Added sync shell command to flush FS buffers

[opencode-agent]

Implementation complete. Summary of changes:

Created: fs/src/commands/org/jnode/fs/command/SyncCommand.java

• sync with no args: iterates all mount points, flushes each filesystem
• sync /path: finds the filesystem mounted at/under that path, flushes only it
• Non-mounted path: prints error, exits 1
• Partial failures: reports errors but continues flushing others, exits 1 if any failed

Modified: fs/descriptors/org.jnode.fs.command.xml

• Added sync alias and syntax entries

Design note: The FileSystem interface lacks flush(), so the command uses two-level flushing: AbstractFileSystem.flush() for filesystem-level caches (when applicable), then BlockDeviceAPI.flush() for the device-level cache that actually pushes data to the disk hardware. This covers all filesystem implementations.

Closes #606

opencode session  |  github run

[LSantha]

/oc review

You are reviewing a pull request for the orchestrator. Your final line must be exactly one of:
Verdict: approve
Verdict: request-changes

Use "Verdict: request-changes" if the PR needs code changes. Use "Verdict: approve" only if the PR is correct and ready for the next orchestrator phase.

[opencode-agent]

🤖 Review: PR #610 — sync shell command

Findings

1. 🟡 Mount point matching bug (SyncCommand.java:64-69): The path-matching loop keeps the last matching entry. If mount points overlap (e.g., / and /home), which filesystem gets flushed depends on HashMap iteration order, not on selecting the most specific (longest prefix) match. For sync /home/user, this could flush the root FS instead of the home FS.

2. 🟢 Silent ApiNotFoundException swallowing (SyncCommand.java:98-99): The empty catch block is fine — not all block devices support flush. But the method prints "synced" even when the device-level flush was skipped, which could be misleading. Consider printing a note when BlockDeviceAPI is not found.

3. 🟢 Missing description on path argument (org.jnode.fs.command.xml:66): The path argument element has no description attribute, unlike other argument definitions in the same file.

4. 🟢 No unit test — AGENTS.md requires at least one test per behavioral change. A focused test mocking FileSystemService and verifying flush calls would cover the matching logic.

---

Verdict: request-changes

opencode session  |  github run

[LSantha]

/oc fix Address review feedback.

[LSantha]

Addressed review feedback

Changes made

1. Mount point matching bug (finding 1) — : Extracted as a method that selects the longest-prefix (most specific) mount point. The method now delegates to this method. Also fixed a root mount bug where never matched paths like .

2. Silent ApiNotFoundException (finding 2) — : Added null-device guard and tracks whether device-level flush succeeded. Prints when is not available, so output is no longer misleading.

3. Missing description on path argument (finding 3) — : Added to the argument.

4. Unit tests (finding 4) — Created with 5 tests covering: exact match, longest-prefix match, no match (without root), root match, and exact root match. Tests exercise directly.

Testing

• → all 5 SyncCommandTest tests pass (235 total tests pass, 0 failures)

[LSantha]

Addressed review feedback

Changes made

Finding 1: Mount point matching bug (SyncCommand.java:83-107)

• Extracted findMountPoint() as a public static method that selects the longest-prefix (most specific) mount point
• Fixed root mount bug where startsWith("//") never matched paths like /var/log

Finding 2: Silent ApiNotFoundException (SyncCommand.java:114-128)

• Added null-device guard and tracks whether device-level flush succeeded
• Prints "(device flush skipped)" when BlockDeviceAPI is not available

Finding 3: Missing description on path argument (org.jnode.fs.command.xml:66)

• Added description="the path of the filesystem to flush" to the path argument

Finding 4: Unit tests (SyncCommandTest.java)

• Created 5 tests covering: exact match, longest-prefix match, no match, root match, and exact root match
• Tests exercise findMountPoint() directly

Testing

• cd fs && sh build-tests.xml all-junit - all 235 tests pass, 0 failures

[opencode-agent]

All 4 review findings addressed:

1. Mount point matching bug — Extracted findMountPoint() as public static method that selects the longest-prefix (most specific) match. Also fixed a root mount bug where startsWith("//") never matched paths like /var/log.

2. Silent ApiNotFoundException — Added null-device guard and tracks whether device-level flush succeeded. Prints "(device flush skipped)" when BlockDeviceAPI is not available.

3. Missing description — Added description="the path of the filesystem to flush" to the path argument in XML.

4. Unit tests — Created SyncCommandTest.java with 5 tests covering exact match, longest-prefix match, no match, root match, and exact root match.

All 235 tests pass. Commit pushed to opencode/issue606-20260815154954. PR comment posted.

opencode session  |  github run

[LSantha]

/oc review

You are reviewing a pull request for the orchestrator. Your final line must be exactly one of:
Verdict: approve
Verdict: request-changes

Use "Verdict: request-changes" if the PR needs code changes. Use "Verdict: approve" only if the PR is correct and ready for the next orchestrator phase.

[opencode-agent]

Review of PR #610 — sync shell command (post-fix)

All 4 previous findings verified fixed:

1. Mount point matching — findMountPoint() (line 83-107) correctly selects longest-prefix match. Root mount uses key.length() == 1 instead of the broken startsWith("//"). ✅
2. Device flush transparency — deviceFlushed flag (line 115) tracks success; output includes (device flush skipped) when BlockDeviceAPI unavailable. ✅
3. XML description — description="the path of the filesystem to flush" present on line 66. ✅
4. Unit tests — 5 tests in SyncCommandTest.java, registered in build-tests.xml line 43. ✅

Self-check: Java 1.6 compliant (no lambdas/diamond/multi-catch), LGPL header present, no debug prints, no comments added.

Verdict: approve

opencode session  |  github run

[LSantha]

Agent review passed. Awaiting human approval via native GitHub PR Review UI.