Fix pre-compile script calls on Windows build (#58) The windows build test on the test server doesn't seem to be updating options appropriately. It looks like the issue is that the *.sh files aren't being properly called -- adding "bash" to the front of them allows them to run properly.

• Summary
• integration.addsan 「1」
• integration.mpi 「1」
• scientific.FlexPepDock
• scientific.abinitio_RosettaNMR_pcs
• scientific.antibody_snugdock
• scientific.glycan_dock
• scientific.make_fragments
• scientific.mp_f19_decoy_discrimination.debug
• scientific.mp_f19_decoy_discrimination
• scientific.mp_f19_tilt_angle.debug
• scientific.mp_f19_tilt_angle
• scientific.mp_lipid_acc
• scientific.protein_data_bank_diagnostic.full
• unit.valgrind

Update SCons for compilation with Python 3.12 (#54) Python 3.12 removed the imp module, which means our current SCons 3.0.4 doesn't work with it. Upgrading to SCons 4 will fix this. However, this removes support for building with Python 2.7 -- I think at this point this is acceptable, as anyone compiling Rosetta will likely have a Python3 version installed for RFDiffusion, etc. anyway. Additionally, the error message printed is relatively user friendly: ``` scons: *** SCons version 4.2.0 does not run under Python version 2.7. Python >= 3.5 is required. ``` SCons 4.2.0 was selected as it's the last version which supports Python 3.5 -- I don't know how many people that will actually help, but we don't need to be bleeding edge. Some additional files were changed to address warning messages with a Python 3.12 build. (Note that the current external/scons-local directory was deleted and then re-created from the official download. No manual changes were made, save for deleting some unnecessary documentation files.)

• Summary
• integration.mpi 「1」
• integration.release_debug 「1」
• scientific.protein_data_bank_diagnostic.cif
• scientific.protein_data_bank_diagnostic.full
• scientific.simple_cycpep_predict
• unit.valgrind

Update benchmark scripts for Python3.12. (#51) Python 3.12 gets rid of the imp module. Moreover, it gets rid of imp.load_source() altogether. We thus need to change around how we do the magic 'import from parent directory' statement. There are also a few other things which were popping up warnings when I ran things locally.

• Summary
• integration.mpi 「1」
• integration.release_debug 「1」
• scientific.glycan_structure_prediction
• scientific.peptide_pnear_vs_ic50.debug
• scientific.peptide_pnear_vs_ic50
• scientific.simple_cycpep_predict.debug

Fix make_fragments and mp_f19_ddG_of_mutation scientific tests (#48) Attempt to fix some of the scientific tests having issues. * make_fragments - restore parameter for the install_dependencies.pl in tool submodule * mp_f19_ddG_of_mutation -- update matplotlib version (and version of python needed.)

• Summary
• integration.addsan 「1」
• integration.mpi 「1」
• scientific.antibody_h3_modeling
• scientific.make_fragments
• scientific.mp_f19_tilt_angle.debug
• scientific.mp_f19_tilt_angle
• scientific.protein_data_bank_diagnostic.cif
• scientific.protein_data_bank_diagnostic.full
• windows.build.debug
• windows.build.release

Update benchmark.py, add how-to-fix explanation to warning message (#50)

• Summary
• integration.mpi 「1」
• integration.release_debug 「1」
• scientific.mp_f19_ddG_of_mutation.debug
• scientific.mp_f19_ddG_of_mutation
• scientific.mp_f19_decoy_discrimination.debug
• scientific.mp_f19_decoy_discrimination
• scientific.mp_lipid_acc
• scientific.protein_data_bank_diagnostic.cif
• scientific.protein_data_bank_diagnostic.full
• windows.build.debug
• windows.build.release

Add ability to specify compiler version ranges in scons settings. (#36) The basic.settings file has had a limitation for a long while where any version-specific settings are tied to an exact version. That is, settings for gcc 9.0 can't also apply to 10.0 (or even 9.1). This makes things difficult when a potential range of different compilers need to be dealt with in the same fashion, especially in a forward-compatible way. (e.g. the lack of -ffor-scope on any GCC newer than 9.0, including ones not yet released) An adjustment to one of the build files allows us to use a pattern matching specification to add parameter blocks for ranges of compilers (e.g. cxx_ver:>9.0 for any compiler version greater than 9.0, or cxx_ver:>=5.2,<7.0 for compiler versions between 5.2 (inclusive) and 7.0 (exclusive).) Should address issue #29 and #28

• Summary
• integration.mpi 「1」
• ubuntu-20.04.clang.python310.scientific.mp_f19_ddG_of_mutation.debug
• ubuntu-20.04.clang.python311.scientific.mp_f19_ddG_of_mutation.debug
• ubuntu-20.04.clang.python312.scientific.mp_f19_ddG_of_mutation.debug
• ubuntu-20.04.clang.python310.scientific.mp_f19_ddG_of_mutation
• ubuntu-20.04.clang.python311.scientific.mp_f19_ddG_of_mutation
• ubuntu-20.04.clang.python312.scientific.mp_f19_ddG_of_mutation
• scientific.protein_data_bank_diagnostic.full
• ubuntu-22.04.gcc.python39.unit 「3」
• windows.build.debug
• windows.build.release

Update CONTRIBUTING.md (#47) Add forgotten link to the Github Workflow documentation

• Summary
• integration.mpi 「1」
• linux.gcc.cxx11thread.serialization.unit 「3」
• linux.gcc.unit 「4」
• windows.build.debug
• windows.build.release

cifparse: Debian missing unistd include (#38) Addresses Issue #31

• Summary
• integration.mpi 「1」
• integration.release_debug 「2」
• windows.build.debug
• windows.build.release

QOL update to the PerResidueEsmProbabilitiesMetric (#26) This short PR: * Adds the proper citation for ESM * Adds some checks for NaNs/Infs to the Perplexity/softmax calculation * fixes a typo in MIFST.cc * Prevents an error in SampleSequenceFromProbabilities when calling the SimpleThreadingMover without mutations * Fixes const behavior in SampleSequenceFromProbabilities --------- Co-authored-by: moritzertelt <moritz.ertelt@gmail.com>

• Summary
• integration.mpi 「2」
• integration.release_debug 「10」
• integration.tensorflow 「1」
• integration.thread 「1」
• integration 「8」

fix PyRosetta build on Ubuntu-22.04 (#45)

• Summary
• integration.mpi 「1」
• integration.release_debug 「1」
• scientific.glycan_structure_prediction
• scientific.make_fragments.debug
• scientific.make_fragments
• scientific.mp_f19_ddG_of_mutation.debug
• scientific.mp_f19_ddG_of_mutation
• scientific.mp_f19_decoy_discrimination.debug
• scientific.mp_f19_decoy_discrimination
• scientific.mp_f19_tilt_angle.debug
• scientific.mp_f19_tilt_angle
• scientific.mp_lipid_acc
• scientific.peptide_pnear_vs_ic50.debug
• scientific.peptide_pnear_vs_ic50
• scientific.simple_cycpep_predict.debug
• scientific.simple_cycpep_predict

rename torch extras and defines (#43) * rename torch extras and defines * fix integration tests

• Summary
• integration.mpi 「1」
• integration.release_debug 「1」
• scientific.protein_data_bank_diagnostic.full

Fix uninitialized variable in InterfaceScoreCalculator (#40) The native_ensemble_best_ variable was uninitialized, which could potentially cause issues for non-XML cases were it doesn't get explicitly set. Make sure we initialize it.

• Summary
• integration.mpi 「1」
• integration.release_debug 「1」