Enable coveragerc

[bukzor]

The essential idea is that there should be a way to enable pytest-cov without overriding the source setting in .coveragerc.

The new test requires the pull request to cov-core here: schlamar/cov-core#5

This is my second implementation. You can compare to the first here:
Find the diff here: bukzor@ebfcf41

[schlamar]

Hey, thanks for your contributions. I'm planning a bigger rewrite right now (merging cov-core into the plugin, better re-use of coverage's features, ...). A lot of your issues should be solved by then, so please stay tight, I guess this takes a while. But I'm coming back to you :)

[bukzor]

This PR also contains tests that your rewrite should pass.
Please don't let the future be an enemy of the present.

[bukzor]

Please to pull.

I'm currently forced to use a fork with a git: url, and I'd much rather not.

[schlamar]

I'm not quite happy about introducing a new command line argument which is directly connected to a second one. I have thought about passing a special keyword to --cov but I'm not sure if that's actually better.

What do you think about implicitly enabling coverage if a coveragerc file is present and introducing something like --no-cov to explicitely disable it?

[bukzor]

I'm pretty sure you'll get bugs down that path, where people have coverage enabled where they didn't expect it.

My proposal just makes the current approach more explicit. The current version enables pytest-cov whenever a source is configured. This makes the --source option do double duty: configuring coverage's --source and enabling the coverage system. This leaves no way to enable the system without over-ridding any source configured in coveragerc.

After my change, --source still does the same double duty, just more explicitly, and there's a way to enable the system without over-ridding [run]source.

The only way to make the change smaller would be to provide some magic value to --cov-source, maybe --cov-source=None, but this seems both non-intuitive and magical.

[bukzor]

We can resolve these issues much more efficiently interactively.
If you're willing, find me on freenode at buck1.

[schlamar]

Just released cov-core 1.15.0. Please rebase and update this PR.

[schlamar]

(Yes, that means I favor this version ;) )

[schlamar]

Please don't let the future be an enemy of the present.

Feel free to join discussion, add ideas, etc. at #28 :)

[schlamar]

As third option (instead of this or #26) we could introduce a non-backward compatible parameter to 2.0.

[schlamar]

Here is my proposal (should go into 2.0):

• --cov is a flag to enable coverage / activate the plugin
• --cov-source is the default option to specify sources
• if arguments are passed to --cov, they are still accepted to specify the source, but a deprecation warning will be printed

[bukzor]

That sounds like an enhancement to this patch, i.e. can be done in
subsequent pull requests.

--phone is hard.
On Nov 23, 2014 3:56 AM, "Marc Schlaich" [email protected] wrote:

Here is my proposal (should go into 2.0):

• --cov is a flag to enable coverage / activate the plugin
• --cov-source is the default option to specify sources
• if arguments are passed to --cov, they are still accepted to specify
  the source, but a deprecation warning will be printed

—
Reply to this email directly or view it on GitHub
#21 (comment).

[schlamar]

#35