The configure scripts have two issues:
(1) in the test to determine HAVE_GLIB, they test GTKMM_LIBS.  I think this should be GLIB_LIBS.
(2) there are a few bashisms involving 'test ... == ...'; POSIX sh requires this to be a single =

--- configure.ac.orig   2007-09-14 12:50:42.000000000 -0600
+++ configure.ac
@@ -119,7 +119,7 @@ AC_ARG_ENABLE( experimental,
   [  --disable-experimental  compile without experimental features],
        config_experimental=$enableval
        ,
-       config_experimental==yes
+       config_experimental=yes
     )
 
 dnl AC_MSG_CHECKING(whether to link statically)
@@ -460,7 +460,7 @@ AC_SUBST(GNET_LIBS)
 AC_SUBST(GNET_CFLAGS)
 
         
-if test "x$GNET_LIBS" == "x"; then
+if test "x$GNET_LIBS" = "x"; then
    config_distribution=no
 fi
 
@@ -588,7 +588,7 @@ AM_GNU_GETTEXT()
 #AM_GLIB_GNU_GETTEXT
 
 dnl Hack because windows gettext package is SEVERELY broken...
-if test "x$platform_win32" == "xyes"; then
+if test "x$platform_win32" = "xyes"; then
         USE_NLS=yes
         AC_DEFINE(ENABLE_NLS,1,[Define to 1 if translation of program messages to the user's native
    language is requested.])
@@ -719,7 +719,7 @@ if test "x$config_with_registry" = "xyes
 fi
 
 HAVE_GLIB=
-if test "x$GTKMM_LIBS" != "x"; then
+if test "x$GLIB_LIBS" != "x"; then
    AC_DEFINE(HAVE_GLIB,,[Define if glib is available])
 fi

Thank you. I don't know what the intention is with GTKMM_LIBS. Were you able to build on/for NetBSD?

My guess is that it was the result of a partial copy-and-paste.

With this patch (particularly the ==), and the patch in bug 684 (to use <sys/wait.h>, I was able to build on NetBSD.

Fixed (I rewrote most of the configure script)