Runtime error with Mac 10.6.5

[Generic User (don't modify these fields)]

I was able to install SHERPA 1.2.3 just fine on Mac 10.6.5. However at runtime I get the following error when linking with the interaction model library.

Erik-Gerwicks-MacBook-Pro:LEP91 erikgerwick$ ~/computational/SHERPA-MC-1.2.3/bin/Sherpa Welcome to Sherpa, Erik Gerwick. Initialization of framework underway. Run_Parameter::Init(): Setting memory limit to 1.90234 GB. Sherpa: Model_Base::InitializeInteractionModel throws not implemented:

Interaction model not implemented

Exception_Handler::Terminate(): Pre-crash status saved to '/Users/erikgerwick/computational/SHERPA-MC-1.2.3/SHERPA/Run/LEP91/StatusFri_Dec_17_11-06-34_2010'. Exception_Handler::Exit: Exiting Sherpa with code (1) Time: 0s on Fri Dec 17 11:06:34 2010

(User: 0s, System: 0s, Children User: 0s, Children System: 0s)

---

Please cite the publications listed in 'Sherpa_References.tex'.

Extract the bibtex list by running 'get_bibtex Sherpa_References.tex' or email the file to ​'slaclib2@slac.stanford.edu', subject 'generate'.

---

Return_Value::PrintStatistics(): Statistics {

Generated events: 0

}

In addition when I run with the SHOW_MODEL_SYNTAX=1 it finds the models correctly, in particular the default SM. The install and config files are attached.

[Generic User (don't modify these fields)]

I confirm this problem for both Sherpa 1.2.3 and 1.2.2.

My system is MacOS 10.6.6 with Xcode 3.2.3.

The c++ compiler is gcc 4.2.1 (Apple Inc. build 5664).

The fortran compiler is g77 (binaries can by found at hpc.sourceforge.net, the binaries I use g77-intel-bin.tar.gz).

Note that the macos 10.6.x the default output of gcc is 64 bits, while the default output of g77 is 32 bits. None of the -m32 and -m64 options are specified in configure; therefore if I simply run "./configure" and "make" the compilation fails to link the libraries because of incompatible architectures. The work around I found is to use:

configure LDFLAGS="-m32" CPPFLAGS="-m32" CXXFLAGS="-m32"

In this way I force gcc to produce 32 bits output and there are no error messages during make and make install.

Unfortunately when I run the Sherpa executable I find the very same error message that is reported in the original message of this thread:

Initialize the Standard Model from / Model.dat Sherpa: Model_Base::InitializeInteractionModel throws not implemented:

Interaction model not implemented

Exception_Handler::Terminate(): Pre-crash status saved to '/Users/lunghi/SHERPA-MC-1.2.2/SHERPA/Run/sm/StatusMon_Jan_17_22-54-52_2011'. Exception_Handler::Exit: Exiting Sherpa with code (1)

This problem should be very easy to reproduce because I have experienced it on a fresh snow leopard install (i.e. no messing around with the system, just xcode and g77).

Thank you very much for any help you can provide.

Enrico Lunghi (​elunghi@indiana.edu)

[Generic User (don't modify these fields)]

(I forgot to add my email in the Cc field).

[hoeth]

Can you try this:

./configure CXX=/usr/local/bin/g++ FC=/usr/local/bin/gfortran CC=/usr/local/bin/gcc

If you don't have g++/gcc in /usr/local/bin you'll need to download the complete package from HPC rather than just the gfortran.

[Generic User (don't modify these fields)]

I installed gcc, g++ and gfortran from HPC (​http://prdownloads.sourceforge.net/hpc/gcc-snwleo-intel-bin.tar.gz) and run ./configure as you suggested.

This time the compilation stops (rather early) with an error. I get the same behavior in both 1.2.2 and 1.2.3.

I attach the Terminal output of configure, make and of the gcc and gfortran versions.

Replying to hoeth:

Can you try this:

./configure CXX=/usr/local/bin/g++ FC=/usr/local/bin/gfortran CC=/usr/local/bin/gcc

If you don't have g++/gcc in /usr/local/bin you'll need to download the complete package from HPC rather than just the gfortran.

[Generic User (don't modify these fields)]

Terminal output for configure, make, /usr/local/bin/gcc - v and /usr/local/bin/gfortran -v

[hoeth]

Great, you now have at least a consistent set of g++ and gfortran, even though it's both development versions. Now to your new problem. Can you #include <stddef.h> in ATOOLS/Math/Random.H and run make again?

[Generic User (don't modify these fields)]

compilation still ends with an error (I attach the complete Terminal output but here is the error):

/usr/local/bin/g++ -g -O2 -pedantic -Wall -o .libs/Combine_Analysis -rdynamic Combine_Analysis.o -Wl,-bind_at_load -L/Users/lunghi/SHERPA-MC-1.2.2/ATOOLS/Phys -L/Users/lunghi/SHERPA-MC-1.2.2/ATOOLS/Math -L/Users/lunghi/SHERPA-MC-1.2.2/ATOOLS/Org /Users/lunghi/SHERPA-MC-1.2.2/ATOOLS/Phys/.libs/libToolsPhys.dylib /Users/lunghi/SHERPA-MC-1.2.2/ATOOLS/Math/.libs/libToolsMath.dylib /Users/lunghi/SHERPA-MC-1.2.2/ATOOLS/Org/.libs/libToolsOrg.dylib -lm -ldl g++: error: unrecognized option '-rdynamic' make[4]: * [Combine_Analysis] Error 1 make[3]: * [all-recursive] Error 1 make[2]: * [all-recursive] Error 1 make[1]: * [all-recursive] Error 1 make: * [all] Error 2

[Generic User (don't modify these fields)]

configure, make (with modified random.H file)

[hoeth]

Google tells me: ​http://lists.apple.com/archives/unix-porting/2005/Feb/msg00025.html So drop that option and see where you get. Maybe you can try to get Sherpa to compile and tell us what changes you needed? I don't have a MacOS 10.6.x available for testing, so for me this currently is all remote guessing and googling, too.

[Generic User (don't modify these fields)]

I removed all reference to the -rdynamic flag and the compilation still ends with an error at the very end (when trying to create libSherpaMain.la and just before creating the executable Sherpa). The error message is

ld: duplicate symbol dyld_stub_binding_helper in /usr/lib/crt1.10.5.o and /usr/lib/dylib1.10.5.o

collect2: ld returned 1 exit status

make[3]: * [libSherpaMain.la] Error 1

I'm not very experienced with c++, do you have an idea of where this come from? I googled around but I didn't find anything useful.

[Generic User (don't modify these fields)]

configure and make for Sherpa 1.2.2 (any reference to -rdynamic have been removed)

[hoeth]

My first guess is that gfortran shouldn't pull crt1.o in explicitly, but it does anyhow. I'll have a closer look tomorrow.

[hoeth]

Two things to try:

Can you remove the explicit -lcrt1.10.5.o from the command that does the final linking (the one that fails) and see if that helps?

If this helps, can you try the tarball from ​http://users.hepforge.org/~hoeth/SHERPA-MC-1.2.3-libtool24.tar.bz2 and check if you run into the same trouble with that one?

[Generic User (don't modify these fields)]

I have run into the same error and followed all of the steps on this thread (CXX etc. configure flags, edited Random.h, removed -rdynamic, used new tarball). I have no problem compiling or linking with this setup, but I still get the "Interaction Model not implemented" error at runtime. As with the others, I am using a fresh snow-leopard install w/ xcode & gcc from HPC.

Adam Martin (​aomartin@fnal.gov)

[Generic User (don't modify these fields)]

I just attached the Shell output of configure, make, make install and of a test run of Sherpa (version 1.2.3). The error message is that "Interaction Model not Implemented". I get the same behavior for both 1.2.3 and 1.2.3-libtool24.

I didn't manage to get Sherpa 1.2.2 to compile with the HPC compilers. Got stuck with the crt1.10.5.o issue.

Incidentally, I can get Sherpa (1.2.2 and 1.2.3) to compile using the native gcc from xcode and g77 from HPC but I have to add the -m32 flag to the c and c++ compilers. Again I get the same runtime error.

[Generic User (don't modify these fields)]

Output of configure, make, make install and ../Sherpa for v 1.2.3

[hoeth]

We got a Snowleopard machine as test box now, so I will be able to look into this myself. I'll let you know what I find out.

[Stefan Hoeche]

Sherpa trunk (1.3.0-to-be) compiles and runs fine on the following configuration:

System Software Overview:

System Version: Mac OS X Server 10.6.6 (10J567) Server Configuration: Advanced Kernel Version: Darwin 10.6.0 Secure Virtual Memory: Not Enabled 64-bit Kernel and Extensions: Yes

Hardware Overview:

Model Name: Mac Pro Processor Name: 6-Core Intel Xeon Processor Speed: 2.66 GHz

Hendrik, can you check whether it works on your local machine too?

The original problem, causing the message 'Interaction model not implemented' was fixed on trunk r17250. This error only occured on Macs.

[Stefan Hoeche]

For version 1.2.3, please use the attached patch. Apply via 'patch -p0 < model_init_2.patch' in the top-level directory of your Sherpa installation and recompile using 'make install'. If the problem persists, please reopen this ticket.

[Generic User (don't modify these fields)]

Starting from an installation of version 1.2.3 where I get the "Interaction Model not Found" error at runtime, I used the patch provided but still find problems. The problem starts when I use the patch command. I get

Adam-Martins-MacBook-Pro:SHERPA-MC-1.2.3 aomartin$ patch -p0 < model_init_2.patch patching file MODEL/Main/Model_Base_GF.C patching file MODEL/Main/Makefile.am Hunk #1 FAILED at 4. 1 out of 1 hunk FAILED -- saving rejects to file MODEL/Main/Makefile.am.rej patching file MODEL/Main/Model_Base.C

that looked bad, but I remade, and tried to run. Instead of the "Interaction model" error, I now get

Adam-Martins-MacBook-Pro:EGamma aomartin$ ../../bin/Sherpa dyld: lazy symbol binding failed: Symbol not found: ZN6ATOOLS15Getter_FunctionIN5MODEL10Model_BaseENS1_15Model_ArgumentsESt4lessISsEEC2ERKSs

Referenced from: /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libModelMain.0.dylib Expected in: flat namespace

dyld: Symbol not found: ZN6ATOOLS15Getter_FunctionIN5MODEL10Model_BaseENS1_15Model_ArgumentsESt4lessISsEEC2ERKSs

Referenced from: /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libModelMain.0.dylib Expected in: flat namespace

[Stefan Hoeche]

Can you make the following change in MODEL/Main/Makefile.am:

Coupling_Data.C Model_Base.C \ 

becomes

Coupling_Data.C Model_Base.C Model_Base_GF.C \ 

[Stefan Hoeche]

Could you please also apply the patch file msg_include_1.patch listed under ticket #135 and give us feedback whether this patch combined with the above one solved your problem? Thanks

[Stefan Hoeche]

Closing this ticket for now, please reopen if necessary

[Generic User (don't modify these fields)]

I edited MODEL/Main/Makefile as well as applying the patch msg_include_1.patch, but I still find the error message

"Hunk #1 FAILED at 4" etc.

when I try applying the model_init_2 patch. Upon running Sherpa, the error message I get is now:

dyld: Symbol not found: ZN6ATOOLS3msgE

Referenced from: /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libSherpaMain.0.dylib Expected in: /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libToolsMath.0.dylib

in /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libSherpaMain.0.dylib

Trace/BPT trap

[Stefan Hoeche]

You should edit Makefile.am, not the Makefile itself

[Generic User (don't modify these fields)]

Sorry, I meant to write that I edited Makefile.am, rather than the Makefile, and I get the above errors

[Stefan Hoeche]

Could you try and use the attached model_init_3.patch instead of model_init_2.patch? Please do not modify the Makefile.am in this case and make sure that MODEL/Main/Model_Base_GF.C does not exist before you apply the patch. Then run 'make clean' in the top level and recompile/install using 'make install'.

[Generic User (don't modify these fields)]

I repeated all the steps above with a fresh download, including the msg_.. patch and the model_init_3 patch. I find no complaints when I install the patches, but at run-time I find:

dyld: lazy symbol binding failed: Symbol not found: ZN6ATOOLS15Getter_FunctionIN5MODEL10Model_BaseENS1_15Model_ArgumentsESt4lessISsEEC2ERKSs

Referenced from: /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libModelMain.0.dylib Expected in: flat namespace

dyld: Symbol not found: ZN6ATOOLS15Getter_FunctionIN5MODEL10Model_BaseENS1_15Model_ArgumentsESt4lessISsEEC2ERKSs

Referenced from: /Users/aomartin/Generators/SHERPA-MC-1.2.3/lib/SHERPA-MC/libModelMain.0.dylib Expected in: flat namespace

[Stefan Hoeche]

This sounds like MODEL/Main/Model_Base_GF.C was not actually compiled into the library. Can you please try

  touch MODEL/Main/Model_Base_GF.C
  cd MODEL/Main
  make

and paste the output here? Thanks

[Frank Siegert]

Is it possible that the autotools (autoconf, automake, libtool) don't exist on your system? In that case the modifications to Makefile.am will get ignored and the file won't be compiled.

[Generic User (don't modify these fields)]

I checked the autotools and reinstalled new versions (autoconf 2.65, automake 1.11, libtool 2.2.6b) & retried everything, getting the same ZN6ATOOLS... error.

going to MODEL/Main and typing make, gets me 'nothing to be done for all'

[hoeth]

I installed the current trunk without any issues on 10.6.0 and can run the LEP91 example. This will be released in the next few days as 1.3.0, so I'll close this ticket. Please reopen if you still have trouble with 1.3.0.

[Generic User (don't modify these fields)]

Could you post the versions of gcc and of the fortran compiler that you are using? I use the HPC versions (see the version output below) and still get the -rdynamic error.

[~]$ /usr/local/bin/gcc -v Using built-in specs. COLLECT_GCC=/usr/local/bin/gcc COLLECT_LTO_WRAPPER=/usr/local/libexec/gcc/x86_64-apple-darwin10.4.0/4.6.0/lto-wrapper Target: x86_64-apple-darwin10.4.0 Configured with: ../gcc-4.6-20101106/configure --enable-languages=c++,fortran Thread model: posix gcc version 4.6.0 20101106 (experimental) (GCC)

[~]$ /usr/local/bin/gfortran -v Using built-in specs. COLLECT_GCC=/usr/local/bin/gfortran COLLECT_LTO_WRAPPER=/usr/local/libexec/gcc/x86_64-apple-darwin10.4.0/4.6.0/lto-wrapper Target: x86_64-apple-darwin10.4.0 Configured with: ../gcc-4.6-20101106/configure --enable-languages=c++,fortran Thread model: posix gcc version 4.6.0 20101106 (experimental) (GCC)

[hoeth]

I used the "gcc" package from fink. This is a complete gcc-4.5.0 installation (including also gfortran).

[Generic User (don't modify these fields)]

I think the problem is solve. Now Sherpa 1.3 compiles and runs fine. The detailed installation instructions are:

1) Install the latest xcode (this is important to get /usr/bin/ranlib as a link to /usr/bin/libtool)

2) Install the HPC compilers (gcc-snwleo-intel-bin.tar.gz) in /usr/local (i.e. just untar in / )

3) Install m4-1.4.14, autoconf-2.65, automake-1.11.1 and libtool-2.2.6b in a directory that comes before /usr and /usr/local in the path - I have them in /sw (it is crucial not to install the new libtool in /usr! If you replace the xcode libtool with the new one you'll end up with errors). To do this follow the instructions at ​http://www.mattvsworld.com/blog/2010/02/install-the-latest-autoconf-and-automake-on-mac-os-10-6/ In addition to these instructions, make sure that you are downloading the version of the programs I listed, configure each program using ./configure --prefix=/sw CC=/usr/local/bin/gcc FC=/usr/local/bin/gfortran CXX=/usr/local/bin/g++

4) In Sherpa-1.3.0 remove everywhere the string -rdynamic . It appears in:

AddOns/Analysis/Scripts/Makefile.am AddOns/Analysis/Scripts/Makefile.in acinclude.m4 aclocal.m4 configure It is because of this step that you have to upgrade the various autotools.

5) ./configure CC=/usr/local/bin/gcc FC=/usr/local/bin/gfortran CXX=/usr/local/bin/g++ --enable-analysis

(you can add your options)

6) make

7) make install

[Stefan Hoeche]

Milestone rel-1.3.0 deleted