I had the same issue (RNGScope) and what worked for me (besides Dirk's clear explanations which I somehow missed) was to run Rcpp.package.skeleton() and compare what changed, which is exactly what Dirk mentioned above (the changes to NAMESPACE and DESCRIPTION). LinkingTo is still necessary (together with Imports) I believe (at least in my case).
Avraham On Wed, Feb 5, 2014 at 2:37 PM, Dan Tenenbaum <dtene...@fhcrc.org> wrote: > > > ----- Original Message ----- > > From: "Dirk Eddelbuettel" <e...@debian.org> > > To: "Dan Tenenbaum" <dtene...@fhcrc.org> > > Cc: rcpp-devel@lists.r-forge.r-project.org > > Sent: Wednesday, February 5, 2014 11:02:10 AM > > Subject: Re: [Rcpp-devel] dataptr / char_get_string_elt not provided by > package 'Rcpp' > > > > > > On 5 February 2014 at 10:45, Dan Tenenbaum wrote: > > | A couple of Bioconductor packages have these errors: > > | > > | package: DOSE > > | > > | error: > > | > > | function 'char_get_string_elt' not provided by package 'Rcpp' > > | Execution halted > > | > > | package: GOSim > > | > > | error: > > | > > | function 'dataptr' not provided by package 'Rcpp' > > > > See below. > > > > > > | I see in > > | https://github.com/RcppCore/Rcpp/issues/92 > > > > That ticket is __closed__ so why do you look there? > > > > Logs for Rcpp testing are at https://github.com/RcppCore/rcpp-logs > > Poke around there, particular in the status/ directory > > > > All that was communicated here. > > > > | Does this indicate the issue has not been resolved? Or is there a > > | workaround? > > > > In most cases, yes. As shown in the repo and here. > > > > To sum up again: > > > > i) Either one of > > > > function 'enterRNGScope' not provided by package 'Rcpp' > > > > function 'char_get_string_elt' not provided by package > > 'Rcpp' > > > > can be fixed by strengthening the Rcpp initialization. > > > > ii) For that I provided patches doing this, noted it in the > > announcements > > and release notes. > > > > You now need > > > > DESCRIPTION: Imports: Rcpp instead of > > Depends: > > > > NAMESPACE: importFrom(Rcpp, evalCpp) > > > > Sometimes a simple import(Rcpp) works, the importFrom(Rcpp, > > evalRcpp) > > is better. It can use any symbol, evalCpp is nice and short. > > > > That should do. The odd thing is that packages sometimes pass (which > > is why > > the defect rate was down to < 10 when we released -- count based on > > my > > machine) but fail on other machines. > > > > In any case, the fix for i) above is to do what is in ii) > > > > Ok? > > > > In your particular case, you may also consider picking up the phone > > and > > making an in-house call to Kevin who tested all of BioC prior to > > release as > > well. And that too was noted in my release announcement. > > > > Indeed, you guys have done a really good job documenting all your work. > You have to forgive me, as I was only told about all these changes > recently, and I did not follow all the communication as it was happening, > but had to come upon it later where it wasn't always clear which document > was the canonical one to read. > > Anyway, your suggestion did the trick for both packages. Thanks! > Dan > > > > Dirk > > > > -- > > Dirk Eddelbuettel | e...@debian.org | http://dirk.eddelbuettel.com > > > _______________________________________________ > Rcpp-devel mailing list > Rcpp-devel@lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel >
_______________________________________________ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
