Merge remote-tracking branch 'origin/main' into feat/matrix-free-obj-constr (fixing clobbered workflow updates)

[gennadiryan]

DTO Issue #113 implementation summary (v1)

Status: Implemented and tested on 2026-06-30 against
DirectTrajOpt.jl@feat/matrix-free-obj-constr. Not yet committed.
Test log archived at ../../tmp/knot_hvp_run1.log.

Companion artifacts:
./dto-matrix-free.md — original design doc;
this summary covers Interface A (KnotHVP) only.

---

Test results

Test	Pass	Total
trait defaults to nothing for every objective	6	6
ConstantLowRankHVP round-trips via KnotPointObjective	3	3
CustomKnotHVP round-trips via KnotPointObjective	4	4
TerminalObjective threads knot_hvp keyword	2	2
default field value is nothing (no-regression smoke)	7	7
Total	22	22

Wall time 4.1s; no warnings, no errors.
17:41:42 [39/1898]

• NEW src/objectives/knot_hvp.jl — KnotHVP abstract,
  ConstantLowRankHVP(A::Matrix{Float64}, core::Symbol),
  CustomKnotHVP(apply!::Function, on_device::Bool), generic
  knot_hvp(::AbstractObjective, ::NamedTrajectory) = nothing, and the
  5 inline @testitem blocks.
• src/objectives/_objectives.jl — exports of KnotHVP,
  ConstantLowRankHVP, CustomKnotHVP, knot_hvp (4 lines), and
  include("knot_hvp.jl") placed before knot_point_objectives.jl
  so KnotHVP is in scope when the struct's field type is parsed.
• src/objectives/knot_point_objectives.jl — added
  knot_hvp::Union{Nothing,KnotHVP} field to the struct,
  knot_hvp=nothing keyword on the canonical outer constructor
  (Outer 1 at :68), threaded through to the inner. The other three
  outer constructors and both TerminalObjective variants forward
  kwargs... so they pick up the keyword automatically. Added the
  trait specialization
  knot_hvp(obj::KnotPointObjective, ::NamedTrajectory) = obj.knot_hvp
  next to Base.show. Docstring updated to replace the phantom
  ∂²Ls line (a doc artifact from the 6aeceee refactor — never an
  actual field) with the real knot_hvp field description.

---

Acceptance criteria (issue #113)

• KnotHVP, ConstantLowRankHVP, CustomKnotHVP defined beside
  get_full_hessian.
• Optional knot_hvp::Union{Nothing,KnotHVP} = nothing on
  KnotPointObjective; threaded through TerminalObjective via
  kwargs....
• Trait with generic default + KnotPointObjective specialization.
• Docstrings define the contract: data layout, core symbol
  convention, apply! signature, on_device host-staging semantics.
• Default returns nothing for every current objective type (T1
  covers KnotPointObjective, QuadraticRegularizer,
  MinimumTimeObjective, GlobalObjective, CompositeObjective,
  NullObjective).
• Minor release bump (v0.9.6 → v0.9.7 per semver, additive) —
  not done in this turn; flag for the commit step.

---

Out of scope (carried forward)

The remaining items from
./dto-matrix-free.md and
./piccolissimo-matrix-free.md remain
unstarted:

1. Piccolissimo Issue #179 — consume knot_hvp(obj, traj) in the
   Altissimo backend; sunset PR #178's per-iterate probe+SVD for
   Tier-1 objectives. Blocked on this PR + compat bump.
2. DTO StepHVP capability — sibling interface for the
   integrator-side constraint-curvature HVP. Independent of Add a KnotHVP capability interface: objectives can declare a matrix-free per-knot HVP #113;
   release-decoupled.
3. Altissimo backend wiring —
   AltissimoOptions.constraint_curvature_hvp flag + closure
   dispatched on step_hvp_capability.
4. Phase 3 — GPU device-path constraint HVP — the big lift,
   in-scope per prior decision.

Full DTO no-regression sweep also deferred per prior
compute-conservation guidance.

[codecov]

Codecov Report

❌ Patch coverage is 90.69767% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/constraints/linear/global_linear_constraint.jl	85.00%	3 Missing ⚠️
src/solvers/constrain.jl	95.65%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'DirectTrajOpt.jl benchmarks'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 481466f	Previous: 0c47f87	Ratio
scaling_N25_d16_ipopt [alloc]	5280639576 bytes	2155582496 bytes	2.45
scaling_N51_d16_ipopt [wall]	43.62511049 s	22.619076088 s	1.93
scaling_N51_d16_ipopt [alloc]	90177961400 bytes	43996002616 bytes	2.05
scaling_N51_d4_ipopt [wall]	0.728783311 s	0.597400298 s	1.22
scaling_N51_d4_ipopt [alloc]	1524505744 bytes	1253140304 bytes	1.22

This comment was automatically generated by workflow using github-action-benchmark.