Fix off-by-one loops over 1-indexed residue/atom/element ranges

[lyskov-ai]

Summary

Audit-pass fix for a family of off-by-one bugs of the same shape across 35 files. Continues the pattern from #695 and #701; this batch covers loops the earlier audits did not reach.

Off-by-one in 1-indexed loops (33 files)

Each loop uses for ( Size i = 1; i < container.size(); ++i ) (or < pose.size() / < pose.total_residue() / < natoms() / < nres) and the body uses i as a direct 1-indexed element accessor — pose.residue(i), vec[i], xyz(i), etc. The last element is therefore silently skipped. Replaced with <= to include it.

Affected functions cover residue iteration in:

• core/energy_methods/RNA_LJ_BaseEnergy.cc (RNA neighbor scan)
• core/io/pose_to_sfr/PoseToStructFileRepConverter.cc (SS element enumeration — a 1-residue SS element at the C-terminus was dropped)
• core/pack/interaction_graph/SurfacePotential.cc ×3 (neighbor count, surface energy sum)
• core/pose/rna/util.cc (heavy-atom contact scan)
• core/select/residue_selector/JumpUpstreamSelector.cc (subset assignment, divergent from sibling JumpDownstreamSelector which uses <=)
• protocols/antibody/residue_selector/CDRResidueSelector.cc
• protocols/canonical_sampling/mc_convergence_checks/HierarchicalLevel.cc (address match count)
• protocols/cartesian/md.cc ×2 (per-atom state save/derivative check)
• protocols/cutoutdomain/CutOutDomain.cc (find_nearest_res)
• protocols/denovo_design/components/FoldGraph.cc
• protocols/docking/DockingEnsemblePrepackProtocol.cc ×2 (chain identity validation)
• protocols/docking/metrics.cc ×2 (cutpoint scan in Irmsd / contact metrics)
• protocols/enzdes/EnzRepackMinimize.cc (movable-residue collection for backrub)
• protocols/features/strand_assembly/StrandBundleFeatures.cc ×2 (per-atom distance scan)
• protocols/fldsgn/MatchResidues.cc ×2 (RMSD/superimpose atom-id map build)
• protocols/frag_picker/GrabAllCollector.hh (clear)
• protocols/frag_picker/VallProvider.cc (find_chunk)
• protocols/frag_picker/quota/QuotaCollector.cc ×5 (per-position pool enumeration)
• protocols/frag_picker/scores/AtomBasedConstraintsScore.cc ×2 (constrainable_atoms map fill, per-row state init)
• protocols/jd3/JobGenealogist.cc
• protocols/legacy_sewing/conformation/Assembly.cc ×2 (pose_loop_anchors, disconnected_segments — the last segment, which by definition has no next, was being skipped — exactly the case the predicate !has_next(i) is meant to catch)
• protocols/loops/util.cc ×3 (non-protein-chunk and per-loop accounting)
• protocols/match/output/UpstreamDownstreamCollisionFilter.cc
• protocols/membrane/MPLipidAccessibility.cc ×2 (slice iteration)
• protocols/moves/PyMOLMover.cc (relevant_residues mask init — last entry left default false despite the intent of "all relevant")
• protocols/pose_metric_calculators/DecomposeAndReweightEnergiesCalculator.cc
• protocols/pose_metric_calculators/SurfaceCalculator.cc (per-residue summary string)
• protocols/pose_sewing/movers/BlockwiseAnalysisMover.cc ×2
• protocols/pose_sewing/movers/OmnibusDisulfideAnalysisLabelerMover.cc ×3
• protocols/rna/movers/RNAIdealizeMover.cc (CoordinateConstraint application)
• protocols/splice/SampleRotamersFromPDB.cc ×2 (rotamer-deduplication scan)
• protocols/splice/SpliceManager.cc (PSSM row check — comment says "go over all the PSSM segments")
• protocols/stepwise/legacy/modeler/rna/StepWiseRNA_WorkingParametersSetup.cc
• protocols/unfolded_state_energy_calculator/UnfoldedStateEnergyCalculatorMover.cc (protein-residue count)

Self-distance copy/paste bug (folded into 2 of the above files)

pose.residue(upstream_res).xyz(2).distance(pose.residue(upstream_res).xyz(2)) is identically 0, so the <= crit_dist_ gate was always satisfied — the distance filter was a no-op. Fixed to .distance(pose.residue(downstream_res).xyz(2)) in:

• protocols/pose_sewing/movers/BlockwiseAnalysisMover.cc
• protocols/pose_sewing/movers/OmnibusDisulfideAnalysisLabelerMover.cc

What was deliberately NOT touched

The audit also turned up many candidate matches that are actually intentional and were left alone:

• Loops with explicit pair access vec[i+1] / pose.residue(i+1) (need to stop at size-1 to stay in bounds — e.g., is_cutpoint scans, SS-transition checks, adjacent-distance checks).
• Membrane-pose iterations that intentionally skip the trailing MEM virtual chain.
• "Separator" output patterns (for i<size: print sep; print last) — e.g., protocols/jd3/InnerLarvalJob.cc, protocols/ligand_evolution/Scorer.cc.
• Suite / connection enumerations where N segments produce N-1 connections by definition.
• Variant-by-cutpoint scans where the variable indexes a between-residue position.
• Self-pacing decompositions that already handle the last index in a special branch after the loop.
• Loops where the analysis script flagged a 0-indexed std::vector body that is in fact correctly bounded (e.g., protocols/cluster/calibur/Clustering.cc, protocols/cluster/cluster.cc — cluster center at [0] plus members at [1..N-1]).
• Commented-out code (in SurfaceInteractionGraph.hh, RNAIdealizeMover.cc, etc.).

Debug build (mode=debug) is clean.

[roccomoretti]

Looks reasonable in principle, but it looks like there's a number of test changes which need to be tracked down to see if they're actual bug fixes, or if we missed some of the deliberately-skipped-the-last-entry cases.

[lyskov-ai]

Closing in favor of splitting this 35-file batch into four smaller, per-subsystem PRs so each can be reviewed and its regression-test impact assessed independently (some of these off-by-one < → <= changes are results-affecting, which is a show-stopper for a single omnibus PR):

• Fix off-by-one in 1-indexed loops in core/ (split 1/4 of #707) #730 — core/ (5 files) — no changed regression tests expected in this subsystem
• Fix off-by-one in 1-indexed loops in frag_picker + accounting (split 2/4 of #707) #731 — frag_picker + job/metric accounting (9 files) — touches unfolded_state_energy_calc
• Fix off-by-one loops + self-distance bug in SEWING/splice/de novo/RNA (split 3/4 of #707) #732 — SEWING / splice / de novo / RNA (10 files), incl. the self-distance copy/paste bug fix — touches pose_sewing
• Fix off-by-one in 1-indexed loops in docking/membrane/misc (split 4/4 of #707) #733 — docking / membrane / antibody / match / loops / misc (11 files) — touches enzdes, inverse_rotamer_remodel, ligand_dock_cholesterol, mp_f19_relax, mpil_load_implicit_lipids

The union of the four PRs is byte-identical to this branch (verified as an exact file partition, no overlaps). The results-affecting changes are concentrated in #733 (and partly #731/#732); #730 should be inert w.r.t. the changed regression set.