Mark Callow

The latest main has a fix for the Windows build errors you are now seeing. The errors are due to an update to ClangCL which adds a new warning about...

@javagl please make some progress on this. I would like to include these changes as part of the eventual 4.4.1 release. I have no date in mind but already have...

> (automatically from the tar.gz downloaded) What is the information contained in the downloaded tar.gz? Is it the name of the git tag?

The only bit that wasn't, and still isn't, clear is "automatically from the tar.gz downloaded". I get you download the artifacts for a particular tag and I get that absent...

`scripts/mkversion` predates use of CMake and creation of `cmake/version.cmake`. I don't know if it is reasonable to combine them. Passing a `--version` argument as you propose seems like a reasonable...

> I could propagate the tag to the mkversion with e.g a special argument like --version or something like this. Would that be enough? Are you working on this?

The intent is that the version emitted by a tool's `--version` option can be fed direct to `git checkout` to checkout the matching source code so the 'v' is required....

@xoofx, In your [PR](https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/69457) you have assimp-dev as one of the dependencies. Assimp is only needed by KTX_FEATURE_LOADTEST_APPS and you are not configuring that.

The name change to `KTX_GIT_VERSION_FULL` breaks the package builds in `CMakeLists.txt` because they rely on `KTX_VERSION_FULL being` set by `cmake/version.cmake`. This won't show up in CI at this point because...

> But as I said earlier, `KTX_VERSION_FULL` is still set in my PR by this line that I haven't changed: Sorry. I missed that line.