Skip to content

Fix failing test suite: restore sorry placeholders in example problems#28

Merged
Cuuper22 merged 1 commit into
mainfrom
claude/tender-lamport-s4wejv
Jun 10, 2026
Merged

Fix failing test suite: restore sorry placeholders in example problems#28
Cuuper22 merged 1 commit into
mainfrom
claude/tender-lamport-s4wejv

Conversation

@Cuuper22

Copy link
Copy Markdown
Owner

What

Restores the unsolved (sorry-placeholder) versions of the four example problem files in examples/. The test suite on main currently fails with 14 failed / 300 passed; with this change it's 314 passed, 9 skipped.

Why

PR #27 added tests/test_operational.py, which asserts that example problem files contain sorry placeholders (test_original_content_is_preloaded, test_intermediate_has_sorry_placeholders). But the same PR committed the example files with completed proofs — solver output evidently leaked into the working tree before commit. The tests it added have been failing ever since.

Solved example files also defeat the project's premise: problems must ship unsolved for the Prover/Critic loop (and the README's mock-mode demo) to have anything to prove.

Changes

Verification

  • python -m pytest tests/ -q → 314 passed, 9 skipped (the 9 skips are environment-gated and unchanged from before).
  • Confirmed the solver writes solutions to a separate work dir (solution_<id>.json + ZIP), never back into examples/, so the current code won't re-introduce this.

Notes for the maintainer

  1. The other 12 failures I saw locally (TestGeminiProvider, provider factory) were a broken cffi wheel in my container, not a code issue — google-generativeai imports fine after reinstalling cffi, and CI installs it cleanly via pyproject deps.
  2. GitHub Actions has zero workflow runs for this repo, despite .github/workflows/ci.yml being configured for pushes/PRs to main and the README sporting a CI badge. That's why this breakage went unnoticed. Worth checking Settings → Actions to confirm Actions is enabled — once it is, this PR should get a CI run.

https://claude.ai/code/session_01KAKnLjpFZ9oF74GNW9QNtN


Generated by Claude Code

PR #27 added tests/test_operational.py, which requires the example
problem files to contain unsolved 'sorry' placeholders, but the same
PR committed those files with completed proofs (solver output leaked
into the working tree). The suite has failed on main ever since:

- test_original_content_is_preloaded: expects 'sorry' in every
  manifest problem
- test_intermediate_has_sorry_placeholders: expects >=5 sorries in
  examples/intermediate.lean

Restore test_theorem.lean, basic_algebra.lean, and number_theory.lean
to their pre-#27 unsolved versions, and replace the proof bodies in
intermediate.lean with sorry while keeping its theorem statements.
Solved files also defeat the project's purpose: problems must ship
unsolved for the Prover/Critic loop to have anything to prove.

Result: 314 passed, 9 skipped (was 14 failed, 300 passed, 9 skipped).

https://claude.ai/code/session_01KAKnLjpFZ9oF74GNW9QNtN
@Cuuper22 Cuuper22 merged commit d94af0f into main Jun 10, 2026
0 of 5 checks passed
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