test(self_test): add unit tests for check_sqlite and check_workspace

[yuxuan-7814]

Summary

Add test coverage for sqlite and workspace diagnostic functions in the self-test command.

• check_sqlite_with_nonexistent_workspace_dir: verifies sqlite behavior with non-existent workspace directory
• check_workspace_with_valid_directory: verifies workspace check with valid directory
• check_workspace_with_file_instead_of_directory: verifies failure when path is a file instead of directory

What changed and why: Fixed the original check_sqlite_with_nonexistent_db test which had:

1. Signature mismatch (passed db_path but function expects workspace_dir)
2. Tautological assertion that could never fail

Scope boundary: Only adds new unit tests, no production code changes.

Blast radius: None - tests are isolated.

Linked issue(s): N/A

Labels: type: test, risk: low, size: XS

Validation Evidence

All three tests pass under cargo test:

cargo test -p zerocode check_sqlite_with_nonexistent_workspace_dir
cargo test -p zerocode check_workspace_with_valid_directory
cargo test -p zerocode check_workspace_with_file_instead_of_directory

• Commands run and tail output: All tests pass
• Beyond CI: Verified test names match behavior and assertions are meaningful

Security & Privacy Impact

• New permissions/file access? No
• External network calls? No
• Secrets changed? No
• PII in tests? No

Compatibility

N/A - test-only change.

Rollback

Low-risk: git revert <sha>.

[WareWolf-MoonWall]

Review — #8210: test(self_test): add unit tests for check_sqlite and check_workspace

Verdict: CHANGES_REQUESTED
Head SHA: 2eedbd3
Author: yuxuan-7814
Milestone: v0.8.2 (assigned — tracking issue #8181)
Existing reviews from others: None

---

Two of the three new tests (check_workspace_*) are solid. The sqlite test has a signature mismatch that makes it misleading and its assertion is too loose to catch regressions. PR template is incomplete, and Quality Gate CI is missing.

---

🔴 Blocking — PR template incomplete

Required sections from .github/pull_request_template.md are absent:

• Validation Evidence — show cargo test output (with the three new test names passing) or the CI run SHA.
• Security & Privacy Impact — state "No" to all items explicitly.
• Compatibility — N/A; state it.
• Rollback — low-risk: "git revert <sha> is the plan".

---

🔴 Blocking — check_sqlite_with_nonexistent_db has a signature mismatch and a tautological assertion

Signature mismatch. The production function signature is:

fn check_sqlite(workspace_dir: &Path) -> CheckResult {
    let db_path = workspace_dir.join("memory.db");
    match rusqlite::Connection::open(&db_path) { ... }
}

check_sqlite takes a workspace directory path and appends "memory.db" internally. The test does:

let db_path = temp_dir.path().join("nonexistent.db");
let result = check_sqlite(&db_path);

The variable is named db_path and paths to a file-like name (nonexistent.db), but it is passed as the workspace_dir argument. The function will internally derive {temp_dir}/nonexistent.db/memory.db — not a nonexistent database file, but an attempt to open a SQLite file inside a directory that itself doesn't exist. The test name check_sqlite_with_nonexistent_db and the variable name db_path are both wrong; what's actually being tested is behavior with a nonexistent workspace directory.

Tautological assertion. The assertion:

assert!(
    result.passed || result.detail.contains("cannot open"),
    "sqlite check should either pass or report open failure"
);

Production only produces two outcomes: passed = true (SQLite opens) or passed = false with a detail message starting with "cannot open". The disjunction covers both — this assertion can never fail. It provides no regression protection.

Suggested fix:

#[test]
fn check_sqlite_with_nonexistent_workspace_dir() {
    let temp_dir = tempfile::tempdir().unwrap();
    // Pass a nonexistent subdirectory as workspace_dir; the function appends "memory.db"
    let workspace_dir = temp_dir.path().join("nonexistent_workspace");
    let result = check_sqlite(&workspace_dir);
    // The parent dir does not exist, so SQLite cannot create the file
    assert!(!result.passed, "nonexistent workspace directory should fail; got: {}", result.detail);
    assert!(
        result.detail.contains("cannot open"),
        "detail should mention open failure, got: {}",
        result.detail
    );
}

This renames the variable and test correctly, and splits the assertion into two non-tautological checks.

---

🟡 Warning — Quality Gate not present in CI status rollup

Only Apply path labels / PR Path Labeler appears. The full Quality Gate (including Test) did not run. Please re-push to trigger it — in particular, check_workspace_with_valid_directory and check_workspace_with_file_instead_of_directory are async tests that should be confirmed passing under cargo test.

---

🟡 Warning — Branch name prefix mismatch

Branch is fix/self-test-add-check-sqlite-tests; PR title is test:. Use test/ prefix.

---

🟢 What looks good — the two workspace tests are clean and correct

check_workspace_with_valid_directory — creates a real temp dir, calls check_workspace, asserts passed and detail.contains("writable"). Cross-check: production emits format!("{} (writable)", workspace_dir.display()) on success. Correct. ✅

check_workspace_with_file_instead_of_directory — writes a file, passes its path to check_workspace, asserts !passed and detail.contains("is not a directory"). Cross-check: production emits format!("{} exists but is not a directory", workspace_dir.display()) for the Ok(_) non-dir branch. Correct. ✅

unwrap() usage — test code only, within #[cfg(test)]. Acceptable. ✅
No PII or real identities in fixtures. ✅
tempfile::tempdir() isolation — clean, no global state. ✅

[Audacity88]

@yuxuan-7814 I re-reviewed current head 5897106a5e4c79c72d1804f6e4920b898e4c0c01 after @WareWolf-MoonWall's review. I checked the live PR state, the prior review, the current one-file patch, the PR body, the existing check_workspace / check_sqlite helper contracts, and the visible GitHub checks. I did not run local Cargo validation.

✅ Resolved — sqlite coverage now matches the helper contract

The new check_sqlite_with_nonexistent_workspace_dir test now passes a value named and shaped as a workspace directory path. That matches the production helper, which appends memory.db internally. The assertions also now check the two useful facts separately: the check must fail, and the detail must mention the open failure. That gives the test real regression signal instead of accepting either production outcome.

✅ Resolved — PR body now carries the missing review context

The PR body now includes the validation commands, scope, blast radius, security/privacy impact, compatibility, and rollback notes for this test-only change. That addresses the missing-template-context part of the earlier review.

🟢 What looks good — workspace tests cover both expected branches

The two check_workspace tests line up with the current helper behavior: a real temp directory must pass and mention writable, while a file path must fail and mention is not a directory. Those are small tests, but they are aimed at the actual observable self-test details rather than only exercising implementation scaffolding.

🟡 Warning — branch refresh and CI still need to happen

GitHub currently reports this branch as conflicting with master, and the only visible check on the current head is Apply path labels; the full Quality Gate / Test jobs have not run. I am not adding an approval while @WareWolf-MoonWall's CHANGES_REQUESTED review is still active, but from this pass I do not see a new source-level blocker beyond the existing branch refresh, CI, and review-state follow-up.

[singlerider]

Moved from the released v0.8.2 to v0.8.3. Test-only coverage; rides with the next point release as carried-over hardening.

[singlerider]

🔴 Same issue as #8208: this is a production behavior change to channel_history_content_for_user_turn, not a test, and it duplicates #8167 (which owns this exact image-marker history fix for the now-CLOSED #8151). This hunk was added after @Audacity88's review of the one-file test patch, so it is unreviewed scope creep. Drop it; let #8167 be the single home for the fix.

[singlerider]

@yuxuan-7814 the test work here is fine: @WareWolf-MoonWall's sqlite signature/tautology concern and @Audacity88's template-context concern were both addressed in the test-only patch they reviewed (5897106a). The problem is what landed on top of that.

🔴 Blocking - unreviewed scope creep duplicating #8167 / #8151

Current head 32789f92 adds the same production change to channel_history_content_for_user_turn (orchestrator) and the same models: None openrouter churn that #8208 carries. This was added after @Audacity88's review, so their "no new source-level blocker" pass does not cover it. The orchestrator change duplicates #8167, which already owns this image-marker history fix (closing #8151, now CLOSED). See the inline note. Drop the orchestrator and openrouter hunks; scope this back to the src/commands/self_test.rs tests only.

🔴 Blocking - merge conflicts

GitHub reports the branch conflicting with master. It must merge cleanly before landing.

Path forward

Reset this PR to just the self_test additions @Audacity88 already validated, rebase to clear the conflict, and let #8167 carry the channel history fix. @WareWolf-MoonWall's CHANGES_REQUESTED is still active, so the test concerns should be re-confirmed once the scope is trimmed.