ci: fix conda_avail circular gate silently skipping conda-full self-tests

[mixmansoundude]

Summary

Follow-up to #390 (merged), found via a direct owner question ("are we sure we aren't shorting conda-full?") that prompted checking real step-level CI conclusions instead of trusting the job's overall green status.

• conda_avail's check (introduced in ci: lane-aware if:always() rollout for remaining 36 gating-lane steps (item 7) #390) assumed the "Bootstrap environment (run_setup.bat)" step performs the first real Miniconda install for the conda-full lane, so gating ~27 downstream real/conda-full-restricted self-test steps on "is conda already on disk" would only skip in the rare case that install genuinely failed.
• That assumption is false. "Bootstrap environment" runs against this repo's own root (no loose .py files there — it exercises the empty-repo/no_python_files graceful-exit path, not a real target app), so it never installs Miniconda at all. Every downstream selfapps step capable of performing the first real install was ALSO gated behind this same check — a circular self-skip with no way out.
• Confirmed directly via the GitHub Actions API against two real runs on ci: lane-aware if:always() rollout for remaining 36 gating-lane steps (item 7) #390 (efd7a5c, fd7a046): all ~27 real/conda-full-only steps showed "conclusion":"skipped" at the identical timestamp on conda-full, every run. The job still reported overall SUCCESS (a skipped step doesn't fail a job), so this merged without the coverage loss ever surfacing.
• Fix: reverted all 36 conda_avail-dependent if: clauses to their pre-item-7 unconditional form (matrix.mode == 'conda-full', no conda_avail dependency), keeping !cancelled() (the part of item 7 that is correct and validated). The conda_avail step itself is left in place, unused, with a comment explaining why.
• Wholistic audit performed on request, not just this one bug: swept every if: condition in the whole 4444-line workflow (all 5 jobs) for the same bug class — none found elsewhere; traced the downstream verdict/enforce/aggregate chain to confirm it has no independent bug (it's just architecturally blind to a step that never ran). Compared the current yml against two real historical baselines (e20d7cd, 2026-07-08, well before this work — conda-full ran 78 min; fd11ff3, immediately before item 7 started — conda-full ran 94 min) via direct condition-by-condition diff: zero steps removed, 25 new steps added (all documented features shipped in that window), and all 73 changed conditions fall into exactly two known, reviewed categories (always()→!cancelled(), or gaining a !cancelled() && prefix) — none reference conda_avail anymore. Full rationale recorded in CLAUDE.md's new Active Backlog item 15.

Test plan

• python -m compileall -q .
• python tools/check_delimiters.py run_setup.bat
• python -m yamllint .github/workflows/
• actionlint -oneline .github/workflows/*.yml
• python -m pytest tests/test_*.py -q (437 passed, 2 skipped)
• Confirmed via GitHub Actions API that the pre-fix state (two runs on ci: lane-aware if:always() rollout for remaining 36 gating-lane steps (item 7) #390) shows the bug; confirmed via direct yml diff that the fix restores exact parity with two known-healthy historical baselines
• Real Windows CI (this PR) — confirm the conda-full lane's ~27 previously-skipped self-tests actually run and pass again, and that job duration returns to the ~80-95 min historical range

Claude-Session: https://claude.ai/code/session_015xbWLPbiaKVsobB9FZy8kS

---

Generated by Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: d88d1cf0-97ba-434a-98e8-a040ec01f114

📥 Commits

Reviewing files that changed from the base of the PR and between 283af5a and 9d360f0.

📒 Files selected for processing (2)

• .github/workflows/batch-check.yml
• CLAUDE.md

📜 Recent review details

⏰ Context from checks skipped due to timeout. (10)

• GitHub Check: analyze
• GitHub Check: auto_merge
• GitHub Check: Batch syntax/run check (conda-full)
• GitHub Check: Batch syntax/run check (real)
• GitHub Check: Batch syntax/run check (cache)
• GitHub Check: Batch syntax/run check (contract-uv)
• GitHub Check: Batch syntax/run check (contract-uv-fail)
• GitHub Check: Batch syntax/run check (uv)
• GitHub Check: Batch syntax/run check (uv-dl-fallback)
• GitHub Check: Batch syntax/run check (justme-test)

🧰 Additional context used

📓 Path-based instructions (4)

**/*

📄 CodeRabbit inference engine (AGENTS.md)

**/*: Use GitHub Actions results, especially Windows runner behavior, as the authoritative source of truth; treat local or sandbox runs as advisory.
Do not weaken tests, remove logging or artifacts, or delete or skip checks to obtain a green build.
Any new observable log line, disk file, or assertion-detectable behavior must produce an NDJSON row and have its artifact path added to the batch-check.yml test-logs upload using both existing slash-style variants.
Follow an explicit Plan -> Check the plan -> Execute -> Self-check/tests loop, and make one change set per loop.
Push every commit before it can be lost; do not leave commits unpushed when a turn ends.
Do not implement scope changes without first creating a backlog item unless explicitly instructed; fix CI failures before proceeding, and add a derived-requirement comment for untraceable code.

**/*: Update the relevant knowledge document in the same commit when a change teaches or invalidates a lesson; edit existing entries rather than only appending.
Implement exactly one missing feature slice and add exactly one missing test per iteration loop; freeze scope and defer new requirements to the backlog.

Files:

• CLAUDE.md

**/*.{md,bat,cmd}

📄 CodeRabbit inference engine (AGENTS.md)

Reference run_setup.bat locations in documentation by stable label or subroutine name rather than exact line numbers.

Files:

• CLAUDE.md

.github/workflows/*.yml

📄 CodeRabbit inference engine (AGENTS.md)

.github/workflows/*.yml: Do not change GitHub Actions workflow triggers, permissions, or retention settings.
Isolate slow, flaky, or environment-dependent diagnostics in their own non-gating (continue-on-error) lane instead of slowing deterministic gating lanes.
Keep the CI interface synchronized with bootstrapper messages and status summaries; update workflow parsers whenever emitted text changes, and vice versa.
Use only the single inline model path in batch-check.yml for future model work; do not add older online-model or extra inline experiments.

Files:

• .github/workflows/batch-check.yml

.github/workflows/**/*.{yml,yaml}

📄 CodeRabbit inference engine (CLAUDE.md)

Validate modified GitHub Actions workflows with both yamllint and actionlint.

Files:

• .github/workflows/batch-check.yml

🪛 markdownlint-cli2 (0.23.0)

CLAUDE.md

[warning] 717-717: Ordered list item prefix
Expected: 2; Actual: 15; Style: 1/2/3

(MD029, ol-prefix)

🔇 Additional comments (5)

.github/workflows/batch-check.yml (3)

273-293: 📐 Maintainability & Code Quality

Confirm both required workflow linters passed.

Please run yamllint and actionlint against .github/workflows/batch-check.yml before merging.

As per coding guidelines, modified GitHub Actions workflows must be validated with both yamllint and actionlint.

Source: Coding guidelines

---

305-336: 🩺 Stability & Availability

Verify the restored conda-full execution on Windows.

The supplied context says real Windows validation is still pending. Inspect a windows-latest run and confirm these steps execute in conda-full rather than merely producing an overall-success job, with their expected NDJSON rows and artifacts present.

As per coding guidelines, GitHub Actions results—especially Windows runner behavior—are authoritative; local or sandbox runs are advisory.

Source: Coding guidelines

---

351-351: LGTM!

Also applies to: 378-422, 557-563, 572-572, 683-773, 907-931

CLAUDE.md (2)

717-762: 📐 Maintainability & Code Quality

Update or supersede the original conda-gating backlog item.

This entry says Item 7’s premise is false, but the change adds Item 15. Verify that Item 7 itself is updated or explicitly marked superseded; otherwise future maintainers may follow stale guidance and recreate the circular skip.

As per coding guidelines, when a change invalidates a lesson, update the existing knowledge entry rather than only appending a new one.

Source: Coding guidelines

---

717-717: 📐 Maintainability & Code Quality

Resolve the MD029 ordered-list warning.

markdownlint-cli2 reports Expected: 2; Actual: 15 on Line 717. Align the prefix with the surrounding list style and rerun the linter, while preserving any stable backlog identifier.

Source: Linters/SAST tools

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes

  • Improved automated validation for conda-full scenarios by allowing self-tests to continue when the informational Conda availability check does not report availability.
  • Prevented valid test scenarios from being skipped due to incorrect gating, covering environment setup, dependency handling, executable checks, runtime behavior, and safety validations.

• Documentation

  • Updated project documentation with follow-up notes explaining the validation workflow adjustment and remaining improvements.

Walkthrough

The workflow makes Miniconda availability diagnostic-only and removes its output from conda-full self-test gating. The active backlog documents the correction and deferred rewiring.

Changes

Conda-full workflow gating

Layer / File(s)	Summary
Diagnostic check and backlog record .github/workflows/batch-check.yml, CLAUDE.md	The Miniconda availability check is retained for diagnostics, while the backlog documents its unused output and deferred rewiring.
Restored self-test execution .github/workflows/batch-check.yml	Self-test, dependency-check, PyInstaller, EXE-smokerun, and final validation steps no longer require conda_avail.outputs.available == 'true'; existing cancellation, cache, and lane predicates remain.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• mixmansoundude/Python_vs_Windows#390: Introduced the conda_avail-based conda-full gating adjusted by this PR.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly names the CI gate fix and the affected conda-full self-tests.
Description check	✅ Passed	The description is directly about the workflow gate fix and matches the changeset.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/pipreqs-fallback-maintenance-49jhr9

---

_{Comment @coderabbitai help to get the list of available commands.}