Add file system mounting support (FUSE, Cloud Files, NFS, FileProvider)

.claude/commands/mount-test-loop.md

@@ -0,0 +1,89 @@
+# Mount Backend Fix & Test Loop
+
+Iteratively fix production bugs, improve test infrastructure, and achieve
+a fully green mount test matrix with zero workarounds.
+
+## Context Files
+
+Read these at the start of every iteration:
+
+1. `PRD.md` — architecture, phases, dependencies, active plan
+2. `PROGRESS.md` — current completion status
+3. `docs/MANUAL_TESTING.md` — full test case descriptions
+4. `docs/TESTING.md` — naming conventions
+
+## Phase Order & Dependencies
+
+```
+A1 (FUSE+SSH fix) ──→ B2 (remove skip macro) ──→ C1 (cross-backend)
+A2 (readonly)     ──→ C1
+A3 (TTL CLI)      ──→ C2 (TTL tests)
+A4 (FP template)  ──→ B4 (FP fixture) ──→ C1
+A5 (TODOs)        → D3
+B1 (polling)      → C1
+B3 (test hacks)   → C1
+B5 (Windows script) — independent
+```
+
+## Key Patterns
+
+### Agent Usage
+
+Use custom agents from `.claude/agents/` per CLAUDE.md:
+- **rust-explorer** for investigating bugs (A1, A2)
+- **rust-coder** for production fixes (A1-A4) and test infra (B1-B5)
+- **code-validator** after any production code change (BLOCKING)
+- **test-implementor** for test rewrites (C1-C2)
+- **test-validator** after test changes (BLOCKING)
+
+### MountProcess + Template
+
+All tests use `#[apply(super::plugin_x_mount)]` with `skip_if_no_backend!`.
+See PRD for full pattern.
+
+### Polling Helpers (Phase B1)
+
+Replace `wait_for_sync()` (2s sleep) with:
+```rust
+mount::wait_until_exists(&ctx, &path);    // polls 200ms, 10s timeout
+mount::wait_until_content(&ctx, &path, expected);
+mount::wait_until_gone(&ctx, &path);
+```
+
+## Iteration Protocol
+
+### Step 1: Select Next Item
+Pick the first `[ ]` item in progress.md whose dependencies are met.
+`[-]` items take priority.
+
+### Step 2: Implement
+- Production fixes: use rust-explorer → rust-coder → code-validator pipeline
+- Test changes: use test-implementor → test-validator pipeline
+- `cargo fmt --all` + `cargo clippy --all-features --workspace --all-targets`
+
+### Step 3: Test
+```bash
+cargo nextest run --all-features -p distant -E 'test(mount::)'
+```
+
+### Step 4: Update Progress
+Mark `[x]` when done. Mark `[-]` with notes if partial.
+
+### Step 5: Report
+```
+== Mount Fix Loop Iteration ==
+Item:    A1 — Fix FUSE+SSH EIO bug
+Status:  [x] Complete
+Details: Root cause was X, fixed by Y
+Next:    A2 — Enforce readonly on WCF + FP
+```
+
+## Rules
+
+- **One item per iteration** (unless items are trivially small)
+- **Always update progress.md**
+- **Respect dependency order**
+- **All tests must pass** before moving on
+- **Use nextest** (not `cargo test`)
+- **Commit after each phase milestone** (A complete, B complete, etc.)
+- **Use custom agents** per CLAUDE.md pipeline

.config/nextest.toml

@@ -13,6 +13,9 @@ max-threads = 4
 [test-groups.pty-tests]
 max-threads = 4
 
+[test-groups.mount-integration]
+max-threads = 8
+
 [profile.default]
 fail-fast = false
 leak-timeout = { period = "500ms", result = "fail" }
@@ -53,3 +56,8 @@ test-group = 'ssh-integration'
 filter = 'test(shell::) | test(spawn::)'
 test-group = 'pty-tests'
 
+# Mount tests use singleton servers that outlive individual test processes
+[[profile.default.overrides]]
+filter = 'test(mount::)'
+test-group = 'mount-integration'
+leak-timeout = { period = "1s", result = "pass" }