VMD-L Mailing List

From: Marc Baaden (baaden_at_smplinux.de)
Date: Tue Feb 14 2006 - 01:27:51 CST

Hi,

just my 2 eurocents on having several VMD versions and on Linux build
systems..

We have found in the past that it often helps the binary portability of
a Linux (but probably also unix in general) application to ship some or
all libraries it depends on with it. The libs can be put in some application
dependent directory and if needed they can be used preferentially over
local libraries by making the directory path the first entry in the
LD_LIBRARY_PATH environment variable. Alternatively they can be used only
for libraries missing on the system by appending them to this variable.
(The quick hack we regularly use is if application X runs on machine Y
 but not on machine Z with a different Linux distribution and if you can
 identify the library that leads to problems, you just need to copy the working
 one from machine Y using the above procedure to make it active without
 generally changing any system libraries)
Maybe this would suffice to make a single VMD binary work on most Linux
distributions.

As to Linux build systems, from personal experience I can strongly recommend
Debian (or Debian-based) distributions. Debian has been ported to many (~10
stable and 5 in development) architectures [1] and having an application build
properly on one of those thus buys you ~15 builds in one go...
It also makes it easy to give instructions for people who want to build from
source. Debian has also been used to provide specific environments (eg via live
CDs) for bio/chemical software [2], scientific computing [3] or medicine
(including bioinformatics) [4] so all libraries or tools one usually needs are
already in there (which helps for build dependencies). There are many other
advantages and strong points, but they may not be so relevant to eg building
VMD.

Marc

[1] http://www.debian.org/ports/
[2] http://www.vigyaancd.org/
[3] http://dirk.eddelbuettel.com/quantian.html
[4] http://www.debian.org/devel/debian-med/

>>> John Stone said:
>>
>> Bogdan,
>> I could certainly setup a VMware build box. Since I'm the only
>> one doing all of this work, I don't really have the time to build
>> any more linux versions of VMD. We make the source code available,
>> so people running versions of linux that are far behind or far ahead
>> of what I'm running here can build their own VMD binaries. It takes
>> a lot of testing to determine that a build of VMD is even worth making
>> available, and so one of the problems with adding VMware and several new
>> linux versions is that it assumes I'll have time to test these additional
>> builds, but in reality I'm already maxed out desparately trying to keep
>> up with incoming VMD related emails and testing the builds I already
>> provide. I would definitely like to make it easier for people to
>> build VMD from source code, so I'm willing to accept any suggestions
>> that would make the source distribution much easier to use.
>>
>> John
>>
>> On Thu, Feb 09, 2006 at 05:51:55PM +0100, Bogdan Costescu wrote:
>> > On Thu, 9 Feb 2006, John Stone wrote:
>> >
>> > > The 32-bit builds of VMD are done on a Redhat 9 box, which is as old
>> > > as I can go (and even this, I had to beg our sysadmin not to upgrade
>> > > it even with the obvious security issues that RH9 has).
>> >
>> > How about changing to CentOS-3 ? This costs nothing, is very similar
>> > to RH9 (IMHO RHEL3 is some cross between RH9 and FC1) and is updated
>> > by the CentOS team including security-wise. Plus that Tru is also part
>> > of the CentOS team, specifically the CentOS-3 part :-)
>> >
>> > If you really want to go even more backwards, then how about
>> > CentOS-2.1 ? This is based on RHEL-2.1 sources which are based on the
>> > RH 7.x releases which seem to still be preferred by a lot of people.
>> >
>> > [ Disclaimer: I am not affiliated with the CentOS project. I am only a
>> > user. ]
>> >
>> > > I may have to distribute two different 32-bit builds of VMD 1.8.4,
>> > > one build on RH9, and one built on RHEL4.
>> >
>> > I'm actually contemplating putting all 3 present CentOS versions in
>> > VMWare "containers" to allow me to compile using the different
>> > libraries and compilers and compare the results; in my case, there is
>> > no graphics involved, so I could even run in these virtual machines.
>> > For graphics (as in the case of VMD using OpenGL), this is still
>> > different than running in these 3 environments on real hardware, but I
>> > think that it is better than compiling in one and just hope for the
>> > best in the others.
>> >
>> > With VMWare (or some other virtualization software), you don't need to
>> > ask your sysadmin for anything. You can be the syadmin... if these are
>> > only used for compilation tests, then they don't need any complex
>> > configuration.
>> >
>> > --
>> > Bogdan Costescu
>> >
>> > IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
>> > Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
>> > Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
>> > E-mail: Bogdan.Costescu_at_IWR.Uni-Heidelberg.De
>>
>> --
>> NIH Resource for Macromolecular Modeling and Bioinformatics
>> Beckman Institute for Advanced Science and Technology
>> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
>> Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
>> WWW: http://www.ks.uiuc.edu/~johns/ Fax: 217-244-6078
>>

Marc Baaden 

-- 
 Dr. Marc Baaden  - Institut de Biologie Physico-Chimique, Paris
 mailto:baaden_at_smplinux.de      -      http://www.baaden.ibpc.fr
 FAX: +33 15841 5026  -  Tel: +33 15841 5176  ou  +33 609 843217