@gtribello I have to revert this since it fails on travis.
Can you please fix it and add it again on v2.4 (then merge, as you did)?

Thanks!

This reverts commit ce4bbf4e.

Check that METAD + TARGET + GRID_SPARSE works correctly.

@chi2, @chi3, @chi4 and @chi5

now an option to avoid clearDependecies.
This allows to have in a single class actionAtomistic and actionWithArguments
and actually use them together

this is used in Action build on MetainferenceBase where requestAtomst(atoms, false)

This add some incompatibility with all former implementations but is far easier to use

* restyling for the reweighting c(t) calculation

* fixed bug in multidimensional grid integration

* fixed regtest for 2D reweighting

* changed reweighting keywords and some doc

@tetsuyamorishita

Regtests

See #422

See #399

and improved regtests

I changed Plumed::invalid() to Plumed::makeInvalid() and added a new function
Plumed::makeValid(). These names are more inline with stdlib names
(e.g. make_shared), though with our conventions for capitalization.

Hopefully there's no code yet using this (beta) interface.

In addition, I made it possible to use a macro to change the behavior
of the default constructor to makeInvalid(). This might be
useful in python wrappers to avoid opening the library twice
(python objects always use the "nullary" - default - constructor first).

now covered by the regtest better

and SCEXP to SCALEINT

todo:
- gpu should only be run by master
- add documentation

but results are very different on cpu or opencl backends

in particular a gpu version of SAXS is added in ISDB
this is just a SAXS collective variable implemented to run on GPU
a regtest is added. This was implemented with an old version of
arrayfire so it should be checked wheter it can be made more modern
a regtest is also added

This bug was reported by Carol Simoes and fixed by Javi Iglesias.

The fix requires resetting a regtest.