Support building without gtest

Now the only dependence is boost, which most modern linux distros have.

5 files changed, 30 insertions, 14 deletions

diff --git a/configure.ac b/configure.ac
index 0fd43e08..efb32284 100644
--- a/configure.ac
+++ b/configure.ac
@@ -11,8 +11,8 @@ CPPFLAGS="$CPPFLAGS $BOOST_CPPFLAGS"
 AC_CHECK_HEADER(boost/math/special_functions/digamma.hpp,
                [AC_DEFINE([HAVE_BOOST_DIGAMMA], [], [flag for boost::math::digamma])])
 
-GTEST_LIB_CHECK
 AC_PROG_INSTALL
+GTEST_LIB_CHECK
 
 AC_ARG_WITH(srilm,
             [AC_HELP_STRING([--with-srilm=PATH], [(optional) path to SRI's LM toolkit])],
diff --git a/decoder/Makefile.am b/decoder/Makefile.am
index 4c86ae6f..19c22bc5 100644
--- a/decoder/Makefile.am
+++ b/decoder/Makefile.am
@@ -1,4 +1,7 @@
-bin_PROGRAMS = \
+bin_PROGRAMS = cdec
+
+if HAVE_GTEST
+noinst_PROGRAMS = \
   dict_test \
   weights_test \
   trule_test \
@@ -6,8 +9,8 @@ bin_PROGRAMS = \
   ff_test \
   parser_test \
   grammar_test \
-  cdec \
   small_vector_test
+endif
 
 cdec_SOURCES = cdec.cc forest_writer.cc maxtrans_blunsom.cc cdec_ff.cc ff_factory.cc timing_stats.cc
 small_vector_test_SOURCES = small_vector_test.cc
diff --git a/training/Makefile.am b/training/Makefile.am
index 944c75f7..b44812e7 100644
--- a/training/Makefile.am
+++ b/training/Makefile.am
@@ -4,9 +4,11 @@ bin_PROGRAMS = \
   grammar_convert \
   atools \
   plftools \
-  lbfgs_test \
   mr_em_train \
-  collapse_weights \
+  collapse_weights
+
+noinst_PROGRAMS = \
+  lbfgs_test \
   optimize_test
 
 atools_SOURCES = atools.cc
diff --git a/training/cluster-ptrain.pl b/training/cluster-ptrain.pl
index 8944ae34..7643d4e5 100755
--- a/training/cluster-ptrain.pl
+++ b/training/cluster-ptrain.pl
@@ -1,14 +1,13 @@
 #!/usr/bin/perl -w
 
 use strict;
-my $SCRIPT_DIR; BEGIN { use Cwd qw/ abs_path /; use File::Basename; $SCRIPT_DIR = dirname(abs_path($0)); push @INC, $SCRIPT_DIR; }
+my $SCRIPT_DIR; BEGIN { use Cwd qw/ abs_path getcwd /; use File::Basename; $SCRIPT_DIR = dirname(abs_path($0)); push @INC, $SCRIPT_DIR; }
 use Getopt::Long;
 
 my $MAX_ITER_ATTEMPTS = 5; # number of times to retry a failed function evaluation
-my $CWD=`pwd`; chomp $CWD;
-my $BIN_DIR = $SCRIPT_DIR;
-my $OPTIMIZER = "$BIN_DIR/mr_optimize_reduce";
-my $DECODER = "$BIN_DIR/../decoder/cdec";
+my $CWD=getcwd();
+my $OPTIMIZER = "$SCRIPT_DIR/mr_optimize_reduce";
+my $DECODER = "$SCRIPT_DIR/../decoder/cdec";
 my $COMBINER_CACHE_SIZE = 150;
 # This is a hack to run this on a weird cluster,
 # eventually, I'll provide Hadoop scripts.
@@ -30,13 +29,15 @@ my $PRIOR;
 my $OALG = "lbfgs";
 my $sigsq = 1;
 my $means_file;
+my $RESTART_IF_NECESSARY;
 GetOptions("cdec=s" => \$DECODER,
-           "run_locally" => \$LOCAL,
            "distributed" => \$DISTRIBUTED,
            "sigma_squared=f" => \$sigsq,
+           "max_iteration=i" => \$max_iteration,
            "means=s" => \$means_file,
            "optimizer=s" => \$OALG,
            "gaussian_prior" => \$PRIOR,
+           "restart_if_necessary" => \$RESTART_IF_NECESSARY,
            "jobs=i" => \$nodes,
            "pmem=s" => \$pmem
           ) or usage();
@@ -61,6 +62,11 @@ if ($parallel) {
 unless ($parallel) { $CFLAG = "-C 500"; }
 unless ($config_file =~ /^\//) { $config_file = $CWD . '/' . $config_file; }
 my $clines = num_lines($training_corpus);
+my $dir = "$CWD/ptrain";
+
+if ($RESTART_IF_NECESSARY && -d $dir) {
+  $restart = 1;
+}
 
 print STDERR <<EOT;
 PTRAIN CONFIGURATION INFORMATION
@@ -83,7 +89,6 @@ my $nodelist="1";
 for (my $i=1; $i<$nodes; $i++) { $nodelist .= " 1"; }
 my $iter = 1;
 
-my $dir = "$CWD/ptrain";
 if ($restart) {
   die "$dir doesn't exist, but --restart specified!\n" unless -d $dir;
   my $o = `ls -t $dir/weights.*`;
@@ -161,6 +166,7 @@ while ($iter < $max_iteration) {
 }
 
 print "FINAL WEIGHTS: $dir/weights.$iter\n";
+`mv $dir/weights.$iter.gz $dir/weights.final.gz`;
 
 sub usage {
   die <<EOT;
@@ -170,6 +176,7 @@ Usage: $0 [OPTIONS] cdec.ini training.corpus weights.init
   Options:
 
     --distributed      Parallelize function evaluation
+    --jobs N           Number of jobs to use
     --cdec PATH        Path to cdec binary
     --optimize OPT     lbfgs, rprop, sgd
     --gaussian_prior   add Gaussian prior
diff --git a/vest/Makefile.am b/vest/Makefile.am
index d7d08133..80ad7218 100644
--- a/vest/Makefile.am
+++ b/vest/Makefile.am
@@ -1,12 +1,16 @@
 bin_PROGRAMS = \
   mr_vest_map \
   mr_vest_reduce \
-  scorer_test \
-  lo_test \
   mr_vest_generate_mapper_input \
   fast_score \
   union_forests
 
+if HAVE_GTEST
+noinst_PROGRAMS = \
+  scorer_test \
+  lo_test
+endif
+
 union_forests_SOURCES = union_forests.cc
 union_forests_LDADD = $(top_srcdir)/decoder/libcdec.a -lz