infrastructure VS ome

In terms of how I got better, it was primarily through project requirements driving improvements in CI automation. I've gone through several CI systems, from Jenkins to GitHub/Travis-CI/AppVeyor, to GitLab CI and others. Developing cross-platform software means testing it at on all the platforms you support. Number one task for each platform is to develop and document how to provision all of the needed third-party prerequisites (example). You can then automate that as appropriate, using whatever tooling is appropriate. For example, Dockerfiles, setup scripts, Ansible/Chef (examples), manual instructions. The aim here is to make it possible to make it possible to recreate a working setup, and keep your setup up-to-date either by applying updates or recreating a fresh copy. When supporting many platforms, maintaining all of the infrastructure can be a full-time job in its own right.

As an example, one dependency is this job: https://gitlab.com/codelibre/ome/ome-common-cpp/-/jobs/4083774435

Files: build log, CI config, common CI config.

I can't claim these are perfect examples of best practice, but feel free to look over here and here for a variety of tests. They provide examples of test fixtures, value-parameterised tests and type-parameterised tests. I got everything directly from the official docs, which are fairly terse, but could have a few more examples.

On Linux, you have the benefit of package managers with all the common libraries ready to install with apt or yum/dnf. Same on FreeBSD with pkg. On MacOS you have MacPorts or Homebrew. On Windows, we now have vcpkg. I've chopped and changed which platforms and package management systems to use over the years depending upon project requirements. Before vcpkg, I wrote my own CMake-based "super-build" for Windows, for the subset of libraries and tools I needed. You can browse an archived copy here.