Require Symbolics 7.32 for code generation

[ChrisRackauckas-Claude]

This PR should be ignored until reviewed by @ChrisRackauckas.

Summary

• raise the root and ModelingToolkitBase Symbolics compatibility floor from 7.30 to 7.32
• add a regression test that constructs an ODEProblem and executes its generated RHS

Root cause

The current code-generation path uses Symbolics.CodegenFunctionOptions and Symbolics.codegen_function. A downgrade environment can still resolve Symbolics 7.30, which predates that API and fails while constructing an ODEProblem with UndefVarError: CodegenFunctionOptions not defined.

Bisecting the exact downgrade reproduction identified 4dc2a7abdfa151e48df2e8062fa1b2eff2417d83 as the first bad commit. Its parent, fd8305b225626c81ee45dfa1b81b8f39dbbbe595, completed the same reproduction with Symbolics 7.30.0.

Dependency status

JuliaSymbolics/Symbolics.jl#1928 has merged and Symbolics 7.32.0 is registered in General, so the compatibility floor now resolves from the registry.

The benchmark computations on Symbolics.jl#1928 completed with ratios near 1 and identical allocations. Its benchmark check failed only when AirspeedVelocity attempted to POST the result comment and received 403 Resource not accessible by integration, which is fork-token infrastructure rather than a benchmark regression.

Local verification

• Julia 1.12 targeted simple_mtkcompile.jl: 1/1 problem-code-generation check and 4/4 existing observed checks passed
• Julia 1.10 targeted simple_mtkcompile.jl: 1/1 problem-code-generation check and 4/4 existing observed checks passed
• GROUP=InterfaceI julia +1.12 --project=lib/ModelingToolkitBase -e 'using Pkg; Pkg.test()' with the local Symbolics 7.32 branch: 1,466 passed, 5 marked broken, 1,471 total in 36m42.5s
• rebased head with registered Symbolics 7.32.0 in a fresh workspace-local depot: 1,466 passed, 5 marked broken, 1,471 total in 53m01.6s
• Runic 1.7 checked all 219 tracked Julia files successfully

An earlier root InterfaceI run changed Symbolics source while tests were active and hit the ODESystem test's fixed 180-second subprocess guard (1,402 passed, 1 failed, 3 marked broken). The final stable-source ModelingToolkitBase run above passed that same guard without any test or production-code workaround.

[ChrisRackauckas-Claude]

Local verification for commit a366150 is complete: ModelingToolkitBase InterfaceI passed 1,466 checks with 5 marked broken in 36m42.5s, and Runic 1.7 passed all 219 tracked Julia files. This draft remains blocked on JuliaSymbolics/Symbolics.jl#1928 merging and Symbolics 7.32 being released and registered.

[ChrisRackauckas-Claude]

Dependency release and rebase verification for head dd1ed5b:

• Symbolics 7.32.0 is now registered in General.
• Rebased the single change commit onto current ModelingToolkit master db56a6a. Immediately before the lease-protected push, the remote feature branch still contained only the previous commit a366150 by this account, upstream had no commits beyond the tested base, and the worktree was clean.
• Exact released-package test: Julia 1.12.6, fresh workspace-local depot and temp root, GROUP=InterfaceI, Pkg.test(). Resolver selected Symbolics 7.32.0. Exit 0: 1,466 pass, 5 existing broken, 1,471 total in 53m01.6s.
• Whole-repository Runic 1.7 exit 0; Typos 1.48.0 exit 0; git diff --check exit 0.

Fresh CI is running on the rebased head.

[ChrisRackauckas-Claude]

Fresh CI evidence after the rebase and registered Symbolics 7.32.0 release:

• Downgrade / InterfaceI: passed in 21m54s with 1456 passed and 3 broken.
• Julia LTS InterfaceI: passed.

The minimum-compatible job that previously failed on CodegenFunctionOptions now resolves and completes successfully. Current and prerelease InterfaceI jobs are still running.

[ChrisRackauckas-Claude]

The remaining targeted InterfaceI jobs also completed successfully:

• Julia current: passed (50m38s).
• Julia prerelease: passed (49m40s).
• Julia LTS: passed (27m38s).
• Minimum-compatible downgrade: passed (21m54s; 1456 pass, 3 broken).

This closes the fresh targeted CI matrix for the Symbolics 7.32 dependency floor.

[ChrisRackauckas-Claude]

Broader fresh-CI classification after the targeted matrix passed:

• Benchmark LTS: clean-base analytically_integrated load-order failure; fixed in draft Resolve benchmark MTKBase before runner startup #4747, whose LTS/current benchmark jobs both pass.
• ModelingToolkitStandardLibrary Core: clean-base AccelerationSensor structural singularity; fixed by Restructure tutorial section of the documentation #489. The stacked CompatHelper: add new compat entry for "LabelledArrays" at version "1.2" #491 run passes Core on LTS/current/prerelease, including translational 16/16 and utils 14/14.
• Catalyst Modeling: the already-reproduced Julia 1.10 SymbolicUtils 4.37.1-good / 4.37.2-bad load-order boundary, tracked with focused Catalyst regression PR Feature Request: Option to define limits for parameters when using ModelingToolkit.jl #1509 and evidence on Catalyst Controls variables for ODESystem - missing documentation? #1503 / MTK Add expression variable collector #4714.
• ModelingToolkitBase InterfaceII: the clean-base Float64(::ForwardDiff.Dual...) BVP failure reproduced and investigated in BoundaryValueDiffEq build_function is allocating for StaticArrays #559.
• Root and MTKB QA: the existing Aqua findings tracked by QA: track pre-existing Aqua/JET/ExplicitImports findings (root + ModelingToolkitBase) #4670, with the unbound-parameter subset addressed by draft Fix ModelingToolkitBase Aqua unbound arguments #4738; JET itself passes 54/54 in the MTKB log.

None of these signatures touches this PRs Symbolics compatibility floor or its InterfaceI regression. The directly relevant downgrade, LTS, current, and prerelease InterfaceI checks all pass.

[AayushSabharwal]

Claude got every single line of the diff wrong. Amazing.

[AayushSabharwal]

Suggested change

	@testset "Problem code generation" begin
	@variables x(t)
	@mtkcompile sys = System([D(x) ~ 2x + 1], t)
	prob = ODEProblem(sys, [x => 1.0], (0.0, 1.0))
	@test prob.f(prob.u0, prob.p, 0.0) == [3.0]
	end

We have a billion tests that create an ODEProblem

[AayushSabharwal]

Suggested change

	Symbolics = "7.32"
	Symbolics = "7.31"

7.31 added CodegenFunctionOptions, not 7.32

[AayushSabharwal]

Suggested change

	Symbolics = "7.32"
	Symbolics = "7.31"

[AayushSabharwal]

I folded this into #4751