Skip to content

refactor(test): the GPU smoke test written once, status.jl split, coverage gaps (T4 of #449) - #477

Merged
ocots merged 1 commit into
mainfrom
refactor/t4-leftovers
Nov 20, 2026
Merged

ocots merged 1 commit into
mainfrom
refactor/t4-leftovers

Conversation

@ocots

@ocots ocots commented Nov 20, 2026

Copy link
Copy Markdown
Member

Refs #449 (T4, the two items left open). Conventions: ct-conventions v0.14.0, ct-agents v0.5.0.

Summary

  • GPU smoke (test/common/ without duplicates, Share the GPU smoke test between OptimalControl and OptimalControlScalarized #445): test_smoke_optimalcontrol.jl and test_smoke_scalarized.jl were the same file but for the backend. Now common/gpu_tests.jl holds the test once, and the two files are wrappers. The GPU contract (TST6: nobody asks CUDA but common/capabilities.jl) scans the shared file too. The lines of the report become test_smoke_gpu … for both modelings (the modeling column tells them apart; nothing reads the former names).
  • common/status.jl (615 lines): what is recorded during a run stays (draft test sets, rows, results, trajectories, the selection of the problems); the files written at the end of a run (reports, results, environments, trajectories) move to common/status_reports.jl, included in the same module TestStatus: no name changes.
  • Coverage (Codecov on main: 92.3 %, 246 lines missed): 220 of them are the ModelingToolkit extension, which the coverage workflow does not run (opt-in environment; to add it, the coverage workflow of CTActions needs a test_args input, a separate change). Of the others, now tested: the show methods of NumericalReference and AnalyticReference (11 lines), the discovery of the problems, run at precompilation (5 lines, now checked against the files). Left: two edge cases of the JuMP solution builder (a grid of one step, a solver without barrier iterations) and two one-line methods the report counts as missed after inlining.

Test plan

  • core and JuMP: passed (core 2347).
  • The GPU files parse; not run locally (no CUDA on macOS): the label run ci kkt-runner runs them on the GPU runner.
  • CI, CPU and GPU.

🤖 Generated with Claude Code

…erage gaps tested (T4 of #449)

The two GPU smoke files were the same but for the backend: common/gpu_tests.jl, two wrappers;
the GPU contract scans it too (TST6). status.jl keeps what is recorded during a run, the
files written at its end move to status_reports.jl (same module). The show methods of the
references and the discovery of the problems, uncovered on Codecov, are tested.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@ocots ocots added run ci github-runner Trigger the CI workflow on GitHub-hosted runners run ci kkt-runner Trigger the CI workflow on the self-hosted kkt runner labels Nov 20, 2026
@ocots
ocots merged commit a4071cf into main Nov 20, 2026
31 of 33 checks passed
@ocots
ocots deleted the refactor/t4-leftovers branch November 20, 2026 18:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

run ci github-runner Trigger the CI workflow on GitHub-hosted runners run ci kkt-runner Trigger the CI workflow on the self-hosted kkt runner

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant