Re: [AMBER-Developers] Confusing clean targets.

From: Jason Swails <jason.swails.gmail.com>
Date: Fri, 29 Jan 2010 17:45:11 -0500

Hello,

On Fri, Jan 29, 2010 at 5:12 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
> Dear All,
>
> Can someone offer a clarification as to why we have a whole plethora of
> different clean targets and specifically, why make clean does not ACTUALLY
> clean things and instead leaves all sort of libraries and modules behind
> that may have been compiled with different compiler versions of even
> compilers?
>
> For example, we have:
>
> clean:
> realclean:
> distclean:
> superclean:
> uninstall:
> uninstall_libraries:

There certainly are a lot of 'clean' rules...

>
> Looks like a little too much has been smoked of late.
>
> I propose doing the following:
>
> make clean
>
> This will do:
>
> clean all src directories
> remove all mod and library files
> clean netcdf, pnetcdf, fftw
> etc etc
>
> Objections?

Problems in the past have arisen when "clean" made things a bit *too*
clean. I propose that two levels of clean be kept on, since make -f
Makefile_at clean is not only necessary before a parallel AT is built,
but also instructed. Thus, what shouldn't happen is a make clean
removing libraries that are required for programs built in serial that
are NOT rebuilt in the parallel build (as I recall, nab especially had
a problem with this, so what I frequently had to do was make at
serial, make clean to make at parallel, then make clean and make
serial again to put the removed libraries back if I forgot to move
them to a safe place first). Thus, my opinion is that the clean rule
should remain the way it is, and perhaps consolidate the rest of the
"cleans" into a single rule that is pretty aggressive in case you want
to, for instance, switch compilers or something of the sort. This
could simply be called "uninstall", since I imagine that a rule that
rips the guts out of an installation should also remove the
executables as well.

Any other thoughts?

Thanks!
Jason


-- 
---------------------------------------
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Jan 29 2010 - 15:00:02 PST
Custom Search