ansaurus

Question

Makefile Makeover -- Almost Complete, Want Feedback

Answer 1

+4 A:

I would rethink calling make recursively, and instead @include the submakefiles in the main makefile. It is much easier to get the dependancies to work right, and it allows you to utilize multiple cores for building (using -j). Take a look at Recursive Make Considered Harmful for all the gory details.

Also, take a look at these questions:
What is your experience with non recursive make?
Recursive make friend or foe

KeithB 2010-07-08 20:24:42

Is implementing a non-recursive make worth the confusion for a project w/ less than 200 files? Another concern is that most of the people in my group are chemical engineers with light computer-sci backgrounds so if I was going to do that I'd have to make it as understandable/maintainable as possible... which seems like it would be tough...

Jason R. Mick 2010-07-08 21:19:34

Actually, its pretty easy. The sub-makefiles are very simple, all of the boiler-plate code is in the main makefile. It reduces a lot of duplication. Once it is setup, adding new object files or executables is trivial. Adding a new subdirectory is not much harder. Also, you may have only 200 files now, but that will only go up. Its easier to make the switch now.

KeithB 2010-07-08 22:55:13

Answer 2

+1 A:

(I won't enter into the recursive/non-recursive debate again. 'Nuff said.)

Why copy all of these files around? I understand that you don't want to hardcode a map of your directory structure in the submake, but you can save a lot of time and trouble by giving the submake access to the original files, either by symbolic link:

molecule_test_prepare_sources: molecule_test_dir
    @echo linking to sources...
    @ln -s $(SOURCE_DIR) $(TEST_DIR)/molecule_unit_test/source  

molecule_test_prepare_makefiles: $(MAKEFILE_DIR)/Makefile.molecule_test
    @ln -s $< $(TEST_DIR)/molecule_unit_test/Makefile

Or by passing a parameter to the submake:

param_test : param_test_prepare_sources param_test_prepare_makefiles \
  param_test_prepare_data_files
    @$(MAKE) -C $(TEST_DIR)/param_unit_test/ SOURCE_DIR=$(SOURCE_DIR) \
      ./bin/param_test # and then use SOURCE_DIR in the submake

The same goes for param_test. I think this actually lets you do away with MOLECULE_UT_SOURCES and PARAM_UT_SOURCES, and good riddance. (I left out the data directories, because I don't know what this code actually does-- maybe it needs a restriced diet, or modifies its input files or something.)

Finally as a matter of style, almost anywhere you see redundancy you can remove it and make the makefile easier to read. For instance,

$(EXE_DIR)/molecule_test : $(OBJ_DIR)/main.o \
  $(OBJ_DIR)/parameter_manager_lj_molecule.o \
  $(OBJ_DIR)/parameter_manager.o $(OBJ_DIR)/parser.o \
  $(OBJ_DIR)/molecule_manager.o $(OBJ_DIR)/molecule_manager_main.o \
  $(OBJ_DIR)/molecule_reader.o \
  $(OBJ_DIR)/molecule_reader_psf_pdb.o
  $(CC) $^ -o $@

can become

OBJECTS := main \
  parameter_manager_lj_molecule \
  parameter_manager parser \
  molecule_manager molecule_manager_main \
  molecule_reader \
  molecule_reader_psf_pdb

OBJECTS := $(patsubst %,$(OBJ_DIR)/%.o, $(OBJECTS)

$(EXE_DIR)/molecule_test : $(OBJECTS)
  $(CC) $^ -o $@

Beta 2010-07-08 23:18:10

Answer 3

+1 A:

This isn't the answer you're looking for but have you considered using the "autotools" suite (automake, autoconf, etc.)?

Once you get the hang of it it's quite wonderful to work with. And it has a lot more functionality than pure Make. Functionality such as checking for libraries needed to build, cross-compiling, etc.

Staffan 2010-07-08 23:21:09

ansaurus

tags:

views:

answers:

Makefile Makeover -- Almost Complete, Want Feedback

related questions