Use std::isnormal(), remove LazyData

[eddelbuettel]

Salut Stephane,

As discussed over in eddelbuettel/bh#80 I am looking in upgrading package BH to a new Boost version. Your package did not build, whereas it does at CRAN, and it turns our to be a pretty bening issue that isnormal now clashes so switching to a more explicit std::isnormal fixes it. (While at I noticed that CRAN notes an unneeded LazyData: true so I removed that too. Feel free to reverse that part.)

It would be great if you could ship an updated gfilogisreg to CRAN as I plan to send BH there too, given that it mostly sails through cleanly.

If you have any questions just ping me.

[stla]

Ok, thanks. Yes, LazyData must be removed now when there's no data in the package. Solaris has been removed from the test platforms? That's good news.

[eddelbuettel]

Yes, the Solaris change was greeted a little bit on social media when users (Henrik may have been first) spotted the change in the repo / sources as documentation updates. An era comes to end, net-net it did of course help but was at times a royal pain.

Thanks so much for quick merge of the (fairly trivial) fix. I recently had similar ones from C++17 bringing us std::data() so flattening the std namespace via using becomes a worse idea by the day ...

[stla]

I implemented an algorithm using the float 128 numbers of BH, and I had to abandon it because it didn't work on Solaris.
I saw there has been a recent update of Eigen, so you have some job ;-)

[eddelbuettel]

Well spotted :) I am more of a user of (Rcpp)Armadillo than (Rcpp)Eigen so that one is more of a pure service, and it is currently 'pending' because I need the various upstreams to act. May need some prodding, hard to say. A much bigger transition is (simple) change to Rcpp (but that reaches to many people): RcppCore/Rcpp#1158