Fix generation of coverage for c++ code

[betatim]

Addresses #1511 (comment) where @ctb noticed that we do not generate coverage for c++ code anymore.

• Is it mergeable?
• make test Did it pass the tests?

[codecov-io]

Codecov Report

Merging #1617 into master will decrease coverage by -25.23%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           master    #1617       +/-   ##
===========================================
- Coverage   95.31%   70.09%   -25.23%     
===========================================
  Files          36       66       +30     
  Lines        2969     8902     +5933     
  Branches      445     3008     +2563     
===========================================
+ Hits         2830     6240     +3410     
- Misses         67     1041      +974     
- Partials       72     1621     +1549

Impacted Files	Coverage Δ
lib/hllcounter.hh	100% <ø> (ø)
lib/traversal.hh	71.42% <ø> (ø)
lib/kmer_hash.cc	86.59% <ø> (ø)
lib/storage.hh	89.69% <ø> (ø)
lib/read_parsers.hh	65.62% <ø> (ø)
khmer/_cpy_nodetable.hh	45% <ø> (ø)
lib/labelhash.cc	52.47% <ø> (ø)
khmer/_cpy_smallcounttable.hh	45% <ø> (ø)
lib/read_aligner.hh	100% <ø> (ø)
lib/assembler.cc	52.87% <ø> (ø)
... and 20 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 42e3f88...bef16bb. Read the comment docs.

[ctb]

Well, something's working, although it looks a bit weird :).

[ctb]

I can only assume this is part of @betatim's master plan.

[betatim]

So it seems like make coverage-debug thinks there is nothing to be done after make test has been run. However without recompiling with -fprofile-arcs -ftest-coverage etc you don't generate the files that gcovr needs to calculate the coverage.

Compare for example: https://codecov.io/gh/dib-lab/khmer/tree/32fcf25137937de2dc52d214724f87312434c51a and https://codecov.io/gh/dib-lab/khmer/pull/1617/tree one of them ignores all the files that aren't .py.

[betatim]

Is there a better way to pattern match these files that has less potential for accidentally deleting stuff that it shouldn't at some point in the future?

[standage]

The risk for inadvertently destroying unrelated files is extremely low.

[betatim]

If someone can think of a nicer rm *gcov pattern I'll do that, and if a second pair of eyes and brains could think about if there is a nicer way of doing this that would also be good. Otherwise I think we should merge. And then spend our time trying to dig ourselves out of this loss of coverage :(