Re: Non-recursive make with Fortran object files

[Dave Goodell]

On Dec 5, 2012, at 2:29 AM PST, Alexis Praga wrote:

>> If you don't need a fully robust strategy that works for parallel make
>> and all sorts of different corner cases (particular files deleted/touched,
>> etc.), then you can likely get by with a simpler set of rules.
> 
> Actually, even when I delete all the .mod file, this does not force a
> rebuild. Is this one of your corner case ? It looks a major issue to me, if
> you do not want to use make clean to recompile everything.

Right, this is exactly the sort of thing that I was talking about.

> So, if i understand correctly, the safer way would be to use only
> dependencies with .f90 and .mod ?

The only robust way that I have found to deal with this is to use the pattern 
that I linked to in my previous email.  Perhaps someone else has a simpler 
approach.

>>> [2] https://gist.github.com/4190680
> 
> One last thing, in this example, why do you use a temporary stamp file,
> instead of creating a rule directly between the library and module files
> and the header ?

I can't remember exactly why that is done, but it was certainly done that way 
to handle some particular case that I was hitting.  I think it ensures that the 
mtime for the "-stamp" file is earlier than the mtime for the ".$(MOD)" file.  
Otherwise make will attempt to rebuild the mod file every time, since it will 
always appear to be out of date w.r.t. its primary dependency.

-Dave