Add informative comment and revert unneeded (?) change

Surely if we need to use -D_GLIBCXX_HAS_GTHREADS on Android we should do that
somehow globally then and not just here in liborcus? But I haven't seen any
compilation errors caused by a lack of that? (Except, if memory serves me
right, for the Android on MIPS platform, but I haven't tried building that for
ages.)

And anyway, setting CPPFLAGS to one value in the environment for configure and
passing a CPPFLAGS setting on the configure command line (in the
--enable-dbgutil case) surely is pointless.

This reverts 84374b2d

Change-Id: I2a280bb24dc0a557c825070866e1969c3176fc3f

liborcus/ExternalProject_liborcus.mk

@@ -47,10 +47,17 @@ endif
 
 else
 
-# must be built with debug STL if --enable-dbgutil
+# Must be built with debug GNU C++ library if --enable-dbgutil has
+# caused the LO code to be built thusly.
+
+# The LIBS setting for Android is needed to get the orcus-xml-dump
+# executable to build successfully. We obviously don't actually need
+# that executable on Android, but we don't want to bother with
+# patching out building it for Android.
+
 $(call gb_ExternalProject_get_state_target,liborcus,build) :
 	$(call gb_ExternalProject_run,build,\
-		$(if $(filter ANDROID,$(OS)),LIBS='-lgnustl_shared -lm' CPPFLAGS="-D_GLIBCXX_HAS_GTHREADS") \
+		$(if $(filter ANDROID,$(OS)),LIBS='-lgnustl_shared -lm') \
 		./configure \
 			--with-pic \
 			--enable-static \