TST: gcov missing a few files in report

@pv correctly notes that a few C source files are missing from our coverage report.

Digging through a recent master branch build log, I can see some “warning” style output for gcov of this nature:

./_configtest.gcda:stamp mismatch with notes file
./build/temp.linux-x86_64-3.6/numpy/core/src/dummymodule.gcda:cannot open data file, assuming not executed
./build/temp.linux-x86_64-3.6/numpy/core/src/private/mem_overlap.gcda:stamp mismatch with notes file
./build/temp.linux-x86_64-3.6/numpy/core/src/private/npy_longdouble.gcda:stamp mismatch with notes file
./build/temp.linux-x86_64-3.6/numpy/core/src/private/ufunc_override.gcda:stamp mismatch with notes file
./empty/_configtest.gcda:cannot open data file, assuming not executed

Possible solutions to this issue have been discussed elsewhere & appear related to build / machine time stamps relative to tests.

When I test locally, this is not reproduced; my current idea is to first open a PR that tries one of the simpler solutions – i.e., using find to recursively touch the generated *.gcda files. I’ll first see if I can reproduce the issue locally by touching the *.c and *.gcno files before running gcov recursively so that the timestamps are messed up. It looks like one user reports that touch may not suffice because the timestamp is actually embedded in the gcda files – I’ll have to check this.

[pv]

Maybe it’s related to the fact that some files are compiled twice (included both in umath and multiarray). If so, gh-10915 should fix it.

[tylerjereddy]

Ok, proposed a fix that works locally in that PR for mem_overlap