Merge pull request #5049 from RosettaCommons/roccmoretti/disable_ligand_dock_performance
Disable the LigandDockProtocol performance test.
This test is one of the noisiest performance test. (Probably because it's attempting to do too much in each cycle.)
Additionally, it's testing an old, outdated protocol.
If we want to re-enable this, we probably want to do more granular testing (different tests for each step),
and test the most recent version of the protocol.
Merge pull request #5060 from RosettaCommons/aleaverfay/hotfix_from_4924
Add a missing inclusion to fix release unit test build
I had a unit test that was not compiling in release mode due to undefined symbols not caught in debug mode
Merge pull request #5027 from RosettaCommons/vmullig/check_cterm_amidation_patch
Debug the Cterm_amidation patch
A few users have complained about this. Adding unit test for desired behaviour. Will debug if unit test fails.
EDIT: Yes, the patch was messed up. It's fixed now, and the unit test will ensure that it stays fixed.
Merge pull request #4945 from RosettaCommons/jkleman/benchmarks
Jkleman/benchmarks:
fixing a bunch of scientific tests but there is more work to do. Just want master to be up to date so I can see what else needs fixing.
Lift uses of `make_shared` that cause MSVC to crash into their own functions (#4966)
Issue Details:
When `make_shared` is used via a using declaration as a default parameter value, it causes MSVC to crash (see https://developercommunity.visualstudio.com/content/problem/1170468/ice-when-using-make-shared-via-a-using-declaration.html).
Fix Details:
Move the use of `make_shared` from the default parameter value into a function, then call that function as the default parameter value.
Merge pull request #4939 from RosettaCommons/roccomoretti/filenotfoundfix
Fix bug in FileNotFound error.
We were putting the source file name in the error message, not the name of the missing data file.
Merge pull request #4928 from RosettaCommons/vmullig/tidy_up_clang_tidy_warnings
Removing unused values flagged by clang-tidy.
Note that the operator() called here has no side-effects; its ONLY effect is to return a value that's never used, so it's safe to delete it.
Additional notes added by Rocco:
Looking at the code and #4890, it looks like the removal is fine. The computation that was previously done with these variables has been moved to the output_interface_deltas().
Presuming the component energies sum to the total scores, then the transformation should be valid. (As the difference of the sums of the components should be equal to the sums of the differences in components.)
... Well, with the caveat that you're either not using the normalization function, or that the normalization function is a linear (not affine) scaling, and the sum of normalized components is equal to the normalized sum of components. -- But looking through the current options for normalization factors, that's the case. (And the normalization option isn't used much, anyway.)
Possible fix to stochastic failure in HRFDyamicsEnergyTests (#4910)
The one thing that I can see is that there's no check that the data read from each line of the initialization file can actually be interpreted as an int and a float (which the code expects). If this fails, the int and the float _should_ be set to 0 and 0.0, respectively, under the cxx11 standard, but earlier standards left the container variables unchanged. I'm wondering if that's what we're seeing here, resulting in stochastic behaviour since the int and the float are uninitialized.
I'm adding a check that the line was properly parsed, and taking out the blank line in the config file. I don't know that this will fix the stochastic failure, but it can't hurt.
Tasks:
- [x] Add the check.
- [x] Confirm that the unit test now fails consistently due to the blank line.
- [x] Remove the blank line in the config file.
- [x] Ensure that test passes.
* Remove blank line in test.
* Update tools repo to point at latest master.
Merge pull request #4761 from RosettaCommons/vmullig/all_to_all_pnear_analysis
Add PNear analysis for many states.
I'm using peptide structure prediction to train ML models that can recognize peptides that fold well. A single structure prediction run actually generates far more data than I'm using. It would be useful to compute and output PNear to all of the negative states in addition to PNear to the state that I'm designing for, to give me more data for training. This PR adds this functionality.
Tasks:
- [x] Add new option.
- [x] Register new option.
- [x] Have the new option do something:
- [x] Add a new step to the `HierarchicalHybridJDApplication` class for this computation:
- [x] Sends a request for each state in turn from the emperor node to all slave nodes.
- [x] Each slave node broadcasts that state to all other slave nodes.
- [x] All slave nodes compute RMSD to that state.
- [x] Emperor collects RMSDs up the hierarchy and carries out PNear calculation.
- [x] Repeat for each relevant state.
- [x] Write out PNear list at end.
- [x] Error if used with `-compute_rmsd_to_lowest option.`.
- ~~Check and address issue #4738~~ --> separate pull request.
- [x] Documentation.
- [x] Integration test.
- [x] Beauty.
Merge pull request #4797 from RosettaCommons/roccomoretti/address_unit_release
Fix the release-mode version of the EPRSpinLabelTest unit test
Convert an exact comparison of a real to a delta comparison.
Merge pull request #4746 from RosettaCommons/roccomoretti/icc_fix
Fix ICC test on test server
ICC 19 is having issues with the boost::accumulator usage, possibly due to the boost update. I'm not quite sure why it's having the issue, but we can work around it by removing the usage.