Improve compilation dependency handling

[jcoupey]

This ticket captures a problem discussed in #76.

TL;DR: dependencies are not handled properly with the current makefile.

This is not a problem when checking out one version and compiling once. But in a dev environment, changing files does not always trigger all required re-compilation. For a more detailed explanation, see @sashakh comment and proposed fix.

I'm by no mean an expert at compilation and makefiles, so I'd be grateful for other views on this to pick a solution.

[krypt-n]

A simple solution would be:

--- a/src/makefile
+++ b/src/makefile
@@ -33,6 +33,7 @@ else
 endif
 
 OBJ = $(SRC:.cpp=.o)
+DEPS = $(SRC:.cpp=.d)
 
 # Main target.
 all : $(MAIN) $(LIB)
@@ -46,8 +47,10 @@ $(LIB) : $(OBJ)
 	$(AR) cr $@ $^
 
 # Building .o files.
-%.o : %.cpp %.h
-	$(CXX) $(CXXFLAGS) -c $< -o $@
+%.o : %.cpp
+	$(CXX) $(CXXFLAGS) -MMD -MP -c $< -o $@
+
+-include ${DEPS}
 
 # Specific target for main (no .h file).
 main.o : main.cpp

This causes the c++ compiler to emit .d files that contain the header-cpp dependency information. These files are then included during the next run of make.

Pro: this does not require manual invocation of make dep, everything is handled automatically

Contra: a lot of .d files in the src-directory. They (aswell as the .o files) should at least be gitignored.

Do you want a PR for this change?

[jcoupey]

Thanks for the input @krypt-n. The advantage I see with this approach is that there is no need to use an additional custom target to generate the dependency files, so this would work out of the box for anyone just using plain make.

The downside of having .d files all around does not look problematic to me, as long as they're ignored by git.

A PR would be really appreciated!

[jcoupey]

Solved by #85

@krypt-n thanks a lot for the PR and the makefile springclean!