[Paraview] MangledMesa & VTK --> we use this

Hank Childs childs3 at llnl.gov
Wed Jan 3 12:27:53 EST 2007


Hi Folks,

I don't think I've seen anybody on the list say "we use this", so I wanted to 
step up to Ken Moreland's "call for relevancy" from yesterday.

Our fundamental problem is having a single binary that sometimes runs with 
Mesa and sometimes runs with GL.

Here's why:
- I compile a binary called VisIt (http://www.llnl.gov/visit) and install it 
onto a shared file system.
- VisIt's parallel server is executed on different parallel machines from this 
single shared file system.
- Some machines have graphics cards and an X-server (obviously, for these, we 
do not use mangled Mesa)
- Some machines have no graphics cards.  Worse, if memory recalls, they don't 
even have an X-server and they don't have a libGL.so that points to Mesa.  
So, for these, we use Mangled Mesa.


An alternative may be to have a script that sets up an LD_LIBRARY_PATH to put 
a non-mangled Mesa in the path for the machines without X, but we opted for 
using the built in VTK Mangled Mesa support.  

It has been a long time since we investigated this, but I also seem to 
remember that the mangled mesa path did a better job of not touching the 
X-server.  Knowing that the mangled Mesa types of the rendering modules are 
essentially the OpenGL ones with some #defines, this doesn't add up to me, 
but maybe Charles Law or Berk remembers something about this that I don't.  
(As I recall, we exchanged a lot of e-mails about this about five years ago, 
so everyone's memory may be fuzzy.) 

Berk commented that the VTK support wasn't the hard thing, it was getting 
recent versions of Mesa to build mangled.  We froze the version of Mesa that 
we use, so that hasn't been an issue for us.  But we plan on updating in the 
next six months or so, so this may be an upcoming issue for us...

-Hank


More information about the ParaView mailing list