[ocots]

Closes #966.

Summary

using OptimalControl no longer brings the module name CTLie into scope. CTLie 0.3.0 (released 2026-10-10, in General) made the @Lie expansion hygienic (control-toolbox/CTLie.jl#40): the generated code holds the modules, not their names. The export was there only for that.

The issue's checklist

1. Compat: CTLie = "0.3" in Project.toml, docs/Project.toml and its snapshot docs/src/assets/Project.toml (copied by docs/make.jl).
2. Export removed: @reexport import CTLie: CTLie and its comment are gone from src/imports/ctlie.jl, replaced by a one-line note. OptimalControl still binds CTLie internally (using CTLie: CTLie in src/OptimalControl.jl), so OptimalControl.CTLie works and test_ctflows.jl's OptimalControl.CTLie is unchanged.
3. Re-export test test/suite/reexport/test_ctlie.jl: the testset "Generated Code Prefix" becomes "Module name not exported" (!is_exported(OptimalControl, :CTLie)).
4. Hygiene test: two top-level modules in the test file, LieImportOnly (import OptimalControl only) and LieMacroOnly (using OptimalControl: @Lie only). They call @Lie [X, Y], @Lie [X, Y] is_autonomous = false and @Lie {H1, H2}. The test asserts that neither module has CTLie or CTBase in scope, and that the results are correct.
5. Docs: the section "using OptimalControl is needed" of the page "The @Lie macro" is dropped, with its hidden check (which asserted the old failure and would now fail).

Choices made

• Compat "0.3" only, not "0.2, 0.3". Without the hygienic expansion, removing the export breaks @Lie for users, so 0.2 must be excluded. Nothing else in the dependency graph depends on CTLie (checked in General: CTFlows, CTDirect, CTSolvers, CTModels, CTParser and CTBase have no CTLie dep), so the bump resolves. This supersedes CompatHelper's CompatHelper: bump compat for CTLie to 0.3, (keep existing compat) #973 and CompatHelper: bump compat for CTLie to 0.3 for package docs, (keep existing compat) #974, which can be closed after merge.
• Also removed import CTLie: CTLie / import CTBase: CTBase in test/suite/indirect/test_shooting_sweep.jl. They were there only because @Lie needed them (comment said so). The Goddard @Lie now runs without them, which is an extra real-world hygiene check.
• Documented as breaking, under a new [Unreleased] / "Breaking changes (unreleased)" section of CHANGELOG.md and BREAKING.md, with a migration snippet (using CTLie: CTLie or OptimalControl.CTLie.…). I did not bump version in Project.toml: the release number (2.3.0-beta or other) is the release PR's call.
• CTLie 0.3's own breaking change is relayed in BREAKING.md. Guard errors of ad/Lift/∂ₜ on HVF / in-place fields are now IncorrectArgument instead of NotImplemented, and OptimalControl users see them through the re-exported functions. Nothing in OptimalControl's tests or docs relied on NotImplemented there.
• docs/src/assets/Manifest.toml not regenerated: docs/make.jl overwrites it on the next docs build. Only the Project snapshot was updated by hand, to stay consistent with docs/Project.toml.
• CTBase stays exported (out of scope, as the issue says); the other module re-exports are Remove the re-exports of CTModels and ExaModels once @def is hygienic #967, blocked on CTParser#356.

Checks

Ran locally (resolved env: CTLie v0.3.0): suite/reexport/*, suite/environment/*, suite/extensions/* and suite/indirect/* give 861/861 passing, including the Goddard shooting sweep that uses @Lie. Full suite and docs build not run locally.

🤖 Generated with Claude Code