Peter Lindstrom

`cudaGetLastError` is quite clearly part of the CUDA API, so there's something more fundamental going on here. It may be useful to `make VERBOSE=1` (not sure what the Windows equivalent...

@kminemur Can you please check to see if the latest changes on the `staging` branch address your issue?

I wonder if the MSVC flags `/wd4146` and `/wd4305` are tripping nvcc up and making it think those are source files. What if you comment those out here: https://github.com/LLNL/zfp/blob/3d2d67a37545c40bc693f7c0e0ba614b54272396/CMakeLists.txt#L80-L81

Update: I can confirm that commenting out those two CMake lines fixes the above nvcc issue (on the `staging` branch). Ideally those warning suppressions would be passed only to the...

This last issue is fixed by substituting `#define inline_ inline` with `#define inline_ static inline` here: https://github.com/LLNL/zfp/blob/3d2d67a37545c40bc693f7c0e0ba614b54272396/src/cuda/shared.cuh#L18 These changes allow me to build zfp on Windows with `-DZFP_WITH_CUDA=ON`, though without...

@kminemur I believe we've now addressed the CUDA build issues on Windows on the `staging` branch. Can you please test it and see if it works for you?

Great to hear. Since we don't have a Windows box with an NVIDIA GPU, would you mind running the tests also (by building with `-DBUILD_TESTING_FULL=ON` and then running `ctest`) to...

@kminemur Thanks for trying. I believe there are a few issues at hand. First, I saw similar google test linkage warnings on my machine, which from a cursory look may...

@kminemur Can you please confirm whether or not you ran `ctest -C Release` and not just `ctest`? On Windows, you have to specify which configuration to test. I see no...

@kminemur Thanks for letting us know. The "failing" tests are the C tests that depend on CMocka and which are not being built, as that requires permission to make symlinks...