[R-SIG-Mac] Remaining problems compiling rinside_sample0.cpp on Xcode 5.1.1 on mavericks 10.9.4.

[Simon Urbanek]

Roger,

it would be much easier if you actually posted your sample Xcode project. Most of the trying you're trying are either irrelevant to the project (Makevars) are not necessary or don't work

In principle the right way to go about this is to use the R framework - that's how you link Xcode projects to R (see the Mac GUI as an example or the REngine Obj-C R API). You have essentially the choice of either using the R build system or using Xcode's build system. It appears and you are trying to mix pieces from both which won't work.

Cheers,
Simon


On Aug 17, 2014, at 11:11 PM, Roger Dean <Roger.Dean at uws.edu.au> wrote:

> Hi again Colleagues,
> Following my earlier query, I carefully followed (and extensively trialled) the helpful leads you gave. (I had to repost this new query, as my first sending was unsuccessful. Sorry about that.)
> 
> In essence, you confirmed that the build failure was due to my having allowed a conflict to develop between String.h from Rcpp and string.h from the standard C library, so that at some points it was finding the wrong one of these two headers. The cumulated advice on this I gleaned was to avoid duplication of RCpp header paths offered in the search dialogue, and to use -idirafter to try to ensure that the R-associated headers are searched after the path containing the standard libstdc++. I did this, and got rid of the numerous errors that previously appeared early in the linking process. But now I still seem to have a residue of the problem appearing when the Rinside_sample0 main function gets into operation.
> 
> I would very much appreciate a further burst of advice on solving this remaining problem.
> 
> FIRST: here is the problem as shown by the error messages from Xcode:
> Three 'Apple Mach-0 Linker (ld) Errors':
> : "RInside::parseEvalQ(std::string const&)", referenced from:
> _main in rinside_sample0.o
> : "RInside::operator[](std::string const&)", referenced from:
> _main in rinside_sample0.o
> Symbol(s) not found for architecture x86_64
> Linker command failed with exit code 1 (use -v to see invocation)
> 
> And the detail:
> Ld /Users/RogerDean/Library/Developer/Xcode/DerivedData/rRinsideXcodetest2-cfwgtxcrotrppifdcdgmneurclwb/Build/Products/Debug/rRinsideXcodetest2 normal x86_64
>    cd /Users/RogerDean/Desktop/RInside-test/rRinsideXcodetest2
>    export MACOSX_DEPLOYMENT_TARGET=10.9
>    /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang++ -arch x86_64 -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk -L/Users/RogerDean/Library/Developer/Xcode/DerivedData/rRinsideXcodetest2-cfwgtxcrotrppifdcdgmneurclwb/Build/Products/Debug -L/Users/RogerDean/Desktop/RInside-test/rRinsideXcodetest2/rRinsideXcodetest2 -F/Users/RogerDean/Library/Developer/Xcode/DerivedData/rRinsideXcodetest2-cfwgtxcrotrppifdcdgmneurclwb/Build/Products/Debug -filelist /Users/RogerDean/Library/Developer/Xcode/DerivedData/rRinsideXcodetest2-cfwgtxcrotrppifdcdgmneurclwb/Build/Intermediates/rRinsideXcodetest2.build/Debug/rRinsideXcodetest2.build/Objects-normal/x86_64/rRinsideXcodetest2.LinkFileList -mmacosx-version-min=10.9 -stdlib=libstdc++ -lRlapack -lR -lRblas.0 -lRInside -Xlinker -dependency_info -Xlinker /Users/RogerDean/Library/Developer/Xcode/DerivedData/rRinsideXcodetest2-cfwgtxcrotrppifd!
> cdgmneurclwb/Build/Intermediates/rRinsideXcodetest2.build/Debug/rRinsideXcodetest2.build/Objects-normal/x86_64/rRinsideXcodetest2_dependency_info.dat -o /Users/RogerDean/Library/Developer/Xcode/DerivedData/rRinsideXcodetest2-cfwgtxcrotrppifdcdgmneurclwb/Build/Products/Debug/rRinsideXcodetest2
> 
> Undefined symbols for architecture x86_64:
>  "RInside::parseEvalQ(std::string const&)", referenced from:
>      _main in rinside_sample0.o
>  "RInside::operator[](std::string const&)", referenced from:
>      _main in rinside_sample0.o
> ld: symbol(s) not found for architecture x86_64
> clang: error: linker command failed with exit code 1 (use -v to see invocation)
> 
> so it seems still to be an issue to do with string operation as far as I can tell. I have tried a range of the architecture options, and currently have 'Native Architecture (x86-64)' as the choice.
> 
> SECOND, here are the places in Xcode and details of the paths I am currently providing (I have tried a range of alternative sites for placing these paths and directives in the Xcode build settings, that either worsened or did not change the outcome):
> ************
> The 'system' paths:
> //:configuration = Debug
> HEADER_SEARCH_PATHS = /$(inherited) /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/include /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include
> 
> //:configuration = Release
> HEADER_SEARCH_PATHS = /$(inherited) /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/include /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include
> 
> //:completeSettings = some
> HEADER_SEARCH_PATHS
> 
> the 'R' paths:
> I have tried these in "User header search paths" with "Always search user paths" = YES, with similar results. The guidance on Always search user paths seems unclear, and also inconsistent with the idea of the -idirafter directive, so I currently have these paths in "Other C++ Flags" as follows:
> 
> //:configuration = Debug
> OTHER_CPLUSPLUSFLAGS = -idirafter/Library/Frameworks/R.framework/Versions/3.1/Resources/library/RInside/include -idirafter/Library/Frameworks/R.framework/Versions/3.1/Resources/library/Rcpp/include
> -idirafter/Library/Frameworks/R.framework/Versions/3.1/Resources/include
> 
> //:configuration = Release
> OTHER_CPLUSPLUSFLAGS = -idirafter/Library/Frameworks/R.framework/Versions/3.1/Resources/library/RInside/include -idirafter/Library/Frameworks/R.framework/Versions/3.1/Resources/library/Rcpp/include
> -idirafter/Library/Frameworks/R.framework/Versions/3.1/Resources/include
> 
> //:completeSettings = some
> OTHER_CPLUSPLUSFLAGS
> 
> ***********
> Third, there is also the direction to use the particular libstdc++ library, and the restriction of macosx types. These I have currently placed in OTHER_CODE_SIGN_FLAGS:
> 
> //:configuration = Debug
> OTHER_CODE_SIGN_FLAGS = -stdlib=libstdc++ -mmacosx-version-min=10.6
> 
> //:configuration = Release
> OTHER_CODE_SIGN_FLAGS = -stdlib=libstdc++ -mmacosx-version-min=10.6
> 
> //:completeSettings = some
> OTHER_CODE_SIGN_FLAGS
> 
> NB Judging by the detail above, the latter part is getting over-ridden somehow so min=10.9.
> 
> 
> Finally, I'll confirm that I also have set up the .R/Makevars according to the best advice I could again glean from your comments, the related SO posts,  and my searches, dealing particularly with selecting the clang++ compiler. As you see above, I chose to put the -idirafter directives into the Xcode settings rather than in Makevars, which seemed more appropriate since I think these directives to some degree specific to this R project and not general to all R projects. I did also try placing all the -idirafter paths in Makevars, without any improvement, Here is my current .R/Makevars:
> 
> CC=clang
> CXX=clang++
> CFLAGS="-mtune=native -g -O2 -Wall -pedantic -Wconversion"
> CXXFLAGS="-stdlib=libstdc++ -mmacosx-version-min=10.6 -mtune=native -g -O3 -Wall -pedantic -Wconversion"
> FLIBS=-lgfortran
> F77="gfortran-4.2 -arch x86_64"
> FC=$F77
> OBJC=clang
> So I do have some duplication, but as the specs are the same I assume this will function as a safeguard. Again, I have fiddled with most of these permutations.
> 
> A subsidiary question I have is whether the libR.dylib and libRInside.dylib are actually needed in the Xcode project, or whether they will/should be built during the compiling?
> 
> I am hoping another round of advice from you experts might allow me to solve my remaining problem. Many thanks again in advance for your advice and time.
> 
> Roger Dean
> 
> 
> Prof Roger Dean, MARCS Institute, University of Western Sydney
> roger.dean at uws.edu.au<mailto:roger.dean at uws.edu.au>
> 
> 
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>