Which architectures to support in a CI like Travis?

[Martin Di Paola]

I would depend on the project.

In the crytoanalysis tool that I developing, "cryptonita", I just
manipule bytes. Nothing that could depend on the distro so my CI picks
one OS and run the tests there.

Project: https://github.com/cryptonitas/cryptonita
CI: https://github.com/cryptonitas/cryptonita/blob/master/.github/workflows/test.yml

On the other extreme I have "byexample", a tool that takes the examples
in your docs and run them as automated tests. It supports different
languages (Python, Ruby, Java, ...) and it works using the interpreter
of each languages.

An there is the challenge for its CI. Because byexample highly depends
on the version of the interpreter, the CI config tries a lot of
different scenarios

Project: https://byexamples.github.io/
CI: https://github.com/byexamples/byexample/blob/master/.github/workflows/test.yml

I don't tests different distros but I should for some cases that I
suspect that it could be differences in how some interpreters behave.

An about OS, I'm planning to add MacOS to the CI because I know that
some users had problems in the past in that platform because how
byexample interacts with the terminal.

So two projects, both in Python, but with two totally different
dependencies on the environment where they run, so their CI are
different.

The two examples are using Gitlab actions but the same applies to
TravisCI.

Thanks,
Martin.


On Sun, Sep 18, 2022 at 09:46:45AM +0000, c.buhtz at posteo.jp wrote:
>Hello,
>
>I am using TravisCI for my project on GitHub. The project is packaged
>for Debian, Ubuntu, Arch and several other distros.
>
>All this distros support multiple architectures and they have their own
>test machines to take care that all packages working on all archs.
>
>On my side (upstream) I wonder which arch I should "support" in my
>TravisCI configuration. I wan't to speed up travis and I want to save
>energy and carbon.
>
>I suspect that my Python code should run on much every platform that
>offers a Python interpreter. Of course there are edge cases. But they
>would be captured by the distros own test environments.
>
>So is there a good and objective reason to support multiple (and maybe)
>exotic platforms in a CI pipeline on upstream?
>
>Kind
>Christian
>-- 
>https://mail.python.org/mailman/listinfo/python-list