On Tue Oct 31 16:07:18 2023 +0000, Giovanni Mascellani wrote:
We don't seem to print anything if a test is skipped due to not
meeting its requirements, should we be? You're based on a rather old master. Commit https://gitlab.winehq.org/wine/vkd3d/-/commit/c1de65a99ba851f97cb5f345b66a2b... should do what you want, if I interpret it well. It wouldn't be bad to rebase on current master anyway, so that the pipeline is run. I haven't read the patches yet (I will do it as soon as I can), but it's likely the commit messages might require some work.
Also, I would reorder the commits in roughly this way: first the tests, then the HLSL compiler, then the SPIR-V and vkd3d bits. The idea is that you implement features in the same order they appear in the pipeline, so that reviewers can follow patches in their order and build on top of earlier patches at each step.
As for commit messages, I think that in general the following features are appreciated: * Terminate each sentence, including the commit subject, with a full stop. * Avoid nominal sentences, like "ROV support", but rather describe what is actually happening with a verb. * When possible, use specific verbs. I won't say that "support" is banned, but often there are better alternatives, like: "Parse ROV types", "Enable EXT_fragment_shader_interlock", "Support SV_PrimitiveID in pixel shaders" (using "pixel shader" rather than "fragment shader" because that change is in the context of the TPF format). * This is probably my taste more than anything, but at least when a certain acronym is often associated with a specific capitalization, I'd use it. Case in point: "SPIR-V".