Docs testing part 1

[wsmoses]

No description provided.

[oscardssmith]

@ChrisRackauckas are doctestsnotworking for this repo?

[wsmoses]

@ChrisRackauckas-Claude identify any/all failing doctests and open issues for them as isolated. Also make a patch to this pr reverting any doctest which currently fails [so we can merge all that pass now, and do the rest as follow ups]

[ChrisRackauckas-Claude]

I isolated the failing docs example blocks from this PR and opened follow-up issues:

• second_order_adjoints docs example fails under Enzyme conversion #1532: docs/src/examples/ode/second_order_adjoints.md fails after the Enzyme conversion with EnzymeRuntimeActivityError: Detected potential need for runtime activity.
• second_order_neural docs example segfaults during Documenter example execution #1533: docs/src/examples/ode/second_order_neural.md exits 139 during Documenter example execution. I reproduced exit 139 both before and after reverting this page to Mooncake, so this likely also needs baseline investigation.
• SimpleChains docs example fails with Enzyme illegal type analysis #1534: docs/src/examples/neural_ode/simplechains.md fails with IllegalTypeAnalysisException: Enzyme compilation failed due to illegal type analysis; this file was not changed by this PR.

I made the requested local patch as commit 389dd081:

Revert failing second-order docs conversions

It reverts only these PR changes back to Mooncake:

• docs/src/examples/ode/second_order_adjoints.md
• docs/src/examples/ode/second_order_neural.md

Local verification run on Julia 1.12.6:

timeout 3600 env JULIA_DEPOT_PATH="$PWD/.julia_depot" /home/crackauc/.juliaup/bin/julia +1.12.6 --project -e 'using Pkg; Pkg.activate("docs"); push!(LOAD_PATH, pwd()); using Documenter, SciMLSensitivity; ENV["GKSwstype"]="100"; using Plots; makedocs(; root="docs", sitename="target", modules=[SciMLSensitivity], clean=true, doctest=false, linkcheck=false, warnonly=[:missing_docs], pages=["target" => "examples/ode/second_order_adjoints.md"])'

Result: exited 139 after reverting, so I am not claiming the page passes. The revert does remove the PR-introduced Enzyme runtime-activity error, but the Mooncake/original page still appears to have a separate crash.

I could not push the patch to this branch from this environment:

remote: Permission to SciML/SciMLSensitivity.jl.git denied to ChrisRackauckas-Claude.
fatal: unable to access 'https://github.com/SciML/SciMLSensitivity.jl.git/': The requested URL returned error: 403

SSH push also failed with Permission denied (publickey). A maintainer with branch write access can cherry-pick local commit 389dd081 or apply this equivalent diff:

diff --git a/docs/src/examples/ode/second_order_adjoints.md b/docs/src/examples/ode/second_order_adjoints.md
index 670dd82d..00e4c011 100644
--- a/docs/src/examples/ode/second_order_adjoints.md
+++ b/docs/src/examples/ode/second_order_adjoints.md
@@ -24,7 +24,7 @@ import OrdinaryDiffEq as ODE
 import Plots
 import Random
 import OptimizationOptimJL as OOJ
-import Enzyme
+import Mooncake
@@ -85,7 +85,7 @@ callback = function (state, l; doplot = false)
     return l < 0.01
 end
 
-adtype1 = OPT.AutoEnzyme()
+adtype1 = OPT.AutoMooncake(; config = Mooncake.Config(; friendly_tangents = true))
 optf1 = OPT.OptimizationFunction((x, p) -> loss_neuralode(x), adtype1)
diff --git a/docs/src/examples/ode/second_order_neural.md b/docs/src/examples/ode/second_order_neural.md
index e578cf67..5d12266d 100644
--- a/docs/src/examples/ode/second_order_neural.md
+++ b/docs/src/examples/ode/second_order_neural.md
@@ -29,7 +29,7 @@ import OptimizationOptimisers as OPO
 import RecursiveArrayTools
 import Random
 import ComponentArrays as CA
-import Enzyme
+import Mooncake
@@ -62,7 +62,7 @@ callback = function (state, l)
     l < 0.01
 end
 
-adtype = OPT.AutoEnzyme(; mode = Enzyme.set_runtime_activity(Enzyme.Reverse))
+adtype = OPT.AutoMooncake(; config = Mooncake.Config(; friendly_tangents = true))
 optf = OPT.OptimizationFunction((x, p) -> loss_n_ode(x), adtype)

[ChrisRackauckas-Claude]

Follow-up to my previous comment: I pushed the patch to the fork and opened it as draft PR #1535 against this branch: #1535

[ChrisRackauckas]

The segfault is EnzymeAD/Enzyme.jl#3292

[wsmoses]

@ChrisRackauckas which one is that so we can remove it and check the others work end to end