On Sat, Feb 6, 2016 at 1:10 PM, zoran matovic <zmatovic.kg.ac.rs> wrote:
> Just recently I installed Amber (including pmemd.cuda(MPI)) as a root
> (using
> su - a; mode that Ross suggests). Everything went fine I mean all the tests
> passed running them as a root. I am running now all the Amber programs as a
> user. So far there were no problems. However if there are something we
> (users) should now here is the place for u people to let us know.
>
I'm not trying to assert that installing as root won't work, or that it
working is unlikely. All I'm saying is that it is largely untested and if
there is a mistake -- either in the build scripts or in user
input/environment setup, the worst-case side effect is a ruined operating
system. And the more commands you run as root, the more potential risk you
open yourself up to. For instance, suppose you run part of the build
script with the following command:
/bin/rm -f $(AMBERHOME)/bin/*
as part of some clean or uninstall rule. If AMBERHOME is not set, this
gets rendered as "delete everything inside /bin". If you run this as a
normal user, the OS laughs as it denies permission. If you run it as root,
have fun reinstalling your operating system. I'm not just making this kind
of thing up -- I have posted a famous example of this *actually* happening
before:
https://github.com/MrMEEE/bumblebee-Old-and-abbandoned/issues/123#issuecomment-1226289.
Of course a more likely side-effect is injecting some clutter, or deleting
some minor component, from important locations... but it's still quite
easily avoided by avoiding root, especially when it's easy to do so.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Feb 06 2016 - 19:00:03 PST
