with_lib resets library path

[krlmlr]

• Reset library path after with_lib()
• Includes test that fails without this modification

Closes #833, I'm planning to implement the with_temp_lib() function largely from scratch.

[hadley]

Could you please squash? (And close #833?)

[krlmlr]

Should be good to merge now. I have no idea what's wrong with the coverage -- I've added explicit tests. @jimhester Any idea?

[jimhester]

It is failing the "patch coverage" test. Patch coverage refers to

The lines changed in each commit must meet this minimum coverage. Recommended value of 90%+.

https://github.com/codecov/support/wiki/Patch-Coverage-Status

Presumably because you did not add any new tests for the changed lines from this pull request. Looks like @hadley set it so that 90% of the changed lines need to have coverage to pass.

[krlmlr]

Thanks @jimhester for the quick reply. My point is:

• I have added a test for the new code
• The test calls with_lib and with_libpaths
• These call (albeit internally) set_lib et al., which are not covered

[jimhester]

Ah I misunderstood the issue, I will try to figure out what is going on
then.
On Jul 3, 2015 9:01 AM, "Kirill Müller" [email protected] wrote:

Thanks @jimhester https://github.com/jimhester for the quick reply. My
point is:

• I have added a test for the new code
  https://github.com/hadley/devtools/pull/836/files#diff-5c28a8e2e00a665df9b1f5938388e7c0R44
• The test calls with_lib
  https://github.com/hadley/devtools/pull/836/files#diff-7793392c33c95f57fb11771ab7b11f2aR153
  and with_libpaths
  https://github.com/hadley/devtools/pull/836/files#diff-7793392c33c95f57fb11771ab7b11f2aR139
• These call (albeit internally) set_lib et al., which are not covered
  https://codecov.io/github/hadley/devtools/R/with.r?ref=20bcb8662ccfdf949d2aa71ed62220513b0a7974#l-144

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

[krlmlr]

Thanks. I probably should have been a bit more specific right from the start.

[jimhester]

@krlmlr So the reason set_lib() was not being covered was because you were forcing the set and reset functions in with_something(). This forces a copy of the functions to be created, and only this copy was being run, not the original function. Removing those lines shows the proper coverage.

[krlmlr]

Removed the force() call as suggested by @jimhester and added a test to satisfy codecov.