Fix travis build running twice the test suite

[64kramsystem]

Fixes #518

[codecov]

Codecov Report

Merging #519 into master will decrease coverage by 50.79%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #519      +/-   ##
==========================================
- Coverage   83.62%   32.82%   -50.8%     
==========================================
  Files          52       52              
  Lines        9111     9111              
==========================================
- Hits         7619     2991    -4628     
- Misses       1257     5967    +4710     
+ Partials      235      153      -82

Impacted Files	Coverage Δ
vm/concurrent_hash.go	0% <0%> (-97.55%)	⬇️
vm/string.go	4.69% <0%> (-91.73%)	⬇️
vm/hash.go	5.9% <0%> (-91%)	⬇️
vm/http.go	0% <0%> (-88.67%)	⬇️
vm/array.go	7.57% <0%> (-86.78%)	⬇️
vm/decimal.go	9.75% <0%> (-86.18%)	⬇️
vm/uri.go	0% <0%> (-82.83%)	⬇️
vm/repl.go	0% <0%> (-81.4%)	⬇️
vm/range.go	10.09% <0%> (-80.77%)	⬇️
vm/concurrent_array.go	0% <0%> (-79.73%)	⬇️
... and 26 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 b0c4238...8be4921. Read the comment docs.

[st0012]

@saveriomiroddi It seems that the new script doesn't import test coverage data properly, so the coverage is now down to 32%.

[st0012]

I should've noticed this before merge. I thought the 32% is about target coverage 😫

[64kramsystem]

It seems that the new script doesn't import test coverage data properly, so the coverage is now down to 32%.

This is very likely caused by the original script having a bug also in the coverage calculation; in this case, it should be because the strategy used:

if [ -f profile.out ]; then
  cat profile.out >> coverage.txt
  rm profile.out
fi

doesn't work as intended, for some reason.

[st0012]

Um.....this should be fixed ASAP. I'm investigating now.

[64kramsystem]

It's hard to work on this area (full testing) without the database setup. I'll work on setting up the database next, and I'll put the instructions on the wiki.

[st0012]

If you already have PG installed and have a proper PG role/user, run these commands and it should work:

psql
create database goby_test;

[64kramsystem]

During the last attempt, things didn't turn out to be trivial overall, I'll need to review it.

[st0012]

I think I found the reason, the command that runs all vm tests doesn’t generate profile result.
We used to use the coverage of the first run and never have the coverage of the second run. And the old first run script doesn’t have the —run flag to filter test cases, so it runs all tests and gives full coverage. But now it only runs those not race detecting tests.
So the solution should be adding the same if statement block that inserting profiling results, just below the second vm test command and above continue (also remember to add coverage flags to second test command)