# v5.8.0 release notes¶

Clawpack 5.8.0 was released on February 4, 2021. See Installing Clawpack.

Permanent DOI: http://doi.org/10.5281/zenodo.4503024

Changes relative to Clawpack 5.7.1 (Sept. 11, 2020) are shown below.

To see more recent changes that are in the the master branch but not yet released, see Changes to master since v5.8.0.

## Changes that are not backward compatible¶

• For AMRClaw and GeoClaw, the data file amr.data now created from setrun.py now includes an additional line with the parameter memsize specifying the initial length of the alloc array used for allocating memory to patches when adaptive refinement is used. This can be specified in setrun.py by setting amrdata.memsize. If it is not set, then default values are used that are similar to past default values; see Specifying AMRClaw run-time parameters in setrun.py. So this is backward compatible in the sense that no changes to setrun.py are required, but the old amr.data files will not work so you may need to do make data to create a new version.

• In GeoClaw, refinement “regions” can no longer be specified implicitly when listing a topo dtopo or qinit file. See the geoclaw section below. Note: You may need to explicitly declare new regions or flagregions to produce the same behavior as in past versions of GeoClaw.

• The GeoClaw transverse Riemann solver rpt2_geoclaw.f has been improved and results in slightly different computated results in some cases. For more details see the riemann and geoclaw sections below.

• For AMRClaw and GeoClaw, an additional short array is saved in a checkpoint file for use in a restart. Due to this change, a checkpoint file created using a previous version of Clawpack cannot be used for a restart with the new version.

## General changes¶

The travis tests that automatically run on pull requests no longer test using Python2, only Python3. See Dropping support for Python 2.7.

## Changes to visclaw¶

• ClawPlotAxes.skip_patches_outside_xylimits does not work properly if there is a mapc2p function defining a grid mapping, so it is now ignored in this case.

## Changes to riemann¶

• The GeoClaw transverse solver rpt2_geoclaw.f was modified to fix some long-standing bugs and change some of the logic.

The new version gives slightly different results on most problems, but extensive testing indicates the new results are at least as good as the old. The new version has also been refactored to make the logic clearer and to avoid some unnecessary work, and generally runs faster. In some cases where instabilities had been observed in long-duration runs (particularly for storm surge), the new version appears to provide better stability.

In particular, the left- and right-going waves are now split up transversely using states in the cell to the left (resp. right) in which the splitting is performed, rather than using Roe averages based on the cell from which the wave originates.

## Changes to amrclaw¶

• An additional short array is saved in a checkpoint file for use in a restart. Due to this change, a checkpoint file created using a previous version of Clawpack cannot be used for a restart with the new version.

• A memsize parameter can now be set in setrun.py, see above and Specifying AMRClaw run-time parameters in setrun.py.

• src/2d/prepc.f was improved to use less storage from the work array alloc that is used for memory allocation for AMR patches. For large-scale problems this can be a substantial savings and allow running larger problems.

## Changes to geoclaw¶

Several changes were made to fix long-standing bugs. These fixes lead to slightly different results than those obtained with previous versions of GeoClaw. In all the tests performed so far the changes are minor and it is thought that the new version is at least as accurate as the old version. Please let the developers know if you run into problems that may be related to these changes.

• In filpatch.f90: The slope chosen for interpolating from a coarse grid to the ghost cells of a fine-grid patch had an index error that could affect the sign of the slope used in momentum components. Also slopes were not always initialized to zero properly at the start of a loop

• Some index errors were fixed in fgmax_interp.f90.

• Changes to riemann/src/rpt2_geoclaw.f90. These cause some change in results but tests have shown the new results appear to be at least as good as previous results and the code may be more stable in some situations. For more detail see the “Changes to riemann” above.

• The new flagregions introduced in v5.7.0 (see Specifying flagregions for adaptive refinement) were not implemented properly in GeoClaw, and in some situations refinement to a maxlevel that was indicated only in flagregion was not allowed as expected. This is now fixed.

• In previous versions of GeoClaw one could implicitly define AMR flag regions that are aligned with the spatial extent of topo, dtopo, or qinit files by specifying minlevel, maxlevel (and in the case of topo files, a time interval t1, t2) when the file name is given. This feature did not always work as advertised and was often confusing. If these values are specified then they are now ignored, as explained in more detail in the following items. Not that you may have to explicitly declare new flag regions now in order to have the expected refinement regions.

• When specifying topo files in setrun.py using the format:

[topotype, minlevel, maxlevel, t1, t2, fname]


the values minlevel, maxlevel, t1, t2 will now be ignored. To avoid warning messages, instead specify:

[topotype, fname]

• When specifying dtopo files in setrun.py using the format:

[topotype, minlevel, maxlevel, fname]


the values minlevel, maxlevel will now be ignored. To avoid warning messages, instead specify:

[topotype, fname]

• When specifying qinit files in setrun.py using the format:

[minlevel, maxlevel, fname]


the values minlevel, maxlevel will now be ignored. To avoid warning messages, instead specify:

[fname]

• A memsize parameter can now be set in setrun.py, see above and Specifying AMRClaw run-time parameters in setrun.py.

• An additional short array is saved in a checkpoint file for use in a restart. Due to this change, a checkpoint file created using a previous version of Clawpack cannot be used for a restart with the new version.

## Changes to PyClaw¶

For changes in PyClaw, see the PyClaw changelog.

See pyclaw diffs

# Other Clawpack Repositories¶

The repositories below are not included in the Clawpack tarfile or pip install, but changes to these repositories may also be of interest.