On Wed, 2014-10-01 at 11:21 -0700, Ross Walker wrote:
> Hi Jason,
>
> Again I reiterate my point. Why are we requiring lots of non-default
> packages to be installed?
>
> Note - a LOT of people have to build AMBER on machines they do not have
> root for. They also have to build it on non-Linux systems - such as AIX.
> Thus I would suggest getting a machine and doing a default install - with
> most Linux systems the packages needed for Xleap etc are there. If they
> are not then right now they get a message asking them to specify -noX11.
> With this setup they will then get another message telling them to specify
> -nopysander - at this point the poor user is like - WTF?? what are all
> these things I am turning off, do I need pysander? What will it break etc
> etc etc... Bad news... Probably we should just have it automatically skip
> non-core packages if it does not find the relevant libraries.
>
> As you had it the configure script would fail with a FATAL error if one
> did not have python-dev installed. This requires root access to install.
>
> I would suggest changing this to be non-fatal - if it does not detect
> python-dev it just skips the building of pysander. This removes the fatal
> dependency on the external package rather than requiring the user to
> explicitly specify -nopysander.
This comes back to the question of when is it better to find out you're
missing a dependency -- at configure time or when you decide to run it?
We (amber devs) had this discussion several years ago about xleap and
the X11 headers, and the people that spoke said that configure-time
notice was better. That's why Tyler implemented that change.
Keep in mind that the "standard" Linux user is more likely to be using a
machine that has the Python headers and development libraries installed
than they are a machine that has gfortran installed... out of the box at
least. And python is deployed via buildbots on all kinds of OSes (AIX
included), so portability is not an issue here.
All the best,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Oct 01 2014 - 12:00:03 PDT
