more ci environments

[ebraminio]

I've provided lots of different environments for harfbuzz https://github.com/behdad/harfbuzz/blob/master/.circleci/config.yml one not merged yet 1. I was wondering if the same is helpful for libui project.

[ebraminio]

of course the most vital one is appveyor which is included on this PR

[simplexidev]

Something like this would be useful. I'm writing (another) C# binding for libui and with no experience building with/using cmake (or any C/C++ compiler for that matter), this would be useful to have authentic (possibly static) libraries for the various OSes supported.

[ebraminio]

@tom-corwin, hi there, if you are using Windows I guess this will help you https://github.com/andlabs/libui/pull/157/files#diff-180360612c6b8c4ed830919bbb4dd459 it is not merged yet and as you mentioned just correctly that it should be done.

[parro-it]

@tom-corwin , I'm the author of the PR you spoke about.
Even if it's not merged yet, I'm publishing prebuilt binaries of libui on my fork:
this is the latest release of libui: https://github.com/parro-it/libui/releases/tag/alpha3.5-004

[simplexidev]

Thanks for the information @ebraminio and @parro-it. I'll definitely be shipping them with my bindings, once the libutf-and-attrstr branch is merged into master. Currently I'm writing them for the upcoming changes.

[andlabs]

I could probably start building ARM and ARM64 packages using Azure Pipelines since I think I figured out a good-enough way of cross-compiling without odd hardcoded lists of extra packages (there is an odd hardcoded list of packages to downgrade, but). Testing those, on the other hand, is another story entirely. But we can cross that avenue once I make libui have a proper automated test framework and not the clunky beast of a manual testing system I have now.

The current CI setup I have covers Windows, Mac, and Linux. Really the most I can ask for from "more CI environments" is "other OSs to CI with". I know Azure Pipelines users often use Docker for these types of things, but it appears you'll have to run your entire pipeline as a giant Docker script, which defeats the point of using the features of your CI system. (I've only really seen NumPy's CI setup, though.) CircleCI appears to do both as per the harfbuzz link at the top, but it's also rather expensive.