Fix 28 mymodule bluegene

[heplesser]

This PR should fix #28 and also partially address trac.526 and trac.617. Specifically

• DynModule is completely removed, as they no longer differed in a meaningful way from SLIModule
• The last code bits left of the never fully implemented module unloading mechanism removed
• When compiling for BlueGene, user modules (based on MyModule) are only built as static libraries and are linked statically into NEST. Other platforms are not affected by this change

Potential reviews: @jougs @tammoippen @apeyser

[tammoippen]

The file sli/interpreter.cc does not conform to clang-format. Can you call clang-format on this file again.

[tammoippen]

The code looks good to me. +1 for merge after TravisCI is satisfied.

[heplesser]

@tammoippen I don't understand the formatting problems. On my machine, I get

[plesser@Hanss-MBP] ~/NEST/code/trunk/src/sli [fix_28_mymodule_bluegene*](0)$ clang-format interpret.cc > interpret.cc.clang
[plesser@Hanss-MBP] ~/NEST/code/trunk/src/sli [fix_28_mymodule_bluegene*](0)$ diff interpret.cc interpret.cc.clang 
[plesser@Hanss-MBP] ~/NEST/code/trunk/src/sli [fix_28_mymodule_bluegene*](0)$ clang-format --version
clang-format version 3.7.0 (tags/google/testing/2015-06-18)

and this line about which clang-format complains on Travis

 out << std::endl << msg << errorname;

seems perfectly fine to me.

I am also surprised that Travis only reports build 89.8 as failing, even though all other builds have the same code.

[tammoippen]

Unfortunately, this is due to the version difference: TravisCI uses clang-format 3.6 and on your Mac you use 3.7 (same with me). Also there is no flag in 3.7, which mimics the 3.6 behaviour.

I am also surprised that Travis only reports build 89.8 as failing, even though all other builds have the same code.

I am also surprised by this: all builds should have failed...

[jougs]

I'm also happy with the code, so +1 for merging. @tammoippen Can you please click the button if you think that the failing Travis is just a glitch in the format checker? Thanks!

[apeyser]

If this is a glitch, shouldn't this branch get pulled and run against clang-3.6 so that we don't introduce failing files into the stream?

[jougs]

@apeyser Sounds reasonable.

[tammoippen]

It seems TravisCI is using different versions of the build.sh script. In the failing build, the changes from #54 are included, which makes wrong formatted code fail. The other builds do not have these changes, but also report the wrong formatted code as before (but do not fail).

I agree with @apeyser, but do not have write permissions to @heplesser's branch.