Merge pull request #1229 from RosettaCommons/everyday847/cleanup
Trivial, semantic-neutral code cleanup effort to core/scoring: I think it makes sense to move forward with this PR: even the scorefile scores, in that one integration test that appears to just get optimized differently, end up the same.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1148 from RosettaCommons/everyday847/dunbrack_cleanup
Clean up core/pack/dunbrack - If I want to do more work on this, I'll do it in a separate PR. May as well get things looking better asap.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1236 from RosettaCommons/roccomoretti/silenttagfile
Add support for -in:file:tagfile to SilentFileInputter.
he relationship between -in:file:tags and -in:file:tagfile is similar to -in:file:s and -in:file:l -- that is, you can now pass a file with a list of tags, rather than having to list all the tags out on the commandline (and without the length limits doing so imposes).
The -in:file:tagfile is a pre-existing flag, only used once in AbrelaxApplication.cc, but with similar semantics. This extends the functionality to JD2 protocols. (Also fixes the option file description.)
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1211 from RosettaCommons/JWLabonte/XRW/HETNAM_records
Post-XRW: Refactoring reading, writing, and storage of HETNAM records
This merge completes a task having been on my to-do list for a long time — a refactoring of `HETNAM` code in Rosetta to be more general / less carbohydrate-centric.
Now anyone should be able to specify using `HETNAM` records the desired base `ResidueType` for a specific three-letter-code at a specific position in the structure. (This was currently only functional for three-letter-codes that are listed in the database as carbohydrate codes.)
A clear example is now found in `source/test/core/import_pose/3-letter_code_ambiguity_resolution.cxxtest.hh
`
Now, Rosetta-format `HETNAM` records are output by default for all non-AA, non-AA, non-virtual, and non-membrane residues. This will change ~50 integration tests.
There is a known issue with `HETNAM` records involved in branched complexes being sorted poorly. I will fix this in a later merge, and it currently only affects sugars.
A flag, `use_pdb_format_HETNAM_records`, can be used to have the Rosetta output PDB-format HETNAM records instead, but for this to be truly useful, we need to have the PDB name for each 3-letter code stored somewhere in Rosetta. (Perhaps a future pull request?)
This merge also cleans up / simplifies `SFR` code that was a mess because about six of us were working on it at the same time during the ChemicalXRW.
(As an extra bonus, it also introduces the (`molfile_to_param.py`-generated) oxidized and reduced versions of flavin into the database.)
All unit tests pass.
Windows PyRosetta test not completed. (See Sergey's e-mail.)
~50 integration test changes from `HETNAM` records in output files
carbohydrates test change from new sorting of `HETNAM` records.
number_of_residuetypes test change from adding two more
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1232 from RosettaCommons/vmullig/symmgly_uninit
Fixing a beta_nov15 issue with symmetric glycine Ramachandran tables
Fixing an issue with symmetric glycine Ramachandran tables and the Shapovalov option: if the option is true (as it is for beta_nov15), then only the data for ss=3 have been loaded, in which case we should only symmetrize those tables (since everything else is uninitialized).
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1231 from RosettaCommons/vmullig/minimizer_uninit
Fixing major uninitialized variable issue unique to lbfgs minimization
This illustrates why, when we change a convention or a Rosetta default behaviour, we REALLY need to make sure that we test the effects of the change very carefully. The switch to lbfgs as default caused a whole host of new Valgrind errors which were previously not revealed because tests were previously not invoking the lbfgs code. This, at least, should fix those issues.
Very minor trajectory changes observed (and expected) in those integration tests that were throwing Valgrind errors. These are probably associated with the first line minimization during the lbfgs minimization trajectory.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1234 from RosettaCommons/vmullig/hg_issue
Small bug in simple_cycpep_predict fixed
Disulfides were not being properly rebuilt following cyclic depermutation.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1226 from RosettaCommons/tlinsky/matchermover_bug
Fix bug where matcher crashed due to no neighbor graph in the pose
Integration test change expected in inverse_rotamer_remodel test due to minor tracer output change
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1227 from RosettaCommons/vmullig/cyclic_packing_bug
Fix a bug in packing cyclic peptides when using beta_nov15
There's an MxN array that should be an NxM array, which throws a debug assertion error.
Tasks:
-- Add unit test to illustrate/catch the problem.
-- Figure out what's causing the problem.
-- Fix it.
-- Beauty.
Update: I think the problem was that a particular array was being set up that was intended to be NxM, where N is the number of rotamers for residue i and M is the number of rotamers for residue j, and i is strictly less than j. (This was for bonded pairs of residues). In the cyclic peptide case, at the cyclization point, i was greater than j, and the NxM matrix had the dimensions swapped. I've added logic to check whether i or j is greater and to swap the indices accordingly.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #1222 from RosettaCommons/jadolfbr/fix_rsd_copy_op
Remove RT copy operation from DisulfideAtomIndices
Fix ResidueType access in DisulfideAtomIndices where we are getting an actual copy of the residue type from a residue instead of a const reference as we should be. This should speed up code a little bit when doing much residue deletion and additions.
Identified by the Time Profiler instrument on a mac during carbohydrate glycosylation profiling. If you have a mac, and want to profile code (number of calls/time spent in particular places, etc.), its super awesome and easy to setup.
notify author
notify list [rosetta-logs@googlegroups.com]