Merge pull request #4692 from RosettaCommons/revert-4624-vmullig/threadsafe_approximate_buried_unsat_penalty
Revert "Make the ApproximateBuriedUnsatPenalty threadsafe" temporarily
Reverts #4624. This PR introduced a bug that I need to track down that's tripping a few people up. Sorry about that.
@bcov77
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #4691 from RosettaCommons/roccomoretti/fix_rotamer_uninitialized
Fix uninitailized variable in StoredRotamerLibrarySpecification.
The uninitialized energy can cause issues if not set later.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #4684 from RosettaCommons/yhsia/ddg_fix_3
Added threshold_min to ddgfilter. This allows the user to set a minimum ddG to the filter in addtion to the already existing threshold option that sets the maximum allowed value.
Added dump_pdbs option to ddg. When set to true, will dump the intermediate pdbs during the ddG run for debugging purposes.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #4680 from RosettaCommons/roccomoretti/params_out_spacing
Fix spacing in params file output.
The params file reader expects a fixed-column format for the atom names, but the current output is giving whitespace separated. This causes issues with CCD types, which don't have fixed-spacing atom names on read-in.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #4656 from RosettaCommons/roccomoretti/dunbrack_precision
Adjust Dunbrack library generation to address test server issues.
It looks like a fair part of the integration test instability on the test server may be due to differences in the Dunbrack rotamer library binary on different machines. (Potentially due to release/debug differences.)
A substantial portion of the difference in release/debug mode binary file generation is due to truncating the spline-fit derivatives to single precision. Changing that to double precision greatly reduces the numeric difference between debug-mode and release-mode generated Dunbrack binaries (at least on my machine).
To further prevent differences in compilation mode from altering integration test trajectories, the way the Dunbrack binaries are stored on the test server was altered, such that each compilation mode gets its own Dunbrack binary.
Both of these changes should hopefully greatly reduce the spurious integration test diffs we've been seeing on the test server.
The Dunbrack binary version has been increased to accommodate for the change in precision. This will result in regenerating the Dunbrack binary.
notify author
notify list [rosetta-logs@googlegroups.com]