[gecode-users] Problem with linear constraint over integers using minimodel ("undef ref")

[Sascha Schnug]

Hi,

i'm a new user of Gecode and i have some trouble.

I'm using Gecode 5.0, compiled with the defaults plus gmp & mpfr (although i don't think it's
needed for my case) under Windows using Cygwin (gcc/g++: 5.4). (If i recall correctly;
i needed to manually modify the 3rd-party code where the rounding-mode is selected).

While i made good progress at first (despite the fact that i'm rediscovering C++), the following
part of my code is not working:

        Excerpt:
        --------

        #include <gecode/int.hh>
        #include <gecode/search.hh>
        #include <gecode/minimodel.hh>
        #include <gecode/float.hh>      // for later use

        using Gecode;

        protected:

        IntVarArray q_start;
        IntVarArray q_end;
        IntVarArray q_length;

        public:

        IntVarArgs q_start_;
        IntVarArgs q_end_;
        IntVarArgs q_length_;

        ..........

        for(int i=0; i<n_q; i++)
        {
            q_start_ << IntVar(*this, 0, 95);
            q_end_ << IntVar(*this, 0, 95);
            q_length_ << IntVar(*this, 0, 95);
        }

        q_start = IntVarArray(*this, q_start_);
        q_end = IntVarArray(*this, q_end_);
        q_length = IntVarArray(*this, q_length_);

        ..........

        for(int q=0; q<n_q; ++q)
        {
            // get length of window / THIS SEEMS TO BE THE PROBLEM!
            rel(*this, 1*q_length[q] == 1*q_end[q] - 1*q_start[q]);  // minimodel-like
        }

I think this follows the examples in the guide as well as other resources.

My Makefile looks like this (following https://gist.github.com/fepa/2187179; also uses the order mentioned in guide):

        # Configure this makefile if needed then run
        # $ make all
        # to compile and link

        # What files to compile
        OBJS=problem.cpp solver.cpp
        # What file to compile to
        MAINFILE=solver
        # What compiler to use
        CC=g++
        # Where is your gecode library?
        # (Gecode's default 'make install' behavior installs gecode libraries and headers under /usr/local)
        GECODEDIR=/usr/local

        # 'make all' does both compiling and link (i.e 'make compile link')
        all: compile link

        # 'make compile' compiles files
        compile: $(OBJS)
        		 $(CC) -I$(GECODEDIR)/include -c $(OBJS)

        # 'make link' links compiled files
        link: $(OBJS)
        	  $(CC) -o $(MAINFILE) -L$(GECODEDIR)/lib  $(MAINFILE).o \
                -lgecodeflatzinc -lgecodedriver \
                -lgecodesearch -lgecodeminimodel -lgecodeset \
        		-lgecodefloat \
                -lgecodeint -lgecodekernel -lgecodesupport

        # 'make clean' cleans up generated files from make commands
        clean:
        	rm -rf *o $(MAINFILE)

Now this is the ugly output i receive:

        $ make
        g++ -I/usr/local/include -c solver_v4.cpp -D_GLIBCXX_DEBUG -D_GLIBCXX_DEBUG_PEDANTIC
        g++ -o solver_v4 -L/usr/local/lib  solver_v4.o \
                -lgecodeflatzinc -lgecodedriver -lgecodefloat \
                -lgecodesearch -lgecodeminimodel -lgecodeset \
                -lgecodeint -lgecodekernel -lgecodesupport
        /usr/local/lib/libgecodeminimodel.a(float-expr.o): In function `Gecode::LinFloatExpr::post(Gecode::Home) const':
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:220: undefined reference to `Gecode::Float::Linear::estimate(Gecode::Float::Linear::Term*, int, Gecode::FloatVal, double&, double&)'
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:220:(.text+0x3df3): relocation truncated to fit: R_X86_64_PC32 against undefined symbol `Gecode::Float::Linear::estimate(Gecode::Float::Linear::Term*, int, Gecode::FloatVal, double&, double&)'
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:223: undefined reference to `Gecode::Float::Linear::post(Gecode::Home, Gecode::Float::Linear::Term*, int, Gecode::FloatRelType, Gecode::FloatVal)'
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:223:(.text+0x4029): relocation truncated to fit: R_X86_64_PC32 against undefined symbol `Gecode::Float::Linear::post(Gecode::Home, Gecode::Float::Linear::Term*, int, Gecode::FloatRelType, Gecode::FloatVal)'
        /usr/local/lib/libgecodeminimodel.a(float-expr.o): In function `Gecode::LinFloatExpr::post(Gecode::Home, Gecode::FloatRelType, Gecode::BoolVar const&) const':
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:205: undefined reference to `Gecode::Float::Linear::post(Gecode::Home, Gecode::Float::Linear::Term*, int, Gecode::FloatRelType, Gecode::FloatVal, Gecode::Reify)'
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:205:(.text+0x4277): relocation truncated to fit: R_X86_64_PC32 against undefined symbol `Gecode::Float::Linear::post(Gecode::Home, Gecode::Float::Linear::Term*, int, Gecode::FloatRelType, Gecode::FloatVal, Gecode::Reify)'
        /usr/local/lib/libgecodeminimodel.a(float-expr.o): In function `Gecode::LinFloatExpr::post(Gecode::Home, Gecode::FloatRelType) const':
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:182: undefined reference to `Gecode::Float::Linear::post(Gecode::Home, Gecode::Float::Linear::Term*, int, Gecode::FloatRelType, Gecode::FloatVal)'
        /cygdrive/x/gecode-5.0.0/gecode-5.0.0/gecode/minimodel/float-expr.cpp:182:(.text+0x468d): relocation truncated to fit: R_X86_64_PC32 against undefined symbol `Gecode::Float::Linear::post(Gecode::Home, Gecode::Float::Linear::Term*, int, Gecode::FloatRelType, Gecode::FloatVal)'
        collect2: error: ld returned 1 exit status
        make: *** [Makefile:24: link] Error 1

I'm kinda lost right now as i have no idea what the problem is. I did not check Gecode's sources yet.
My makefile uses some of GCC's STL-debugging tools, but they are not the reason for this.

Another remark:
    In my previous compilation (without gmp + mpfr; added now) i build all the examples too.
        -> all examples using float & minimodel are behaving strange:
            => they all seem to run without errors, but do not produce any output (clean shell)

    I did not research the Script-like usage of gecode a lot, but i expect those examples should
    produce output too like all the others.

Thanks for taking the time to read this! I hope i did not forget to mention relevant information.

-
Sascha-Dominic Schnug