Overhaul Travis & Codecov setup

[fingolfin]

For Codecov, actually enable status messages in pull requests (so far, all were turned off, but for some reason codecov ignores this; still, better to specify actual intent in the file).

For Travis, adapt scripts/build_pkg.sh to work with packages not using autoconf, so that we can use the same script across many packages. Also don't copy the tested package into GAP's pkg dir; instead, put it into its own GAP root which we tell GAP to look at before its usual GAP root.

Modify scripts/gather-coverage.sh and scripts/run_tests.sh to use this modified list of GAP root dirs

Simplify scripts/build_gap.sh significantly by using BuildPackages.sh.

[codecov[bot]]

Codecov Report

Merging #5 into master will not change coverage. The diff coverage is n/a.

@@           Coverage Diff           @@
##           master       #5   +/-   ##
=======================================
  Coverage   84.63%   84.63%           
=======================================
  Files           7        7           
  Lines        1946     1946           
=======================================
  Hits         1647     1647           
  Misses        299      299

[fingolfin]

Please do not merge this just yet: the profiling compilation failed (due to clang++ issue), which is a problem in itself, but also the fact that this did not cause the build to be marked as failed.

Of course it's a questionable why this package is tested with clang in the first place, as it contains no compiled code?

[fingolfin]

This is now waiting on https://github.com/gap-system/gap/pull/2022

[olexandr-konovalov]

Of course it's a questionable why this package is tested with clang in the first place, as it contains no compiled code?

Setup was inherited from some other package, which used compiled code. It makes no harm to test a package without compiled code on different setting as well, isn't it? For example, FactInt, I assume, may exercise GMP heavily.

[fingolfin]

I don't think it exercises GMP more heavily than the GAP test suite. It mostly performs elementary operations only.

Anyway, this PR is now IMHO ready for merge.

[olexandr-konovalov]

OK - I want to edit the history a bit, should that be done now, or after?

[fingolfin]

Which history? That of the master branch of this package? Or of this PR? I am fine either way, just if you do it before merging the PR, let me know, because I'll have to rebase it.

That said, actually I'd like to edit this PR some more to make it trivial to add tests against the stable branch and/or HPC-GAP in the future (if desired). So, just go ahead with editing the master history :-)

[olexandr-konovalov]

@fingolfin I want to edit master history to make changes within last two weeks atomic. I will ask you and @Stefan-Kohl to reset your repositories then.

[Stefan-Kohl]

Should I merge this?

[olexandr-konovalov]

@fingolfin not yet, please- I want first to understand why Example package fails, with the same setup.

[fingolfin]

Rebased