Re: dependency issue

[Josh Bialkowski]

> > > > > >     I use the following rules to generate dependency automatically
> > > > > >
> > > > > > %.P : %.cpp
> > > > > >          $(QUIET_MM)rm -f $@; \
> > > > > >          $(CPP) -M $(CPPFLAGS) $(CXXFLAGS) $(INCLUDES)  $<  >  address@hidden; \

The rule from the make manual:

%.d: %.c
    @set -e; rm -f $@; \
    $(CC) -M $(CPPFLAGS) $< > address@hidden; \
    sed 's,\($*\)\.o[ :]*,\1.o $@ : ,g' < address@hidden > $@; \
    rm -f address@hidden


I find that rule to be overly complicated and confusing. Rather, I 
follow the advice in http://mad-scientist.net/make/autodep.html to 
handle auto dependency. In particular, why should you have a separate 
rule for the dependency file? I use the following style rule:

%.o: %.c
    gcc -Wp,-M,-MF$*.d,-MT$@ -c -o $@ $<

-Wp says to pass the following options on to the preprocessor
-M says to generate dependency
-MF says to ouput the dependency file to <stem>.d instead of stdout

(note, this is an example, probably you want something with CFLAGS and 
so forth)

The idea is this: build the dependency file at the same time as the 
object file. If the object hasn't been built yet, then there is no need 
for a dependency file, because make already has to build the object 
file. We only actually care about the object file's dependencies if the 
object file exists. I include the dependency files with "-include" 
instead of "include" so that the dependencies are only included if they 
actually exist. The construction makes more sense to me, and it cuts 
down on the number of processes required in a build scenario.

In any case, that page is a good read an is worth understanding if 
you're doing automatic dependency generation.


Josh