Re: Automatic dependencies and clean rule

[Philip Guenther]

On Wed, Mar 27, 2013 at 8:53 PM, Nick Andrik <address@hidden> wrote:
>>> My problem is that every time that I run make clean, the rules for
>>> creating DEPS are executed again.
>>> Is there any way to tell the Makefile to ignore the final include line
>>> (or empty the DEPS var) when I run make clean?
>>
>> The really short answer is that you should read
>>         http://make.paulandlesley.org/autodep.html#advanced
>
> I have laready been through that, but I found no way to get rid of the
> .d generation when I call clean

That webpage explain how: don't have the .d files as targets!  The .d
files should only be generated as a *side effect* of the generation of
the .o files.


...
>> When there are file that require two stages to build, the second of
>> which may require dependencies that you cannot calculate until the
>> first is done (the lexer.c file might not be compilable without the
>> parser.tab.h file generated by bison).  So, to auto-generate
>> dependencies for the .o files you'll need to previously inform make of
>> the relevant ".c: .y" and ".c: .l" dependencies.
>
> Sorry for not being clear up front. I do have rules for generating all
> the .cpp .hh .hpp files needed for the other rules. I didn't include
> these rules in my initial example because I wanted to keep it kind of
> simple. We can consider that there are rules for every generated c++
> and header files.

Ah, good.


>> My preference is to declare the generated .h files to be order-only
>> dependencies of *all* the .o files.  That way make will know that it
>> has to generate the .h before it tries to compile the .o's, but once
>> the .h exists the .o's will only be recompiled if the auto-dependency
>> bits generate a *real* dependency on the .h.
>>
>> ${OBJS} : | parser.tab.h
>>
>> (...or whatever the name of the generated .h file(s) are)
>
> Hmm, that's an interesting approach, but I still need to use the
> -include directive, no?

Yes, you still need the -include...

> If I do, then my problem persists: .d files are created for each make
> clean invocation

...but this won't happen if the .d files are not targets.


> By the way, I might go on with some double calling of the makefile and
> check if I can get away with the include directive

Ick, that's strictly worse.


Philip Guenther