[makedoc]

When there are zero atoms in one of the DD domains,
some arrays are allocated to be zero sized and later
accessed with &array[0].

Whereas this is harmless, it triggers an out-of-boundary
error when plumed is compiled with --enable-debug-glibcxx

The size of those arrays is now guaranteed to be at least 1.

[makedoc]

This reverts commit 564bd00f.

(useless)

this required a few fixes:
- use export COLUMNS instead of stty
- install flex +universal

The latter is needed for some misterious reason. I think
it is a bug in the libmatheval port file

I changed the settings so as to use 10.11 on osx on non-macports builds.

Macports builds fail for some unknown reasons so I leave the with 10.9

[makedoc]

gaussian with grids, the check for sigma_min and
grid was not working properly

regtest reset

[makedoc]

the gaussian support was wrong

lapack and blas README files were not shown.

I guess this happened with some doxygen update that ignores README files.
Solved renaming to README.md.

[makedoc]

[makedoc]

[makedoc]

evaluate gaussian was using always the periodicity of argument 0
plus some reorganisation of the code

[makedoc]

It works creating a local Portfile so as to test
non released plumed versions.

[makedoc]

a message reported by ./configure was wrong

[makedoc]

I made it consistent with other libraries.

Now, plumed looks first for the required functions and, if not
found, adds a standard -l flag. This allows as in the other cases
to override the paths using the LIBS variable.

As a side effect, it might be that -lgslcblas does not get linked.
This could happen if linked blas already define a cblas interface
(as far as I know, this happens with mac for me and not with linux).

[makedoc]