check _WIN32 rather than WIN32

[kevinushey]

For new Rtools compatibility.

[eddelbuettel]

Thanks mich!

Explain to me again how this did not break before? Because Rtools used to support both?

[jonathon-love]

hi,

this PR isn't enough, it still won't compile with -std=c++11. let me investigate.

looks to me like WIN32 was and is a legit macro, but under -std=c++11 this is no longer the case.

[jonathon-love]

oh, ok, it looks like the issue is that the R headers (in 3.2.3 at least) are still using #ifdef WIN32

line 77 here:
http://docs.rexamine.com/R-devel/RStartup_8h_source.html

are you interested in following this up @eddelbuettel ? brian ripley always hurts my feelings.

[eddelbuettel]

Hm, why are tossing around URLs to a 3rd party site on friggin' GitHub when there is wch's well known mirror?

As you are the one troubled by this, could you possibly try something with all of

• r-release
• r-devel
• whatever R you can use with Rtools using 4.9.* ?

Anything else we are missing, @kevinushey ?

Otherwise I may just revert and close this as Rtools with 4.9.* is not yet operational and we do releases here for CRAN. Which passed. And still pass.

That said, we may as well plan for a world with a real C++ compiler so your probing is appreciated, @jeroenooms !

[kevinushey]

I don't think there's anything else that needs to be done on the RInside side. You could probably just add something like:

#ifdef _WIN32
# define WIN32
#endif

before including the R headers. That said, I think that _WIN32 is the 'canonical' way of declaring Windows, although I forget the nuances of when / if / who should define WIN32 vs. _WIN32 and so on...

[eddelbuettel]

I am pretty I went over this once in the past. This site was always a good resource, and it has _WIN32 as the canonical one. The R 3.2.3 source and the R-devel source appear to have both.

[jonathon-love]

Hm, why are tossing around URLs to a 3rd party site on friggin' GitHub when there is wch's well known mirror?

apparently news hadn't made it to my search engine :P

As you are the one troubled by this, could you possibly try something with all of

looking here, there's still the #ifdef Win32

https://github.com/wch/r-source/blob/trunk/src/include/R_ext/RStartup.h#L35

so i'm pretty sure the issue would persist in r-devel, etc.

[eddelbuettel]

And FWIW what is currently in the master branch here just passed fine on win-builder.

[jonathon-love]

yup, it's only an issue with -std=c++11

[eddelbuettel]

I am not sure what your point is. Do you think we need

#ifndef WIN32
#ifdef _WIN32
#define WIN32
#endif
#endif

It is not our job to make R itself ready for g++ 4.9.* on Windoze. We look after a particular package here, and it builds with the tools I have at my disposal namely win-builder. You need to submit easier to understand, and more clearly argued, requests for changes.

[kevinushey]

Right, I mean that if @jonathon-love wanted to work around the issues with R's headers locally he could use something like

#ifdef _WIN32
# define WIN32
#endif

#include <RInside.h>

I don't think there's any more changes to be made on the RInside side; R likely will need to change something to accommodate the new toolchain and C++11.

[eddelbuettel]

👍

[eddelbuettel]

And C++11 is totally uncharted territory on Windows as we have been stuck at g++ 4.6.* for ages.

There is nothing more to say here, unless I miss something.

[jonathon-love]

yup, i think this is all rounded off. thanks for that suggestion @kevinushey, i'm going to make use of it.

[eddelbuettel]

👍