building bio++ on Mavericks

[gsw...@g.uky.edu]

Hello,

I'm trying to build bio++ from source under OSX Mavericks using the instructions from the wiki. I've cloned the git repositories and successfully run cmake to obtain my Makefile, however when I actually try to run make I start running into problems:

$ make
-- Configuring done
-- Generating done
-- Build files have been written to: src/bpp/bpp-core
Scanning dependencies of target bppcore-shared
[ 0%] Building CXX object src/CMakeFiles/bppcore-shared.dir/Bpp/Exceptions.cpp.o
In file included from src/bpp/bpp-core/src/Bpp/Exceptions.cpp:42:
src/bpp/bpp-core/src/Bpp/Exceptions.h:60:17: error: implicit
instantiation of undefined template 'std::__1::basic_string<char,
std::__1::char_traits<char>, std::__1::allocator<char> >'
std::string message_;
            ^

I was able to get past this error by adding an #include <string> to Exceptions.cpp. After that fix I was able to finish compiling bpp-core and got it to install. However, when I moved onto bpp-seq I started running into even more issues during the compilation, this time with the linker complaining about undefined symbols.

I must admit that I haven't built many things using cmake before, so I'm not sure if this is my problem or not. Any ideas would be greatly appreciated. 

Thanks

Grady

[Julien Yann Dutheil]

Dear Grady,

Can you please mention which tool you are using to compile Bio++ (gcc or sthg else?) and the version of the compiler.

As for CMake, it is important to specify where the libs have been installed, for instance:

cmake -DCMAKE_INSTALL_PREFIX=$HOME/.local

using the same location for all libraries.

We would also need the error msg from bpp-seq (at least the first ones).

Best,

Julien.

--
You received this message because you are subscribed to the Google Groups "Bio++ Development Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biopp-devel-fo...@googlegroups.com.
To post to this group, send email to biopp-de...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/279e9825-0960-42e0-ac7b-d5d2be7393d0%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
Julien Y. Dutheil, Ph-D
0 (+49) 6421 178 986

§ Max Planck Institute for Terrestrial Microbiology
Department of Organismic Interactions
Marburg -- GERMANY

§ Intitute of Evolutionary Sciences - Montpellier
University of Montpellier 2 -- FRANCE

[gsw...@g.uky.edu]

Hi Julien,

I am using the default setup with Mavericks, which uses clang as the compiler.

g++ --version
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 5.1 (clang-503.0.38) (based on LLVM 3.4svn)
Target: x86_64-apple-darwin13.1.0
Thread model: posix

I begin by building bpp-core. I followed the procedure described in the "Development Version" section of http://biopp.univ-montp2.fr/wiki/index.php/Installation I used the following call to cmake (after setting up $bpp_dir)

cmake -DCMAKE_INSTALL_PREFIX=$bpp_dir -DCMAKE_LIBRARY_PATH=$bpp_dir/lib -DCMAKE_INCLUDE_PATH=$bpp_dir/include -DBUILD_TESTING=FALSE .

This runs OK, but when I run make I get the error given previously, so something is already going wrong. 

I was able to continue compiling and by including <string> in Exceptions.bpp. I'm not at all confident that this is the best fix, but it did get it to compile.

I then did a make install and moved on to bpp-seq. Same cmake invocation goes OK and finds the core files fine. When I run make the linker errors out with:

Linking CXX shared library libbpp-seq.dylib
Undefined symbols for architecture x86_64:
  "bpp::LetterAlphabet::LETTER_UNDEF_VALUE", referenced from:
      bpp::DefaultAlphabet::DefaultAlphabet() in DefaultAlphabet.cpp.o
      bpp::CaseMaskedAlphabet::CaseMaskedAlphabet(bpp::LetterAlphabet const*) in CaseMaskedAlphabet.cpp.o
      bpp::NucleicAlphabet::NucleicAlphabet() in DNA.cpp.o
      bpp::ProteicAlphabet::ProteicAlphabet() in ProteicAlphabet.cpp.o
      bpp::NucleicAlphabet::NucleicAlphabet() in RNA.cpp.o
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)

[gsw...@g.uky.edu]

I believe that I may have found the real problem: some sort of compatibility issue between libc++ and libstdc++. 

I added -stdlib=libstdc++ to CMAKE_CXX_FLAGS in both bpp-core/src and bpp-seq/src and things seem to be working now.

[gsw...@g.uky.edu]

I seem to have been able to convince bppsuite+dependencies to build under clang.

I had to change all the relevant CMakeLists files to the following:

SET(CMAKE_CXX_FLAGS "-Wall -Weffc++ -Wshadow -Wconversion -stdlib=libstdc++")

This let me make install everything succesfully. I ended up with bin/ lib/ include/ in my $bpp_dir, which seem to be properly populated. 

However when I tried to run any of the executables in bin/ I now get the following (for example)

dyld: Library not loaded: libbpp-phyl.9.dylib
Referenced from: ./bppseqgen
Reason: image not found
Trace/BPT trap: 5

In order to get the executables to run, I had to create simlinks in bin/ to the files in lib/

This seems highly hacky, but it does work.

I haven't tried anything more exotic than getting the programs to print out their welcome messages, but at least I have some binaries now.

[Julien Yann Dutheil]

Dear Grady,

Thanks a lot for the libstdc++ trick, this will certainly be useful to other users!

As for the execution issue, you can set the environment variable

DYLIB_LIBRARY_PATH=$bpp_dir/lib

this should work (instead of making links in the bin dir).

Best regards,

Julien.

--
You received this message because you are subscribed to the Google Groups "Bio++ Development Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biopp-devel-fo...@googlegroups.com.
To post to this group, send email to biopp-de...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/09d44cd5-1b0f-4fac-a9da-0a3a5b7052ee%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[gsw...@g.uky.edu]

Thanks Julien,

I'll give the path variable a try later.

Is there an easy way to get CMake to build the binaries with static linkage?

[Julien Yann Dutheil]

Hi,

You can enable the BUILD_STATIC  (either using the cmake interface, or passing -DBUILD_STATIC=ON ).

Best,

Julien.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/507a1155-255d-4907-b016-4f8a660b1220%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[gsw...@g.uky.edu]

I spent a bit more time this evening looking into the cause of the error in compiling bpp-seq that I ran into and this seems to be a minimal example of the problematic code.

#include <vector>

class Foo{

  static const int bar = -99;

public:

  std::vector<int> bazv;

  Foo():bazv(10,bar) {}

};

int main(){

  Foo f;

  return 0;

}

This code will compile with libstdc++, but fails with libc++, with an error similar to the one I described previously. Interestingly, if you change the constructor to 

Foo():bazv(bar) {}

Then it will compile under either library (although it gives a runtime error due to the -99).

[gsw...@g.uky.edu]

I managed to confound several colleagues with this, but one of them was eventually able to put me on the right path. Apparently, the definition of static members should be moved out-of-class, rather than specified in the declaration.

So, LetterAlphabet.h gets modified to contain

      static const int LETTER_UNDEF_VALUE;

And I created a new file LetterAlphabet.cpp with

const int bpp::LetterAlphabet::LETTER_UNDEF_VALUE = -99;

Now It looks like I can build everything completely natively.

--

Grady

[Julien Yann Dutheil]

Dear Grady,

That is absolutely correct, and I am very surprised we had that mistake in the libs... compiler should have complained about that!

I have included a fix (the one you found yourself). It will be part of the next release.

Many thanks!

Julien.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/92fa8f5d-5eb8-40fc-964a-7e9e65309f57%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[gsw...@g.uky.edu]

Hi Julien,

Great! Don't forget that I also had to include <string> in one of the bpp-core files, as I mentioned previously.

If you would like me to test anything else on Mavericks before the next release feel free to send me an email.

--

Grady Weyenberg

Department of Statistics

University of Kentucky, USA

[Mathieu Groussin]

Dear Julien, Grady and all,

I have similar problems when compiling bpp-core on a mac os Mavericks... After reading this discussion, I did find a solution to my problems...

I first got the first error that Grady obtained. Indeed, I solved the issue by including #include <string> in Exceptions.cpp. But contrary to Grady, it seems that this problem concerns a lot of classes. The next problematic class is with BppApplications.cpp and then, after including #include <string> in BppApplications.cpp, it goes on with BppODiscreteDistributionFormat.cpp etc...

Additionally, I got a very long list of warnings... I attached to this email an exemple of errors (with BppApplications.cpp) and warnings I've just obtained.

I downloaded the dev libraries today with git, the g++ version is 4.2.1, the cmake version is 2.8.12.2.

If you have any ideas, I'd really appreciate (warnings and errors...)!

Thanks,

Mathieu

[Julien Yann Dutheil]

Hi Mathieu,

Thanks a lot, will give it a look.

J.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/9ba242f8-c8fa-4570-bbe2-2990d8821185%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Julien Yann Dutheil]

Matthieu,

I fixed the first warnings (I hope). Can you give it a try and send me the new list of warnings (privately) ?

Cheers,

J.

On Mon, Apr 28, 2014 at 9:20 PM, Mathieu Groussin <mgro...@gmail.com> wrote:

To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/9ba242f8-c8fa-4570-bbe2-2990d8821185%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Julien Yann Dutheil]

Ok, done pour bpp-core, et c'est parti pour un premier jet sur bpp-seq...

J.

On Mon, Apr 28, 2014 at 9:20 PM, Mathieu Groussin <mgro...@gmail.com> wrote:

To view this discussion on the web visit https://groups.google.com/d/msgid/biopp-devel-forum/9ba242f8-c8fa-4570-bbe2-2990d8821185%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Julien Yann Dutheil]

Oups, my mistake, stroke the wrong reply button :D We're still removing those Maverick warnings...

J.