See nexedi/slapos@22cc8e7e

See merge request nexedi/slapos!1251

Includes buildout, configuration, test, and resiliency support

See merge request nexedi/slapos!1298

Allow ffmpeg to get the correct libvpx path

Using "Sort Lines Ascending" from Theia or `:'<,'>!sort -f` from vim
produce this diff.

See merge request !1279

See merge request nexedi/slapos!1295

This is more coherent with other SR. At some point, we hsould get this
file from nginx installation.

See merge request nexedi/slapos!1294

depends on nexedi/erp5!1695

See merge request nexedi/slapos!1290

The coverage of an individual test is after all not useful, only the
coverage of all tests combined is significant.

This replaces the broken --coverage argument from runUnitTest, coverage
needs to be started earlier and also introduces a coverage plugin to
collect coverage data for code in ZODB; python scripts and components
are supported.

To use on test nodes, set up a web dav server somewhere, configure the
instance parameters of the test suite on ERP5 to enable coverage and
upload individual results to this webdav server and then combine the
coverage data and produce a report, using the bin/coverage script from
the software release.

For the steps below, it is necessary to change working directory to the
root of the software folder.

Step 1: combine the coverage data:

    ./bin/coverage combine \
      --keep \
      /path/to/all/coverage/files/*coverage.sqlite3

( using --keep is optional, but it helps in case of mistakes )

Step 2: build an html report:

   ./bin/coverage html \
     --skip-covered \
     --omit parts/erp5/product/ERP5/Document/UnitTest.py \
     --directory /path/for/html/report/

Note that we want to omit UnitTest.py that is created during test
(see testBusinessTemplate.TestDocumentTemplateItem) and get coverage
result because it is executed in the test, but it does not exist as a
source file in the repository, so we skip it.

Of course, to produce a correct html report from a test that have been
running on test nodes, it's necessary that the software release used to
produce the html report has exactly the same version ans the one that
has been running on test nodes.

Another simpler, but slower approach is to run all the tests on the same
machine, then only running step 2 is necessary.