hobbes1069 at gmail.com
Fri Oct 28 17:54:21 UTC 2011
On Thu, Oct 27, 2011 at 8:44 PM, David Gaden <davegaden at hotmail.com> wrote:
> Looks like I'm not going to be much help.
>>1. The build from source documentation is pretty bad. The "System
>>Requirements" need to be broken up between build requirements and
>>runtime requirements. I.e, do I need <blah> or <blah>-devel... It has...
> I don't know the full list. The distro install instructions list packages required for each supported OS, suggesting someone has the list you need.
Yup, just need to find it.
>>2. The documentation seems to have instructions that are not strictly
>>needed for installation, but I could just be confused.
> I don't see anything unnecessary in this: http://www.openfoam.org/git.php
That looks like a slightly cleaner version of what I was using from
the download link:
>>3. Wanting to be compiled in the location it's installed to is bad...
> There's still a lot of old-school left in OpenFOAM. Object files go into subdirectories within each library. The #include strategy is particularly stupid - each library requires a directory with shortcuts to all files in its scope. There are scripts that attempt to maintain these, but they don't always work, and the fail mode of name collisions is apparently to destroy both files.
>>4. Along those lines, is the entire source tree really needed during use?
> The library is broken down into multiple sub-libraries. It shouldn't bee too hard to pull it apart to make large, logically grouped extension packages, like openFOAM-core, openFOAM-chemistry, openFOAM-lagrange, and so on.
Not really the answer to what I was asking, but good to know. What I
meant is the way that the install works, or rather doesn't work.
There's no "make install" method, you extract the source from its
archive directly to the install location and build it in place, which
makes packaging difficult.
So my question really was: Is the complete source needed after
building the binaries and libraries? The current install method would
suggest it is.
>>Now more Fedora specific issues:
>>5. The ThridParty package is completely unacceptable. Anything in
>>there that isn't currently already in Fedora will need to be packaged
> Outside my realm of expertise - I don't know much about the third party stuff. Some are pretty standard: openmpi, cmake. But then again, some of them are pretty old too, and I guess we'd have to package these ourselves.
I think most of what's in the ThirdParty archive is already available.
I know openmpi and cmake are. I did a quick search for ParaView and
paraview-data.noarch : Data files for ParaView
paraview.x86_64 : Parallel visualization application
paraview-mpich2.x86_64 : Parallel visualization application
paraview-openmpi.x86_64 : Parallel visualization application
I assume we would need paraview, paraview-data, and paraview-openmpi.
>>6. I'm not sure if the built-in wmake tool is acceptable, it seems to
>>come from openwatcom.org, it will also need to be packages separately.
> I think wmake is native to OpenFOAM - although openwatcom's description of theirs is identical. The wmake source files have OpenFOAM's standard comment headers. If you have to get rid of it, you are right about FreeFoam. Last I heard, the project is still unfinished, but the guys who manage the OpenFOAM-extend fork are borrowing FreeFoam's cmake solution for their Windows port. I understand the Windows port is a huge struggle, and is also unfinished... so I'm not sure how much of the struggle is associated with the cmake aspect.
I'm not a cmake expert but as a packager it's my preferred building
tool. If there was a good windows build and install file I could
probably convert it for *nix.
>>7. The 'copy the entire source-tree with the compiled binaries and
>>libraries' issue from #3. I'm not sure it will be acceptable.
>>Now once all those are taken care of, all the 3rd party stuff will
>>have to be ripped out during build to make sure it's not used and the
>>makefiles patched to use the system libraries where necessary.
>>All-in-all, a lot of work. Do any of the recent changes solve any of
> I didn't realize how much work would be involved. I'd say we give it a little more time to mature. The community is growing, and there's pressure to modernize the project.
Works for me, especially since I have no use for it but just like
packaging software. This piqued my interest because my major professor
and some of the graduate students at my alma mater were doing a lot of
More information about the users