Re: [AMBER-Developers] Confusing clean targets.

From: Gustavo Seabra <gustavo.seabra.gmail.com>
Date: Tue, 2 Mar 2010 13:13:39 -0300

On Fri, Jan 29, 2010 at 9:23 PM, Scott Brozell <sbrozell.rci.rutgers.edu> wrote:
> On Fri, Jan 29, 2010 at 06:11:34PM -0500, case wrote:
>> On Fri, Jan 29, 2010, Ross Walker wrote:
>> >
>> > clean
>> > -----
>> > This removes the libraries (.a file) from the lib directory.
>>
>> This is problematic: NAB is a compiler, and uses some of the .a files in
>> amber11/lib to compile programs.  If you remove all *.a files in lib, you
>> effectively have to recompile NAB after each "make clean".
>>
>> If there are particular *.a files that should be removed with "make clean",
>> you can get rid of those.  But "make clean" should not remove all of them.
>>
>> >
>> > distclean
>>
>> This is probably a bad name for what might be a good target.  "distclean"
>> implies that you get back to the "distribution" (nothing beyond what was in
>> the tar files), but you don't really do that.
>>
>> But I don't really see the need for this.  (1) the test directory can be
>> "cleaned" in its own directory; (2) does removing config.h really serve a
>> useful purpose?
>
> Yes, it makes cross platform QA testing from one sandbox less error
> prone.  There is a need for a correct distclean and superclean:
> amber has become so big that it is not practical to re-checkout
> the tree for nightly testing, and having these targets makes it easy to
> have simple and consistent (relative to fast interactive testing and
> relative to other software) scripts for QA.
>
> Most of the cleans are standard targets (from googgle):
> http://www.ccp4.ac.uk/peter/programming/makefile-targets.html
>
> And imo two additional 2 or 3 line targets at the tail of Makefile
> does not increase Amber's software complexity.
>
> The main issue is getting these to do what their standard names imply.
>
> Scott

I was wondering if I missed any e-mails on this discussion. After all
the talk, Makefile still has:
  o clean
  o netcdf_clean
  o uninstall
     + uninstall.serial, uninstall.parallel, uninstall.cuda
  o superclean (== uninstall)

While Makefile_at has:
  o clean
  o netcdf_clean
  o uninstall
     + uninstall_at, uninstall_libraries
  o superclean ( == uninstall + rm config_file)

According to Scott's link, we don't approach the supposed standards of
clean, realclean, distclean and uninstall, although our "superclean"
(which doesn't seem to be part of any standard or customary use)
apparently is the closest thing to the standard "distclean". Are there
any plans to work on that yet? Also, should we also have a
"distclean"(or superclean) in the amber's Makefile?

Cheers,
Gustavo.

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Mar 02 2010 - 08:30:02 PST
Custom Search