Skip to content

Require strict SciMLTesting 2.4 QA - #146

Merged
ChrisRackauckas merged 1 commit into
SciML:mainfrom
ChrisRackauckas-Claude:codex/strict-scimltesting-24-basemodelica
Jul 27, 2026
Merged

ChrisRackauckas merged 1 commit into
SciML:mainfrom
ChrisRackauckas-Claude:codex/strict-scimltesting-24-basemodelica

Conversation

@ChrisRackauckas-Claude

Copy link
Copy Markdown
Member

Ignore until reviewed by @ChrisRackauckas.

Removes all BaseModelica-specific Aqua and ExplicitImports QA waivers and runs plain run_qa(BaseModelica) under SciMLTesting 2.4. Direct imports now use the declaring public APIs rather than ModelingToolkit/Symbolics reexports or PythonCall builtins.

Also enables doctests and export coverage in Documenter, documents the public API at its definition sites, and ensures it is rendered on the API page.

Local verification:

  • Julia 1.12 Pkg.test() passed: 75 Julia-parser, 138 ANTLR-parser, and 25 error-message tests.
  • Plain strict SciMLTesting 2.4 QA passed: 20/20.
  • Documenter doctests, checkdocs = :exports, and HTML rendering passed.
  • Runic check and git diff --check passed.

No [sources] entries or path dependencies were added.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
@ChrisRackauckas
ChrisRackauckas marked this pull request as ready for review July 27, 2026 04:44
@ChrisRackauckas
ChrisRackauckas merged commit 1c5ff39 into SciML:main Jul 27, 2026
8 of 10 checks passed
ChrisRackauckas added a commit that referenced this pull request Aug 5, 2026
#146 added SciMLBase, SymbolicIndexingInterface, SymbolicUtils and
Symbolics as direct dependencies. Downgrade CI only pins direct
dependencies, so those four went from "resolved to newest compatible"
to "pinned at their declared floor", which selected Symbolics 7.18.1.

At that version the diode model in the ANTLR suite produces a system
whose W matrix is singular, so every Rosenbrock step falls through to
LinearSolve's QR fallback and the integrator never terminates. The run
is not slow, it does not converge: one hour of wall time produced 11.3
billion allocations while still inside solve!.

Raising only the Symbolics floor fixes it. The solver stack is not
involved: GROUP=Core passes with SciMLBase still at 3.1.0,
OrdinaryDiffEqCore at 4.3.0 and LinearSolve at 3.87.0.

Co-authored-by: ChrisRackauckas-Claude <accounts@chrisrackauckas.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants