Don't write score file entries for unscored poses.
Such entries were previously incomplete and contained only total score = 0, and potentially some data attached to the job by movers and filters. This was problematic when a score file contained a mix of entries for scored and unscored poses.
The structure is still written, and as before with an empty energies table.
Unit Test Status: All Pass
Integration Test Status: Cosmetic file name changes to:
AnchoredPDBCreator
database_jd2_compact_io
database_jd2_io
InterfaceAnalyzer
InterfaceAnalyzer_bothpack
InterfaceAnalyzer_prepack
InterfaceAnalyzer_resfile
InterfaceAnalyzer_tracer
ld_converter
oop_create
remodel
rings
scaffold_matcher
unfolded_state_energy_calc
notify author
notify list [rosetta-logs@googlegroups.com]
Adding build support for clang 3.5 (Ubuntu 14.04).
Fixing warning: comparison of constant ... is always false [-Werror,-Wtautological-constant-out-of-range-compare].
notify author
notify list [rosetta-logs@googlegroups.com]
JobDistributor's main loop no longer persists a pose across jobs.
The "master pose" which is passed to the mover the JD runs on is no longer kept outside the main while loop's scope. This means it's reinitialized with each new pose, which makes this loop a bit more exception safe--and fixes a bug where exceptions that end with this pose in an invalid state were leading to skipped jobs.
notify author
notify list [rosetta-logs@googlegroups.com]
[minor update of ResidueTypeConstraintMover]
- requires 'AA_name3' only to run ResidueTypeConstraintMover instead of
requiring both 'AA_name3' and 'favor_bonus'
- should not change any integration tests
[minor update of SandwichFeatures]
- newly added write_resfile_to_minimize_too_much_hydrophobic_surface
- should not change any integration tests other than features and features_database_schema (if any)
notify author
notify list [rosetta-logs@googlegroups.com]
Fix two bugs with NCAAs and the PackerTask, change behavior of EMPTY
Adds a new Resfile action called RESET and changes behavior of EMPTY. Previously, once a position became a NCAA, it could not be mutated back to a CAA. Also, once a NCAA was enabled at a position, it could not be disabled. The RESET action enables all CAAs at a position like the default state when a ResidueLevelTask is created. This allows the PIKAA action to behave properly. The RESET action also disables any NCAAs that have been enabled, allowing you to reenable any you want turn on using the NC action. The behavior of EMPTY has been changed to also disables any NCAAs that have been enabled, allowing you to reenable any you want turn on.
To design both CAAs and NCAAs at the same position use RESET, for example...
5 A RESET NC HLU PIKAA AVL
To design just NCAAs use EMPTY, for example...
5 A EMPTY NC HLU
This fix has the packer tack comparing interchange ability group rather than AA type to determine if we are designing at a position. The hydroxyproline patches had to change slightly so that they did not produce a residue type with identical AA and varient types, I believe this was a bug. Additioanl syntactic changes to the packer task to get around some complicated pointer assignment when used with std::list::erase.
Unit test status: [Pass]
Integration test status: [1]
[posttranslational_io]: decrease in the number of rotmaers during packing from the HYP bug
notify author
notify list [rosetta-logs@googlegroups.com]
[minor update of SandwichFeatures]
- total update of resfile generation based on more accurate native
data set
- better namings of resfiles for easier output process
[integration tests]
- should not change
notify author
notify list [rosetta-logs@googlegroups.com]