Fix missing-braces issue for Rcomplex

[MichaelChirico]

Closes #134. This approach avoids the need for old-vs-new preprocessor branching.

[eddelbuettel]

No, not a fan of a preprocessor macro based approach, sorry.

[eddelbuettel]

Rejecting as using macros is step backwards. An inline function with preprocessor step for clang > whateverthreshold may work.

[MichaelChirico]

Sure! Made it a routine instead, PTAL.

[MichaelChirico]

preprocessor step for clang > whateverthreshold

I'm taking the approach of branching on R version because that's where the actual typedef for Rcomplex changes, a branch on the compiler would just be to avoid compiler warnings. If we think the warning is good+correct, the R version branching is better.

[eddelbuettel]

Micro nags. We're getting there.

[eddelbuettel]

Yup, that's pretty much what I would have done and the check on the R version may actually be net better than what I had in my first attempt. We can live with this.

[eddelbuettel]

Can you please check why macOS is crying?

[MichaelChirico]

Argh. Sorry. I can't tell why. Do they give you and GHA artifacts that would include 00install.out? I didn't see any but I might just lack permissions.

Don't spend more than a few minutes on this, I'll try and figure out another way if it's causing you headache.

[eddelbuettel]

Do they give you and GHA artifacts that would include 00install.out?

We have to add a block such as towards the end of the file ci.yaml:

      - name: View Logs
        run: .run.sh dump_logs
        if: failure()

Otherwise I can try mac-builder later but I presume you have the requisite tar.gz handy?

[MichaelChirico]

The earlier Mac issue was probably due to my sloppy copy-pasting, I had

makeComplex(double *re, double *im)
// then later
makeComplex(re, im); // Should be &re, &im!

Obviated the issue by just making copies (which I think we need to do and IINM is what the original code was doing).