Thanks for pointing this. I am not sure, but I could remember the current way is to compile to object files and then linking together to a feature-test binary.
Besides, I think the current compile time is like 80% of linking + 20% of compiling objects. Best xunzhang Send from my iPhone > 在 2016年7月8日，10:23，Paul Guo <[email protected]> 写道： > > I'm adding some feature tests into hawq code, however I found currently the > makefile code compiles all of the test cpp files each time, regardless of > single file change or no file change. This is too time-consuming given in the > future there will be more and more cases. See more below. > > g++ -I/usr/include -I/usr/local/include -I/usr/include/libxml2 > -I/data2/github/incubator-hawq/src/test/feature/ > -I/data2/github/incubator-hawq/src/test/feature/lib/ > -I/data2/github/incubator-hawq/src/interfaces/libpq > -I/data2/github/incubator-hawq/src/interfaces > -I/data2/github/incubator-hawq/src/include > -I/data2/github/incubator-hawq/depends/thirdparty/googletest/googletest/include > > -I/data2/github/incubator-hawq/depends/thirdparty/googletest/googlemock/include > -Wall -O0 -g -std=c++11 test_main.cpp ExternalSource/test_errortbl.cpp > ExternalSource/test_exttab.cpp ManagementTool/test_hawq_register.cpp > PreparedStatement/TestPreparedStatement.cpp UDF/TestUDF.cpp > ao/TestAoSnappy.cpp catalog/test_alter_owner.cpp catalog/test_alter_table.cpp > lib/command.cpp lib/data_gen.cpp lib/file_replace.cpp lib/hawq_config.cpp > lib/psql.cpp lib/sql_util.cpp lib/string_util.cpp lib/xml_parser.cpp > parquet/test_parquet.cpp partition/test_partition.cpp > query/test_aggregate.cpp query/test_insert.cpp query/test_prepare.cpp > query/test_sequence.cpp testlib/test_lib.cpp type/TestType.cpp > -L../../../src/port -L../../../src/port -Wl,--as-needed > -L/data2/github/incubator-hawq/depends/libhdfs3/build/install/data2/github/hawq-install/lib > > -L/data2/github/incubator-hawq/depends/libyarn/build/install/data2/github/hawq-install/lib > -Wl,-rpath,'/data2/github/hawq-install/lib',--enable-new-dtags > -L/usr/local/lib -L/usr/lib -L/data2/github/incubator-hawq/src/test/feature/ > -L/data2/github/incubator-hawq/src/test/feature/lib/ > -L/data2/github/incubator-hawq/src/interfaces/libpq > -L/data2/github/incubator-hawq/depends/thirdparty/googletest/build/googlemock > -L/data2/github/incubator-hawq/depends/thirdparty/googletest/build/googlemock/gtest > -lpgport -ljson-c -levent -lyaml -lsnappy -lbz2 -lrt -lssl -lcrypto -lz > -lreadline -lcrypt -ldl -lm -L/usr/local/lib -lcurl -lyarn -lkrb5 -lgtest > -lpq -lxml2 -ltest -o feature-test > > > PROG = test_main.cpp $(wildcard */*.cpp) > RM = rm -rf > > .PHONY: all sharelib sharelibclean distclean clean doc > > all: sharelib > $(MAKE) -C lib all > $(CXX) $(CPPFLAGS) $(CXXFLAGS) $(PROG) $(LDFLAGS) $(LIBS) -o feature-test > > Above is related code in Makefile. since share lib is a .PHONY target without > dependency, so > above code will always be executed. > > So we could compile each test_$case.cpp into one .o and link link them into > one. > The will make compile faster, besides, take the feature test cases under ao/ > for example, > we could make the relative path starts with $test_root/ao for the ao/ cases. > This > make more sense. > > TEST_F(TestAOSnappy, Create1048576) > { > hawq::test::SQLUtility util; > util.execSQLFile("ao/sql/ao_crtb_with_row_snappy_1048576.sql", > "ao/ans/ao_crtb_with_row_snappy_1048576.ans", > initFile); > } > > Any idea?
