On Jul 13, 2011, at 9:43 AM CDT, Nick Bowler wrote: > On 2011-07-13 07:59 -0400, Jeff Squyres wrote: >> On Jul 12, 2011, at 12:23 PM, Dave Goodell wrote: >>> Is there a better way? I really expected version 2 to work and was >>> surprised when it didn't. >> >> I've always a) assumed that Automake doesn't have Fortran dependency >> analysis, and b) coped with it by adding explicit dependency rules >> myself (which is icky). Something like this: >> >> foo_SOURCES = foo.f bar.h >> foo.f: bar.h > > Note that this isn't quite right: the source file does not actually > depend on the header. Such a dependency would imply that if you update > the header file, the source file needs to be regenerated. But what one > usually wants is for the *object* file to be regenerated when a header > is updated: IOW, the compiled object is what depends on the header, not > the source file itself. > > So this would perhaps be better written as > > foo_obj = foo.$(OBJEXT) # might need to be foo.lo if using libtool. > $(foo_obj): bar.h
Your correction of Jeff's solution made me realize that my "version 3" solution is wrong. It expresses a foo-->bar.h dependency instead of foo.o-->bar.h dependency. Your approach is what I was looking for. Unfortunately this solution gets even uglier looking if foo has any per-target flags such as "foo_FFLAGS=-I.". -Dave
