[a_soli]

Dear all,

I am absolute OpenFOAM novice and my Linux skills are not great either. I need to install OpenFOAM 2.4.0 on Ubuntu 16.10 and I followed step-by-step this guide thought for Ubuntu 16.04:

https://openfoamwiki.net/index.php/I...u#Ubuntu_16.04

Now, everything is fine, until point 10.3 where I experienced some issues while building ParaView. OpenFOAM itself works just fine, as I could see from the execution of the "Lid-driven cavity" tutorial. Can someone help me? I am attaching the log.makePV

P.S. : I wasn't sure if this should have been posted here or in the ParaView section.

[rdbisme]

Originally Posted by a_soli

Dear all,

I am absolute OpenFOAM novice and my Linux skills are not great either. I need to install OpenFOAM 2.4.0 on Ubuntu 16.10 and I followed step-by-step this guide thought for Ubuntu 16.04:

https://openfoamwiki.net/index.php/I...u#Ubuntu_16.04

Now, everything is fine, until point 10.3 where I experienced some issues while building ParaView. OpenFOAM itself works just fine, as I could see from the execution of the "Lid-driven cavity" tutorial. Can someone help me? I am attaching the log.makePV

P.S. : I wasn't sure if this should have been posted here or in the ParaView section.

I think this is related to using Gcc6.

I'm not a cmake expert but in VTK/CMake/GenerateExportHeader.cmake cmake seems to check only for GCC versions up to 5.

    string(REGEX MATCH "[345]\\.[0-9]\\.[0-9]"
      _gcc_version "${_gcc_version_info}")
    # gcc on mac just reports: "gcc (GCC) 3.3 20030304 ..." without the
    # patch level, handle this here:
    if(NOT _gcc_version)
      string(REGEX REPLACE ".*\\(GCC\\).* ([34]\\.[0-9]) .*" "\\1.0"
        _gcc_version "${_gcc_version_info}")
    endif()

Try to use a lower gcc version (e.g. 4.8.2).
http://askubuntu.com/questions/26498...-and-g-version

[a_soli]

Thanks tidusuper91, I installed gcc and g++ 4.7 and I can now launch paraFoam

[UDS_rambler]

Hello Everybody,

I have installed OpenFoam 2.4 to run on a VirtualBox Ubuntu 16.10 machine. I compiled everything and testing icoFoam -help I did not get any error message, so I figured out that I compiled it properly.

Nevertheless, when I used HelyxOS, using simpleFoam solver, I realized that I did not complete the compiling, and some of the solvers work, some other does not.

I attach here the log file, maybe someone has an idea of what is wrong?

Thank you in advance,


Giuseppe

[wyldckat]

Greetings to all!

I've moved Giuseppe's post to this thread, because it's related to the same issue: Building OpenFOAM 2.4.0 in Ubuntu 16.10.

@Giuseppe: The problem is that GCC 6 is not fully supported to be used for building OpenFOAM 2.4.0. Specifically, the error was this:

ensight/part/ensightPart.C: In member function ‘bool Foam::ensightPart::isFieldDefined(const Foam::List<double>&) const’:
ensight/part/ensightPart.C:54:54: error: ‘isnan’ was not declared in this scope
             if (id >= field.size() || isnan(field[id]))
                                                      ^

In order to fix this, you will need to one of the changes mentioned here: http://bugs.openfoam.org/view.php?id=2041

I'll try to have instructions on the wiki later this week.

Best regards,
Bruno

[alexeym]

Hi,

In fact ::isnan problem is just a beginning of troubles with GCC 6 and OpenFOAM 2.4.x.

Then user will encounter troubles with optimisation of reference address comparison with NULL (it is assumed as always true) and all ifs bodies in, for example, polyMesh.C like:

    if (&points)
    {
        points_.transfer(points());
        bounds_ = boundBox(points_, validBoundary);
    }

are executed unconditionally, and it leads to rather funny things (sometimes solver runs, sometimes solver crashes). When this behaviour has reported for clang, bug was closed with comment "to much work to fix it in 2.4.x, fixed in 3.0 with nullObject".

IIRC after adding nullObject to 2.4.x and changing all "&points"-like code to "notNull(points)"-like, it was possible to get stable running build.

[wyldckat]

Greetings to all!

Step-by-step instructions on how to build OpenFOAM 2.4.0 in Ubuntu 16.10 are now available here: http://openfoamwiki.net/index.php/In...u#Ubuntu_16.10

Originally Posted by alexeym

When this behaviour has reported for clang, bug was closed with comment "to much work to fix it in 2.4.x, fixed in 3.0 with nullObject".

IIRC after adding nullObject to 2.4.x and changing all "&points"-like code to "notNull(points)"-like, it was possible to get stable running build.

Many thanks Alexey! I suspected there were more issues, but I was hoping that the other fix would be enough. I've also updated the http://openfoamwiki.net/index.php/In...ibility_Matrix wiki page accordingly, by placing a reference link to your post.

From what I could find, the closest related report and commits to what you're referring to are as follows:

• https://github.com/OpenFOAM/OpenFOAM...eb7eca92784929
• http://bugs.openfoam.org/view.php?id=1973

The commit seems to be the closest to your description, because the bug report is a lot more recent.
As for:

Originally Posted by alexeym

When this behaviour has reported for clang, bug was closed with comment "to much work to fix it in 2.4.x, fixed in 3.0 with nullObject".

Unfortunately there is no funding dedicated to maintaining the previous releases (at least as far as I know of), if there is a current release already under maintenance. And since this was a major change, it would need to go through the test/tutorial loop... which takes time to run and check


Best regards,
Bruno

[alexeym]

Hi,

Originally Posted by wyldckat

Unfortunately there is no funding dedicated to maintaining the previous releases (at least as far as I know of), if there is a current release already under maintenance. And since this was a major change, it would need to go through the test/tutorial loop... which takes time to run and check

And this is a consequence of no strictly (or at least somehow) defined release life cycle.

Surely one can always check commit activity on github; yet, for example, can we assume 3.0.x as dead since the last commit (https://github.com/OpenFOAM/OpenFOAM...ee3b73b0f308e4) in that branch was in August (and quite bizarre commit - fixes are for 3.0.x, version in header comment is 4.x)?

[wyldckat]

Quick answer (although somewhat late):

Originally Posted by alexeym

Surely one can always check commit activity on github; yet, for example, can we assume 3.0.x as dead since the last commit (https://github.com/OpenFOAM/OpenFOAM...ee3b73b0f308e4) in that branch was in August (and quite bizarre commit - fixes are for 3.0.x, version in header comment is 4.x)?

Sorry, mea culpa on that commit. This was because there was a weird issue between OpenFOAM 3.0.x and 4.x, reported here: https://bugs.openfoam.org/view.php?id=2179 - so the commit was a direct copy of the script files in 4.x to 3.0.x, for fixing the typo in said scripts.

Nonetheless, the commit before last was also bit strange as well... although it was to make things a bit more consistent between the versions available back then, I guess...

Either way, long term support is something I'm still yet to see to be done by any provider of OpenFOAM versions/forks.