GitLab

PEP_MACHINE_DIR must be set. It was removed by accident in a previous
commit.

More closely follow the make commands according to the Engine docs.

This adds the command "make db" as well as using "make all" instead of
just "make"

Due to a bug[0] in one of the Sequoia makefiles which tries to use
`ginstall` for BSD & Darwin systems, we have to make a symlink in our
$PATH from "install" to "ginstall"

[0] https://gitlab.com/sequoia-pgp/sequoia/-/blob/365d00a08bec6a5a48d48a7c7893d78c27092b59/openpgp-ffi/Makefile#L23-24

We should now have no problem using the latest version of Sequoia's
"pep-engine" branch, now that we are using Rustup to manage our Rust
versions.

This reverts commit c05d7917.

The web server on fdik.org is not sending the interemediate chain for
its Let's Encrypt certificate. We have to download it manually, so that
wget doesn't complain about an unauthorized cert.

Sequoia requires newer versions of Rust than what's available in the apt
repository. We can use Rustup to install and switch between Rust
versions.

Run apt-get upgrade on our base image (toolchain), to make sure we're
running with the latest packages.

head of pep-engine is failing, so we roll back to last version known to
build.

Another missing "&&" causing build to fail

Old versions of Engine, libpEpAdapter, and Sequoia were being built. Now
they are using the latest versions as of the date of this commit.

Removed the unused Adapter version variables from the Makefile while I
was at it.

Dangling "&&" was causing build to fail.

We want to keep things as close as possible between the build files for
each distro. This brings the Debian 10 Dockerfiles up to par with the
CentOS/RHEL8 Dockerfiles.

Make the dependency script fail quickly in case of problems.

house cleaning

Setting this DEBIAN_FRONTEND variable to "noninteractive" makes things
like apt behave better when in a non-interactive shell.

The old job from before the jobs were split, was still around. Deleted
this and also fixed a typo in the debian10:toolchain title.

Use heck's toolchain/dependencies script from the root of this repo.
It's slightly modified to be more succinct, and use some apt arguments
which are better for running in Docker. "sudo" is removed from the list
of packages, as it is not needed.

Copying the pattern set-out in the Centos8 subdirectories, this commit
splits up the Debian 10 builds into their own directories, with their
own makefiles. The jobs in the .gitlab-ci.yml have been updated to match
accordingly, and are overall more standardized.

In an effor to more closely follow Sequoia's canonical CI environment,
let's install clang and explicitly check versions of cargo and rustc.

ETK CLI is not included anymore, so description should not mention it.

If the directory tree for storing the packages/binaries does not exist
on the remote host, then we will create it.

rsync is a better choice than curl for internal hosts.

paths needed more updating due to the structure of the new depot host

Have a new depot directory structure, so the path in the rsync commands
needed updating.

This removed YAML redundancy in our .gitlab-ci.yml file.

The merge is of this remote branch:
https://pep-security.lu/gitlab/hartmut/Internal-Deployment/-/tree/master

Thanks hartmut!

Build script for PyAdapter now uses "make" instead of "setup.py"
directly. This adheres to the recommended way of building PyAdapter.

ETK builds are failing and preventing package creation, so this commit
disables it.

Scripts needed a bit of updating to work with latest JNIAdapter build.

Also the "Basic.java" test needed an additional import. Thanks to heck
for the tip.

- Add "hidden jobs" for defining common building blocks (e.g. ensuring
  docker is installed)
- Add a "hidden job" `standard_job` providing a common
  `before_script`.
- Base all jobs on `standard_job`, removing `tags` and `before_script`
  there.

  Only exception: For job `centos8:etk-cli-connector` keep the
  `before_script`, which currently differs from the standard one -
  anyway reusing the building blocks for common steps.

As recommended by the manual
<https://docs.gitlab.com/ee/ci/yaml/README.html#special-yaml-features>:

  In most cases, the extends keyword is more user friendly and should be
  used over these special YAML features. YAML anchors may still need to
  be used to merge arrays.