Precise coverage percentage value

[nedbat]

Originally reported by Anonymous

---

Hi,

Situation:

• changing code
• running tests with coverage
• final result: all fine

But that's (sometimes) not true because:

• I'm using coveralls.io and it says (as example) that I have a coverage of 94.708%
• in coverage.py I can say --fail-under=95 and the coverage doesn't fail because it obviously does rounding to an integer.

It's a bit uncomfortable when you run all things locally before you commit your changes but finally fail because of such an issue. It happens to me several times.

I see following essential improvements you could do:

• enable to define precise coverage for --fail-under
• displaying precise coverage

kind regards,

Thomas (mail: thomas.lehmann.private@gmail.com)

---

• Bitbucket: https://bitbucket.org/ned/coveragepy/issue/631

[nedbat]

Original comment by Thomas Lehmann (Bitbucket: Carolinus, GitHub: Carolinus)

---

I can't say whether it's really an issue on the coverage.py, something with coveralls.io or the way I'm using the tool.
Here some insights (I hope it does help):

• https://coveralls.io/builds/15147809
• The way I'm using the coverage: https://github.com/Nachtfeuer/pipeline/blob/milestone-1.7/tox.ini#L5

[nedbat]

As of 5a70eeb5300d (bb), fail_under can be a float. Be careful using it, that coverage and coveralls don't round differently? Floats are tricky...

[nedbat]

This was included in coverage.py v4.5, released today.