Because we were generating a fontconfig file without the default
configuration, non optimal fonts where selected during substitutions,
for example instead of the very common Arial, we were substituting a
serif font which looks completly different.

See merge request !753

setUpModule is expensive and also the snapshot in case of error would
fail if called more than once

This is a loop invariant.

Is there an option to only ignore KEYEXPIRED ?

Currently, we could just list gettext but for maintainability,
it's easier to compare the list with only python&gcc's buildout.cfg

- pylint and other linters were not working in the new version of python extensions.

- configure URLs for webviews in something compatible with our frontends.

See merge request !755

Because we were generating a fontconfig file without the default
configuration, non optimal fonts where selected during substitutions,
for example instead of the very common Arial, we were substituting a
serif font which looks completely different.

This disable a theia security feature of using a different hostname for
each webview. By defaut, for each webview, thiea generate an unique
hostname that when using vifib frontends would be something like
https://uuid.webview.softinstXXX.host.vifib.net but that's not usable in
our case because we cannot create frontends for a subdomain (also we
don't have certificates for such domain).

Configure THEIA_WEBVIEW_EXTERNAL_ENDPOINT to something less secure, but
working in our environment. This fixes embedded jupyter notebooks and
other web views.

This reverts commit 7e3bc5ee.

Theia does not seem to be ready for this version, several things are not
working:
 - selecting another python interpreter
 - linting in the editor
 - at startup there's always a "Python is not installed.Please download
and install Python before using the extension." warning . I think this
happened sometimes with the previous version too.

jedi was working fine, but too maybe critical features are broken.

add more tests for KVM and fix them

See merge request !752

fixes 794185d0

Otherwise it's installed by python setup.py outside of buildout

This is partially WIP but I let @jerome, @Nicolas and @tomo to review it. 

Thie WIP is resilient tests included on it, it may cause the test runs for several hours.

My current idea is release all except resilient one, them use the Test case as a Library and recreate additional folders/softwares, which will run resilient tests. 

It still required to see if we can optimise the compilation maximazing the use of shared.

See merge request !748

This MR is a complement of nexedi/slapos.core!214, however it can be merged before the release of slapos.core because it won't be used until slapos.toolbox (actuall runner code) is changed to call them.

/cc @jerome @tomo @Nicolas

See merge request !751

This reverts commit de1e9ea0.

2.7.1+slapos006 revealed a bug in buildout bug, breaking resilient KVM.

Commit f355e0af didn't work because
GCC building ignores LDFLAGS when linking libgfortran.so
whereas the latter does not even need zlib and would be linked to
system zlib (if it exists) for nothing.

In our case, there seems to be no zlib code inside the installed files.