[Rcpp-devel] Defining template specialisation for wrap on Windows 64bit (Compilation error: 'result_type' does not name a type)

[Renaud Gaujoux]

The change/commit that fixes the error in the header file is this one:
https://github.com/renozao/RcppOctave/commit/59375a038e5abeeb7fceb01ab7ffbc98cfe8cccf

>From the diff, it feels like more stuff changed, because I also moved some
irrelevant-to-this-issue macro declarations but this is really only about
not including the Octave headers before including Rcpp.h, and substituting
them by a forward declaration before declaring the specialisation. Octave
headers are then included _after_ Rcpp.h.

I will try the (x)emacs/cygwin trick. Cmd.exe is simply not reliable and
frustrating. Even running R from cmd.exe (not Rgui) breaks horribly with
i/o issues. However, isn't there any conflict between plain cygwin and the
files installed by Rtools that are to be used when building/checking
packages?

Renaud



On 4 October 2013 20:05, Steve Jaffe <sjaffe at riskspan.com> wrote:

> I don't think the "typename" is necessary ... As I said, I could see no
> reason for needing it, it was just something to try in case I was
> overlooking something. Now that Renaud has found a real solution I think we
> can forget about it (in fact, based on my understand of the rules, I think
> adding the "typename" would itself be an error :-)
>
> Steve
>
> -----Original Message-----
> From: Dirk Eddelbuettel [mailto:edd at debian.org]
> Sent: Friday, October 04, 2013 12:12 PM
> To: Renaud Gaujoux
> Cc: Steve Jaffe; rcpp-devel at lists.r-forge.r-project.org
> Subject: Re: [Rcpp-devel] Defining template specialisation for wrap on
> Windows 64bit (Compilation error: 'result_type' does not name a type)
>
>
> On 4 October 2013 at 17:45, Renaud Gaujoux wrote:
> | I posted the follow up email adding [SOLVED] to the subject, but I
> thought it
> | would appear in the same thread. Here it is:
> |
> | http://lists.r-forge.r-project.org/pipermail/rcpp-devel/2013-October/
> | 006555.html
> |
> | Indeed, one of the suggested solution out there was to add "typename" in
> the
> | typdef statement, but I don't control that piece of code, which is in
> Rcpp
> | itself. Maybe Dirk and Romain can look into the implications of that for
> the
> | next release?
>
> Could you possibly provide a diff output (or 'svn diff' output) to make it
> more clear exactly what changes where you are talking about?
>
> | I eventually managed to get RcppOctave working on Windows!!! The library
> issue
> | got solved by using proper gcc linking flags, making the libraries nicely
> | dynamically loaded.
>
> Very awesome.
>
> | I want to start another thread on the Rcp and RcppOctave list, to ask
> willing
> | windows users to actually test the binary package I built on R-3.0.2.
>
> Yup.
>
> | I see strange things happening when building/checking the package in the
> crappy
> | cmd.exe terminal on my VirtualBox Windows7 (which makes me even happier
> to run
> | Ubuntu ;) ), so I'd rather ask plain R/Windows users a hand.
>
> I have / once had WinXP inside a kvm virtualization, which was even less
> of a
> resource hog, but it looks like I busted that one -- last time I tried R
> and
> compilations it didn't quite.  Need a new machine with more ram...
>
> Dirk
>
> --
> Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20131004/4c992738/attachment.html>