Kilian Weishaupt authored 7 years ago and Timo Koch committed 7 years ago

* the test failed compared to the ref solution generated on the old
  master
* the reason for this is the change of how rho in context of gravity
  is calculated

* using rho = 0.5*(elemVolVars[scvf.insideScvIdx()].density(phaseIdx)
            + elemVolVars[scvf.outsideScvIdx()].density(phaseIdx));

  instead of interpolating the value of rho onto the face made the test
  pass w.r.t the the old solution

* we prefer interpolationg rho, however, therefore the ref solution is
  updated
* using the interpolated rho also decreases the number of timesteps from 8 to 7
  (compared to the old master)

Kilian Weishaupt authored 7 years ago and Timo Koch committed 7 years ago

* Created on old master by changing the names in model.hh
* remove output of viscosity

[staggered][freeflow] Add class to calcute the fluxes over a plane

See merge request !668

* only works for staggered free flow

Feature/adpativity next

See merge request !542

throw statements are no const expressions

Publicly stating of numComponents and numPhases has to be implemented
anyway. Should be done already in all fluid systems.

throw statements are no const expressions

Publicly stating of numComponents and numPhases has to be implemented
anyway. Should be done already in all fluid systems.

gcc7.2 interpreted the brace initializer as initializer list and wrongly
constructed the pivar vector with only one entry. The solution in this
patch works for all tested compilers.

Dennis Gläser authored 7 years ago and Timo Koch committed 7 years ago

For this test one needs to be able to update the point source map from outside.
As a first workaround, the computePointSourceMap() function in the FvProblem
does not get a map as input anymore but takes its own map. We should discuss
how this should look in general.

Dennis Gläser authored 7 years ago and Timo Koch committed 7 years ago

The test in the old master used element-wise distributed permeabilities
and material law parameters. This makes a big difference for the solution
on adaptive grids with the box scheme. In order to get closer to the reference
solution obtained on master, we have to do the same here.

Dennis Gläser authored 7 years ago and Timo Koch committed 7 years ago

In adaptive grids it might be necessary to store more interaction volume
data. Memory reallocation must be avoided as the stored pointers lose
validity otherwise.

Dennis Gläser authored 7 years ago and Timo Koch committed 7 years ago

this will be necessary in the grid data transfer for adaptive 2p models

Martin Schneider authored 7 years ago and Timo Koch committed 7 years ago

The fvGeometry and volVars vectors need to be updated
before calling the indicator update function

Feature/tracer on next

See merge request !619