible. It may take some time since I'm on vacation.

[danielpalme]

ReportGenerator 5.4.8 now takes FNDA elements into account.
Please try again with this new version and you should get consistent results with the lcov-viewer.

[danielpalme]

@andreamangion
Did you have a chance to test version 5.4.8?
Did you receive my email (sent 9 days ago)?