[vbezhenar]

May be I'm missing something, but CI looks like the thing that's trivially implemented by a few bash scripts. It won't be as pretty and as reliable compared to battle-tested service, but generally it'll work. I did that in the past and it worked pretty well for our uses.

[hrktb]

The most valuable part of CI where I work right now is it being run on a reference setup, with persisted builds and logs attached to each merge requests.

We are all supposed to run tests, coverage and benchmarking localy. The point in CI is to guarantee nobody cut corners and keep verifiable proof of good builds.

We could fulfil the same requirements with a hand made setup on a box somewhere uploading subsequent artefacts, but I am not sure we would be happier on any specific aspect. Even cost wise, setup and maintenance could non trivial compared to the amounts we're paying now.

[johannes1234321]

A few bash scripts work up to some size. For many projects at some time demands increase and you got longer test runs, which eventually require more machines and and different independent build steps (maybe you don't want to build the core libraries which are rarely touched for each commit of other parts etc.?) and at some point it breaks apart and you need something "proper" especially if you then also want to integrate test results with code review and track them from the single planning tool.