Description: Upstream changes to support big-endian architectures
 This patch consists of upstream revisions 2071, 2079, 2080 and 2140.
Author: Greg Landrum
Origin: upstream
Bug-Debian: http://bugs.debian.org/670054
Last-Update: 2012-07-28

--- rdkit-201203.orig/CMakeLists.txt
+++ rdkit-201203/CMakeLists.txt
@@ -26,6 +26,9 @@ option(RDK_INSTALL_STATIC_LIBS "install
 option(RDK_BUILD_SLN_SUPPORT "include support for the SLN format" ON )
 option(RDK_TEST_MULTITHREADED "run some tests of multithreading" OFF )
 
+include(TestBigEndian)
+TEST_BIG_ENDIAN(RDK_BIG_ENDIAN)
+
 # At build time put runtime binaries in the bin subdirectory
 set(RDK_RUNTIME_OUTPUT_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/bin")
 set(RDK_LIBRARY_OUTPUT_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/lib")
--- rdkit-201203.orig/Code/CMakeLists.txt
+++ rdkit-201203/Code/CMakeLists.txt
@@ -15,7 +15,11 @@ add_subdirectory(GraphMol)
 add_subdirectory(Query)
 
 add_subdirectory(DataManip)
-add_subdirectory(SimDivPickers)
+if(RDK_BIG_ENDIAN)
+  message("Skipping build of SimDivPickers on big endian system")
+else(RDK_BIG_ENDIAN)
+  add_subdirectory(SimDivPickers)
+endif(RDK_BIG_ENDIAN)
 
 add_subdirectory(ML)
 add_subdirectory(ChemicalFeatures)
--- rdkit-201203.orig/Code/DataStructs/BitVect.cpp
+++ rdkit-201203/Code/DataStructs/BitVect.cpp
@@ -40,7 +40,7 @@ void BitVect::initFromText(const char *d
   
   // earlier versions of the code did not have the version number encoded, so
   //  we'll use that to distinguish version 0
-  ss.read((char *)&size,sizeof(size));
+  RDKit::streamRead(ss,size);
   if(size<0){
     version = -1*size;
     if (version == 16) {
@@ -52,12 +52,12 @@ void BitVect::initFromText(const char *d
     else {
       throw ValueErrorException("bad version in BitVect pickle");
     }
-
-    ss.read((char *)&size,sizeof(size));
+    RDKit::streamRead(ss,size);
   } else if( !allowOldFormat ) {
     throw ValueErrorException("invalid BitVect pickle");
   }
-  ss.read((char *)&nOn,sizeof(nOn));
+
+  RDKit::streamRead(ss,nOn);
   _initForSize(static_cast<int>(size));
 
   // if the either have older version or or version 16 with ints for on bits
@@ -65,13 +65,13 @@ void BitVect::initFromText(const char *d
       ( (format == 1) && (size >= std::numeric_limits<unsigned short>::max()) ) ) {
     boost::uint32_t tmp;
     for(unsigned int i=0; i<nOn; i++){
-      ss.read((char *)&tmp,sizeof(tmp));
+      RDKit::streamRead(ss,tmp);
       setBit(tmp);
     }
   } else if (format == 1) { // version 16 and on bits stored as short ints
     boost::uint16_t tmp;
     for(unsigned int i=0; i<nOn; i++){
-      ss.read((char *)&tmp,sizeof(tmp));
+      RDKit::streamRead(ss,tmp);
       setBit(tmp);
     }
   } else if (format == 2) { // run length encoded format
--- rdkit-201203.orig/Code/DataStructs/BitOps.cpp
+++ rdkit-201203/Code/DataStructs/BitOps.cpp
@@ -21,18 +21,20 @@
 
 #include <boost/lexical_cast.hpp>
 
+using namespace RDKit;
+
 int getBitId(const char *&text,int format,int size,int curr){
   PRECONDITION(text,"no text");
   int res=-1;
   if( (format==0) || 
       ( (format == 1) && (size >= std::numeric_limits<unsigned short>::max()) ) ) {
     int tmp;
-    tmp = *(int *)text;
+    tmp = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)text);
     text += sizeof(tmp);
     res=tmp;
   } else if (format == 1) { // version 16 and on bits sotred as short ints
     unsigned short tmp;
-    tmp = *(unsigned short *)text;
+    tmp = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(unsigned short *)text);
     text += sizeof(tmp);
     res=tmp;
   } else if (format == 2) { // run length encoded format
@@ -54,7 +56,7 @@ bool AllProbeBitsMatch(const char *probe
   int refFormat=0;
   int version=0;
 
-  int probeSize = *(int *)probe;
+  int probeSize = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)probe);
   probe+=sizeof(probeSize);
   if(probeSize<0){
     version = -1*probeSize;
@@ -67,11 +69,11 @@ bool AllProbeBitsMatch(const char *probe
     else {
       throw("Unknown version type for the encode bit vect");
     }
-    probeSize = *(int *)probe;
+    probeSize = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)probe);
     probe+=sizeof(probeSize);
   }
 
-  int refSize = *(int *)ref;
+  int refSize = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)ref);
   ref+=sizeof(refSize);
   if(refSize<0){
     version = -1*refSize;
@@ -84,14 +86,14 @@ bool AllProbeBitsMatch(const char *probe
     else {
       throw("Unknown version type for the encode bit vect");
     }
-    refSize = *(int *)ref;
+    refSize = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)ref);
     ref+=sizeof(refSize);
   }
 
 
-  int nProbeOn = *(int *)probe;
+  int nProbeOn = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)probe);
   probe+=sizeof(nProbeOn);
-  int nRefOn = *(int *)ref;
+  int nRefOn = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)ref);
   ref+=sizeof(nRefOn);
 
   int currProbeBit=0;
@@ -121,7 +123,7 @@ bool AllProbeBitsMatch(const T1 &probe,c
   const char *text=pkl.c_str();
   int format=0;
   int nOn=0,size,version=0;
-  size = *(int *)text;
+  size = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)text);
   text+=sizeof(size);
   if(size<0){
     version = -1*size;
@@ -134,10 +136,10 @@ bool AllProbeBitsMatch(const T1 &probe,c
     else {
       throw("Unknown version type for the encode bit vect");
     }
-    size = *(int *)text;
+    size = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)text);
     text+=sizeof(size);
   }
-  nOn = *(int *)text;
+  nOn = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(*(int *)text);
   text+=sizeof(nOn);
 
   int currBit=0;
--- rdkit-201203.orig/Code/DataStructs/DiscreteValueVect.cpp
+++ rdkit-201203/Code/DataStructs/DiscreteValueVect.cpp
@@ -1,6 +1,6 @@
 // $Id: DiscreteValueVect.cpp 1528 2010-09-26 17:04:37Z glandrum $
 //
-//  Copyright (C) 2004-2008 Greg Landrum and Rational Discovery LLC
+//  Copyright (C) 2004-2012 Greg Landrum and Rational Discovery LLC
 //
 //  @@ All Rights Reserved @@
 //  This file is part of the RDKit.
@@ -143,8 +143,15 @@ namespace RDKit {
     streamWrite(ss,tInt);
     tInt=d_numInts;
     streamWrite(ss,tInt);
+
+#if defined(BOOST_BIG_ENDIAN)
+    boost::uint32_t *td = new boost::uint32_t[d_numInts];
+    for(unsigned int i=0;i<d_numInts;++i) td[i]=EndianSwapBytes<HOST_ENDIAN_ORDER,LITTLE_ENDIAN_ORDER>(d_data.get()[i]);
+    ss.write((const char *)td,d_numInts*sizeof(tInt));
+    delete [] td;
+#else    
     ss.write((const char *)d_data.get(),d_numInts*sizeof(tInt));
-    
+#endif    
     std::string res(ss.str());
     return res;
   };
@@ -175,7 +182,15 @@ namespace RDKit {
     d_numInts=tInt;
     boost::uint32_t *data = new boost::uint32_t[d_numInts];
     ss.read((char *)data,d_numInts*sizeof(boost::uint32_t));
+
+#if defined(BOOST_BIG_ENDIAN)
+    boost::uint32_t *td = new boost::uint32_t[d_numInts];
+    for(unsigned int i=0;i<d_numInts;++i) td[i]=EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(data[i]);
+    d_data.reset(td);
+    delete [] data;
+#else    
     d_data.reset(data);
+#endif    
 
   };
 
--- rdkit-201203.orig/Code/DataStructs/testDatastructs.cpp
+++ rdkit-201203/Code/DataStructs/testDatastructs.cpp
@@ -125,9 +125,13 @@ template<typename T> void Test(T arg){
   delete t5;
 
   std::string pkl=t1.toString();
+  BOOST_LOG(rdInfoLog) << " a" << std::endl;
   const char *pkl64=Base64Encode(pkl.c_str(),pkl.size());
+  BOOST_LOG(rdInfoLog) << " b" << std::endl;
   T t6(t1.getNumBits());
+  BOOST_LOG(rdInfoLog) << " c" << std::endl;
   t6.initFromText(pkl64,strlen(pkl64),true);
+  BOOST_LOG(rdInfoLog) << " d" << std::endl;
   delete [] pkl64;
   TEST_ASSERT(t6==t1);
 }
@@ -149,10 +153,14 @@ template<typename T> void TaniTest(T &ar
   int idx=0;
   for(int i=0;i<4;i++){
     T v1(256);
+    BOOST_LOG(rdInfoLog) << " a "<<i << std::endl;
     FromDaylightString(v1,fps[i]);
+    BOOST_LOG(rdInfoLog) << " b"<<i << std::endl;
     for(int j=i;j<4;j++){
       T v2(256);
+      BOOST_LOG(rdInfoLog) << "     c " << j << std::endl;
       FromDaylightString(v2,fps[j]);
+      BOOST_LOG(rdInfoLog) << "     d " << j << std::endl;
       double tani=TanimotoSimilarity(v1,v2);
       TEST_ASSERT(feq(tani,dists[idx]));
       tani = TverskySimilarity(v1,v2,1.,1.);
@@ -1174,15 +1182,9 @@ int main(){
 
   ss.write((const char *)&v1,sizeof(v1));
   ss.write((const char *)&v2,sizeof(v2));
-#if 0
-  ss.close();
-  fstream ss2("blah.bin",ios_base::binary|ios_base::in);
-  ss2.read((char *)&v3,sizeof(v3));
-  ss2.read((char *)&v4,sizeof(v4));
-#endif
   ss.seekp(0,ios_base::beg);
-  ss.read((char *)&v3,sizeof(v3));
-  ss.read((char *)&v4,sizeof(v4));
+  RDKit::streamRead(ss,v3);
+  RDKit::streamRead(ss,v4);
   
   TXTMSG("v3",v3);
   TXTMSG("v4",v4);
@@ -1190,12 +1192,16 @@ int main(){
   BOOST_LOG(rdInfoLog) << " SPARSE -----------------------------------" << std::endl;
   SparseBitVect sparseFoo(10);
   Test(sparseFoo);
+  BOOST_LOG(rdInfoLog) << " 1" << std::endl;
   TaniTest(sparseFoo);
+  BOOST_LOG(rdInfoLog) << " 2" << std::endl;
   ProbeTest(sparseFoo);
   BOOST_LOG(rdInfoLog) << " Explicit ----------------------------------" << std::endl;
   ExplicitBitVect explicitFoo(10);
   Test(explicitFoo);
+  BOOST_LOG(rdInfoLog) << " 1" << std::endl;
   TaniTest(explicitFoo);
+  BOOST_LOG(rdInfoLog) << " 2" << std::endl;
   BOOST_LOG(rdInfoLog) << " Done" << std::endl;
   
   BOOST_LOG(rdInfoLog) << " Test DiscreteValue Vectors 1 ----------------------------" << endl;
--- rdkit-201203.orig/Code/DataStructs/SparseBitVect.cpp
+++ rdkit-201203/Code/DataStructs/SparseBitVect.cpp
@@ -317,10 +317,15 @@ std::string SparseBitVect::toString() co
 
   boost::int32_t tInt = ci_BITVECT_VERSION*-1;
   RDKit::streamWrite(ss,tInt);
+std::cerr<<"version: "<<tInt<<std::endl;
   tInt=d_size;
   RDKit::streamWrite(ss,tInt);
+std::cerr<<"size: "<<tInt<<std::endl;
+
   tInt=getNumOnBits();
   RDKit::streamWrite(ss,tInt);
+std::cerr<<"nOn: "<<tInt<<std::endl;
+
 
   int prev = -1;
   unsigned int zeroes;
--- rdkit-201203.orig/Code/RDGeneral/StreamOps.h
+++ rdkit-201203/Code/RDGeneral/StreamOps.h
@@ -16,10 +16,72 @@
 #include <sstream>
 #include <iostream>
 #include <boost/cstdint.hpp>
-
+#include <boost/detail/endian.hpp>
 
 namespace RDKit{
-    
+  // this code block for handling endian problems is from :
+  // http://stackoverflow.com/questions/105252/how-do-i-convert-between-big-endian-and-little-endian-values-in-c
+  enum EEndian
+    {
+      LITTLE_ENDIAN_ORDER,
+      BIG_ENDIAN_ORDER,
+#if defined(BOOST_LITTLE_ENDIAN)
+      HOST_ENDIAN_ORDER = LITTLE_ENDIAN_ORDER
+#elif defined(BOOST_BIG_ENDIAN)
+      HOST_ENDIAN_ORDER = BIG_ENDIAN_ORDER
+#else
+#error "Failed to determine the system endian value"
+#endif
+    };
+
+  // this function swap the bytes of values given it's size as a template
+  // parameter (could sizeof be used?).
+  template <class T, unsigned int size>
+  inline T SwapBytes(T value)
+  {
+    union
+    {
+      T value;
+      char bytes[size];
+    } in, out;
+
+    in.value = value;
+
+    for (unsigned int i = 0; i < size / 2; ++i)
+      {
+        out.bytes[i] = in.bytes[size - 1 - i];
+        out.bytes[size - 1 - i] = in.bytes[i];
+      }
+
+    return out.value;
+  }
+
+  // Here is the function you will use. Again there is two compile-time assertion
+  // that use the boost librarie. You could probably comment them out, but if you
+  // do be cautious not to use this function for anything else than integers
+  // types. This function need to be calles like this :
+  //
+  //     int x = someValue;
+  //     int i = EndianSwapBytes<HOST_ENDIAN_ORDER, BIG_ENDIAN_ORDER>(x);
+  //
+  template<EEndian from, EEndian to, class T>
+  inline T EndianSwapBytes(T value)
+  {
+    // A : La donnée à swapper à une taille de 2, 4 ou 8 octets
+    BOOST_STATIC_ASSERT(sizeof(T)==1 || sizeof(T) == 2 || sizeof(T) == 4 || sizeof(T) == 8);
+    if(sizeof(T)==1) return value;
+
+    // A : La donnée à swapper est d'un type arithmetic
+    //BOOST_STATIC_ASSERT(boost::is_arithmetic<T>::value);
+
+    // Si from et to sont du même type on ne swap pas.
+    if (from == to)
+      return value;
+
+    return SwapBytes<T, sizeof(T)>(value);
+  }
+  // --------------------------------------
+  
   //! Packs an integer and outputs it to a stream
   inline void appendPackedIntToStream(std::stringstream &ss, boost::uint32_t num) {
     int nbytes, bix;
@@ -27,7 +89,7 @@ namespace RDKit{
     char tc;
     
     CHECK_INVARIANT(num >= 0, "");
-    res = num;
+    res=num;
     while (1) {
       if (res < (1<<7)) {
         val = (res<<1);
@@ -53,9 +115,10 @@ namespace RDKit{
         break;
       }
       else {
-        CHECK_INVARIANT(0, "ERROR: Integer to big to pack\n");
+        CHECK_INVARIANT(0, "ERROR: Integer too big to pack\n");
       }
     }
+    //val = EndianSwapBytes<HOST_ENDIAN_ORDER,LITTLE_ENDIAN_ORDER>(val);
     
     for (bix = 0; bix < nbytes; bix++) {
       tc = (char) (val & 255);
@@ -100,6 +163,7 @@ namespace RDKit{
       offset = (1<<7) + (1<<14) + (1<<21);
     }
     num = (val >> shift) + offset;
+    //num = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(num);
     return num;
   }
 
@@ -147,18 +211,22 @@ namespace RDKit{
       offset = (1<<7) + (1<<14) + (1<<21);
     }
     num = (val >> shift) + offset;
+    //num = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(num);
     return num;
   }
   
   //! does a binary write of an object to a stream
   template <typename T>
     void streamWrite(std::ostream &ss,const T &val){
-    ss.write((const char *)&val,sizeof(T));
+    T tval=EndianSwapBytes<HOST_ENDIAN_ORDER,LITTLE_ENDIAN_ORDER>(val);
+    ss.write((const char *)&tval,sizeof(T));
   }
   //! does a binary read of an object from a stream
   template <typename T>
     void streamRead(std::istream &ss,T &loc){
-    ss.read((char *)&loc,sizeof(T));
+    T tloc;
+    ss.read((char *)&tloc,sizeof(T));
+    loc = EndianSwapBytes<LITTLE_ENDIAN_ORDER,HOST_ENDIAN_ORDER>(tloc);
   }
  
   //! grabs the next line from an instream and returns it.
--- rdkit-201203.orig/Code/ML/CMakeLists.txt
+++ rdkit-201203/Code/ML/CMakeLists.txt
@@ -1,4 +1,8 @@
-add_subdirectory(Cluster)
-add_subdirectory(FeatureSelect)
+if(RDK_BIG_ENDIAN)
+  message("Skipping build of clustering and cmim libraries on big endian system")
+else(RDK_BIG_ENDIAN)
+  add_subdirectory(Cluster)
+  add_subdirectory(FeatureSelect)
+endif(RDK_BIG_ENDIAN)
 add_subdirectory(InfoTheory)
 add_subdirectory(Data)
--- rdkit-201203.orig/Code/ML/FeatureSelect/Wrap/testCMIM.py
+++ rdkit-201203/Code/ML/FeatureSelect/Wrap/testCMIM.py
@@ -31,13 +31,13 @@ class TestCase(unittest.TestCase):
      examples.append([0,bv,1])
 
      r = FS.selectCMIM(examples,2)
-     self.failUnless(r==(2,4))
+     self.failUnlessEqual(r,(2,4))
 
      r = FS.selectCMIM(examples,1)
-     self.failUnless(r==(2,))
+     self.failUnlessEqual(r,(2,))
 
      r = FS.selectCMIM(examples,3)
-     self.failUnless(r==(2,4,-1))
+     self.failUnlessEqual(r,(2,4,-1))
 
    def test1FromList(self) :
      examples = []
@@ -63,13 +63,13 @@ class TestCase(unittest.TestCase):
      examples.append([0,bv,1])
 
      r = FS.selectCMIM(examples,2)
-     self.failUnless(r==(2,4))
+     self.failUnlessEqual(r,(2,4))
 
      r = FS.selectCMIM(examples,1)
-     self.failUnless(r==(2,))
+     self.failUnlessEqual(r,(2,))
 
      r = FS.selectCMIM(examples,3)
-     self.failUnless(r==(2,4,-1))
+     self.failUnlessEqual(r,(2,4,-1))
 
 
 if __name__ == '__main__':
--- rdkit-201203.orig/rdkit/DataStructs/VectCollection.py
+++ rdkit-201203/rdkit/DataStructs/VectCollection.py
@@ -230,12 +230,13 @@ class VectCollection(object):
   # set up our support for pickling:
   #
   def __getstate__(self):
-    pkl = struct.pack('I',len(self.__vects))
+    pkl = struct.pack('<I',len(self.__vects))
     for k,v in self.__vects.iteritems():
-      pkl += struct.pack('I',k)
+      pkl += struct.pack('<I',k)
       p = v.ToBinary()
       l = len(p)
-      pkl += struct.pack('I%ds'%(l),l,p)
+      pkl += struct.pack('<I',l)
+      pkl += struct.pack('%ds'%(l),p)
     return pkl
 
   def __setstate__(self,pkl):
@@ -245,12 +246,12 @@ class VectCollection(object):
     self.__needReset=True
     szI = struct.calcsize('I')
     offset = 0
-    nToRead = struct.unpack('I',pkl[offset:offset+szI])[0]
+    nToRead = struct.unpack('<I',pkl[offset:offset+szI])[0]
     offset += szI
     for i in range(nToRead):
-      k = struct.unpack('I',pkl[offset:offset+szI])[0]
+      k = struct.unpack('<I',pkl[offset:offset+szI])[0]
       offset += szI
-      l = struct.unpack('I',pkl[offset:offset+szI])[0]
+      l = struct.unpack('<I',pkl[offset:offset+szI])[0]
       offset += szI
       sz = struct.calcsize('%ds'%l)
       bv = DataStructs.ExplicitBitVect(struct.unpack('%ds'%l,pkl[offset:offset+sz])[0])
--- rdkit-201203.orig/rdkit/SimDivFilters/__init__.py
+++ rdkit-201203/rdkit/SimDivFilters/__init__.py
@@ -3,4 +3,8 @@
 #    All Rights Reserved
 #
 from rdkit import rdBase
-from rdSimDivPickers import *
+try:
+  import rdSimDivPickers
+  from rdSimDivPickers import *
+except ImportError:
+  rdSimDivPickers=None
--- rdkit-201203.orig/rdkit/ML/test_list.py
+++ rdkit-201203/rdkit/ML/test_list.py
@@ -5,7 +5,7 @@ tests=[
   ("python","UnitTestAnalyzeComposite.py",{}),
   ]
 
-for dir in ['Cluster','Composite','Data','DecTree','Descriptors','FeatureSelect','InfoTheory','KNN','ModelPackage','NaiveBayes','Neural','SLT']:
+for dir in ['Cluster','Composite','Data','DecTree','Descriptors','InfoTheory','KNN','ModelPackage','NaiveBayes','Neural','SLT']:
     tests.append(('python','test_list.py',{'dir':dir}))
 
 longTests=[
--- rdkit-201203.orig/rdkit/ML/InfoTheory/BitClusterer.py
+++ rdkit-201203/rdkit/ML/InfoTheory/BitClusterer.py
@@ -3,6 +3,8 @@
 #
 
 from rdkit.SimDivFilters import rdSimDivPickers as rdsimdiv
+if rdsimdiv is None:
+   raise ImportError,'rdSimDivPickers not built'
 from rdkit import DataStructs
 import numpy
 
--- rdkit-201203.orig/rdkit/ML/InfoTheory/testCorrMatGen.py
+++ rdkit-201203/rdkit/ML/InfoTheory/testCorrMatGen.py
@@ -2,7 +2,11 @@
 
 from rdkit import RDConfig
 import unittest
-from rdkit.ML.InfoTheory import rdInfoTheory, BitClusterer
+from rdkit.ML.InfoTheory import rdInfoTheory
+try:
+  from rdkit.ML.InfoTheory import BitClusterer
+except ImportError:
+  BitClusterer=None
 from rdkit.ML.Data import DataUtils
 from rdkit import DataStructs
 import random
@@ -64,6 +68,8 @@ class TestCase(unittest.TestCase):
         assert 2*navr/self.nbits == 158.3,2*navr/self.nbits
 
     def test1Cluster(self) :
+        if BitClusterer is None:
+            return
         cmg = rdInfoTheory.BitCorrMatGenerator()
         cmg.SetBitList(self.blist)
         for fp in self.fps:
--- rdkit-201203.orig/rdkit/ML/Cluster/UnitTestCluster.py
+++ rdkit-201203/rdkit/ML/Cluster/UnitTestCluster.py
@@ -14,7 +14,11 @@
 from rdkit import RDConfig
 import unittest
 from rdkit.ML.Cluster import Standardize,ClusterUtils
-from rdkit.ML.Cluster import Clusters,Murtagh
+from rdkit.ML.Cluster import Clusters
+try:
+  from rdkit.ML.Cluster import Murtagh
+except ImportError:
+  Murtagh=None
 import numpy
 import cPickle
 
@@ -57,6 +61,7 @@ class TestCase(unittest.TestCase):
     assert indices==[8,7,5,6],'bad index order'
 
   def testMurtaghUPGMA(self):
+    if Murtagh is None: return
     nPts = 5
     sz = 5
     dataP = numpy.random.random((nPts,sz))
--- rdkit-201203.orig/rdkit/ML/FeatureSelect/UnitTestCMIM.py
+++ rdkit-201203/rdkit/ML/FeatureSelect/UnitTestCMIM.py
@@ -1,4 +1,7 @@
-from rdkit.ML.FeatureSelect import CMIM
+try:
+  from rdkit.ML.FeatureSelect import CMIM
+except ImportError:
+  CMIM = None
 from rdkit import DataStructs as DS
 from rdkit import RDConfig
 import unittest
@@ -8,6 +11,8 @@ class TestCase(unittest.TestCase):
       pass
 
    def test0FromList(self) :
+     if CMIM is None:
+       return
      examples = []
 
      bv = DS.ExplicitBitVect(5)
--- rdkit-201203.orig/rdkit/ML/DecTree/BuildSigTree.py
+++ rdkit-201203/rdkit/ML/DecTree/BuildSigTree.py
@@ -12,7 +12,10 @@
 import numpy
 from rdkit.ML.DecTree import SigTree
 from rdkit.ML import InfoTheory
-from rdkit.ML.FeatureSelect import CMIM
+try:
+  from rdkit.ML.FeatureSelect import CMIM
+except ImportError:
+  CMIM=None
 from rdkit.DataStructs.VectCollection import VectCollection
 import copy
 import random
@@ -136,7 +139,7 @@ def BuildSigTree(examples,nPossibleRes,e
     nBits = fp.GetNumBits()
     ranker = InfoTheory.InfoBitRanker(nBits,nPossibleRes,metric)
     if biasList: ranker.SetBiasList(biasList)
-    if useCMIM > 0 and not ensemble:
+    if CMIM is not None and useCMIM > 0 and not ensemble:
       ensemble = CMIM.SelectFeatures(examples,useCMIM,bvCol=1)
     if random:
       if ensemble:
--- rdkit-201203.orig/rdkit/ML/NaiveBayes/CrossValidate.py
+++ rdkit-201203/rdkit/ML/NaiveBayes/CrossValidate.py
@@ -9,14 +9,17 @@ and evaluation of individual models
 """
 from rdkit.ML.NaiveBayes.ClassificationModel import NaiveBayesClassifier
 from rdkit.ML.Data import SplitData
-from rdkit.ML.FeatureSelect import CMIM
+try:
+  from rdkit.ML.FeatureSelect import CMIM
+except ImportError:
+  CMIM=None
 
 def makeNBClassificationModel(trainExamples, attrs, nPossibleValues, nQuantBounds,
                               mEstimateVal=-1.0,
                               useSigs=False,
                               ensemble=None,useCMIM=0,
                               **kwargs) :
-  if useCMIM > 0 and useSigs and not ensemble:
+  if CMIM is not None and useCMIM > 0 and useSigs and not ensemble:
     ensemble = CMIM.SelectFeatures(trainExamples,useCMIM,bvCol=1)
   if ensemble:
     attrs = ensemble