Run codecov in appveyor ci

[atugushev]

Increases the test coverage.

[blueyed]

This will likely cause problems with codecov timing out though - if there are non already (given the huge number of builds for linux already).
We're having trouble (likely due to flags and number of builds) in pytest (and other projects) since months.

So in general it would be better to only run selective envs with coverage, by using a coverage factor in tox (see pytest's tox.ini).

[codecov]

Codecov Report

Merging #745 into master will increase coverage by 1.19%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #745      +/-   ##
==========================================
+ Coverage   86.07%   87.27%   +1.19%     
==========================================
  Files          33       33              
  Lines        2126     2082      -44     
  Branches      303      303              
==========================================
- Hits         1830     1817      -13     
+ Misses        231      201      -30     
+ Partials       65       64       -1

Flag	Coverage Δ
#linux	85.83% <ø> (?)
#windows	86.4% <ø> (?)

Impacted Files	Coverage Δ
piptools/__main__.py	83.33% <0%> (-2.39%)	⬇️
piptools/scripts/sync.py	52.83% <0%> (-0.88%)	⬇️
piptools/repositories/local.py	90.69% <0%> (-0.8%)	⬇️
tests/conftest.py	93.75% <0%> (-0.7%)	⬇️
piptools/utils.py	89.32% <0%> (-0.11%)	⬇️
piptools/scripts/compile.py	87.76% <0%> (-0.09%)	⬇️
piptools/cache.py	94.44% <0%> (-0.08%)	⬇️
piptools/resolver.py	94.59% <0%> (-0.08%)	⬇️
tests/test_writer.py	100% <0%> (ø)	⬆️
tests/test_cli_compile.py	100% <0%> (ø)	⬆️
... and 9 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c45862b...fc69f3a. Read the comment docs.

[atugushev]

We're having trouble (likely due to flags and number of builds) in pytest (and other projects) since months.

@blueyed could you please tell me what kind of problems do you mean?

[blueyed]

@atugushev
When looking at PRs via codecov comments or the UI in general it might time out on their backend ("connection reset").
See pytest-dev/pytest#4846 (comment).

[blueyed]

Here https://codecov.io/gh/jazzband/pip-tools/pull/745?src=pr&el=h1 is not timing out, although there are a lot of builds.
Source code size might also be a factor - I don't know.

[blueyed]

So I suggest adding flags ("windows" / "linux" (or rather $TRAVIS_OS_NAME)) here after all.

[atugushev]

@blueyed

So in general it would be better to only run selective envs with coverage, by using a coverage factor in tox (see pytest's tox.ini).
See pytest-dev/pytest#4846 (comment).

I see. Running selective envs with coverage sounds reasonable to me now. I'll try to test it and propose a PR soon.

So I suggest adding flags ("windows" / "linux" (or rather $TRAVIS_OS_NAME)) here after all.

Builds section looks pretty good now, see the latest codecov report. Thanks for the suggestion, Daniel!

[blueyed]

For reference: extra coverage due to this: https://codecov.io/gh/jazzband/pip-tools/compare/c45862bd3e99df71fc020f828395fc7f6766be7c...fc69f3a2eab8d54c1a9850dcb2fdd9ec3d6930c5/changes

But it could be achieved with a single (or few) jobs only probably also.