Windows: fix Makefile.Windows

[gvanem]

Extend the Windows-build (for CI) with:

• Add rules for target/client.exe and target/server.exe.
• src/crustls.h is no longer generated (why should it?)
• The C-library is now named target/release/rustls_ffi.lib.
• Reduce warning with -D_CRT_NONSTDC_NO_WARNINGS.

[gvanem]

After fixing the test.yaml file (that now works), there are issues with an #include <sys/uio.h> in tests/client.c.
Ref: https://github.com/rustls/rustls-ffi/runs/4137631017?check_suite_focus=true

Another PR to clean-up the mess in tests/*.[ch] coming soon.

[jsha]

This looks good, but still fails in CI with:

link -nologo -incremental:no -debug -map -verbose -out:target/client.exe  common.obj client.obj target/release/rustls_ffi.lib advapi32.lib userenv.lib ws2_32.lib > link.tmp
link: unknown option -- n

According to some googling, it sounds like this is most likely because there is a copy of the MinGW linker in the path ahead of the MSVC linker. I'm not sure how best to resolve. If we can figure out the page we can move it aside.

[gvanem]

... there is a copy of the cygwin linker in the path ahead of the MSVC linker.

Add a del /Q <Cygwin-root>\bin\link.exe 2> NUL? But what is Cygwin's root on Github Actions?

[jsha]

I added link /h to the build and got this output:

Microsoft (R) Incremental Linker Version 14.29.30136.0
Copyright (C) Microsoft Corporation.  All rights reserved.

LINK : warning LNK4044: unrecognized option '/h'; ignored
LINK : warning LNK4001: no object files specified; libraries used
LINK : warning LNK4068: /MACHINE not specified; defaulting to X64
LINK : fatal error LNK1561: entry point must be defined
Error: Process completed with exit code 1.

So it seems like we actually are getting the right linker? I'm pretty mystified about why we're getting the "unknown option" error.

[jsha]

Thanks for doing this! I messed around with GitHub Actions and got things working. I put the resulting fixes in #211. Closing this PR out in favor of that one.

[gvanem]

I'm pretty mystified about why we're getting the "unknown option" error.

The unknown option -- n is possibly due to a MAKEFLAGS env-var in some other step.