Maik Riechert

@WardF How are you progressing on this one? If you need someone for testing just let me know!

It's been a while, any updates on this?

Please see my earlier comment https://github.com/Unidata/netcdf-c/issues/554#issuecomment-354823851. You essentially tried the same here. It would be good if you get a local setup to test it yourself, otherwise it takes too...

Just to clarify, if control flow guard (CFG) was enabled during compilation of a static library (here MSMPI), does that mean that dependent libraries or executables have to be compiled...

I just checked with dumpbin, indeed msmpi.lib does not have the `__guard_check_icall_fptr` symbol. Is this expected? Regarding CFG in gcc/gfortran, there's no support for it as far as I see,...

@jithinjosepkl You wanted to check with the compiler team. What was the result of that? In the meantime, I found the following work-around which simply stubs out the unavailable CFG...

Which runtime exactly? Do you have a DLL filename? MinGW links against MSVCRT.

Still not there. Will this ever be fixed? It's kind of broken as it is now. Homebrew uses the [CMake](https://github.com/Homebrew/homebrew-core/blob/master/Formula/netcdf.rb) variant and hence doesn't have a working nf-config.

> So if we implement this, I think it should be by exposing the current memory used by the historical query cache, per-handle, and implementing a size-based culling policy as...

`write_only` is weird, since you can also read. I would mirror the existing naming with `make_endpoint` and leave the existing adapter as is, while adding a new one as `read_only_adapter_v*`