Fix flaky MountVolume tests: replace fixed delays with polling loops

[Copilot]

Description

Directory_Delete_MountVolume and Directory_ReparsePoints_MountVolume are flaky on loaded CI machines because fixed-duration waits are insufficient for NTFS mount point operations to propagate.

Delete_MountVolume.cs

7 locations used Task.Delay(300).Wait() before asserting !Directory.Exists() after deleting through a mount point. 300 ms is not enough under load.

• Added WaitForDirectoryGone(string path): polls Directory.Exists every 100 ms for up to 10 s using Environment.TickCount64 for accurate elapsed tracking
• Replaced all 7 fixed delays with WaitForDirectoryGone(<path>)
• Replaced Task.Delay(300).Wait() in the DeleteDir retry loop with Thread.Sleep(300); removed unused System.Threading.Tasks import

// Before
Directory.Delete(dirNameReferredFromMountedDrive, true);
Task.Delay(300).Wait();
Eval(!Directory.Exists(dirName), "Err_20387g! ...");

// After
Directory.Delete(dirNameReferredFromMountedDrive, true);
WaitForDirectoryGone(dirName);
Eval(!Directory.Exists(dirName), "Err_20387g! ...");

ReparsePoints_MountVolume.cs

DeleteDir (called in finally blocks after MountHelper.Unmount) had no retry logic, so transient IOException during volume teardown would fail cleanup silently or throw.

• Replaced the single-shot Directory.Delete with a retry loop: catches IOException specifically (which can occur transiently when the volume is still being unmounted — the reparse point directory may be momentarily locked by the kernel while the mount is being torn down), retries up to 10× with 200 ms back-off
• Added explanatory comment on the catch (IOException) block documenting the observed transient failure mode
• Added using System.Threading

Changes

• Directory/Delete_MountVolume.cs — polling helper + 7 delay replacements
• Directory/ReparsePoints_MountVolume.cs — robust DeleteDir with IOException retry and explanatory comment

Testing

Tests are Windows-only ([PlatformSpecific(TestPlatforms.Windows)], requires NTFS) and require an elevated environment with mount point access. Validation requires observing reduced flakiness on CI.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>MountVolume test flakiness: Delete_MountVolume and ReparsePoints_MountVolume</issue_title>
<issue_description>> [!NOTE]

This issue was created with Copilot assistance.

Tracking flaky MountVolume tests that are hitting blocking-clean-ci:

• System.IO.Tests.Directory_Delete_MountVolume.RunTest failure #125295 — Directory_Delete_MountVolume.RunTest (16 hits/month)
• Directory_ReparsePoints_MountVolume test failures #125624 — Directory_ReparsePoints_MountVolume.runTest (4 hits/month)
• System.IO.Tests.Directory_Delete_MountVolume.RunTest occasionally fails in CI #125780 — closed as dupe of System.IO.Tests.Directory_Delete_MountVolume.RunTest failure #125295

Root cause (Delete_MountVolume): 7 locations use a fixed Task.Delay(300) after Directory.Delete through NTFS mount points, then assert !Directory.Exists(). 300ms is insufficient on loaded CI machines.

Root cause (ReparsePoints_MountVolume): Needs investigation — no Task.Delay in the test; may be mount setup/teardown timing.

Proposed fix: Replace fixed delays with a polling loop (retry Directory.Exists with backoff, ~10s timeout).

Known Issue Error Message

Fill the error message using step by step known issues guidance.

{
  "ErrorMessage": "",
  "ErrorPattern": "",
  "BuildRetry": false,
  "ExcludeConsoleLog": false
}

</issue_description>

<agent_instructions>create a pr with the recommended test-only fix. the PR should be set to close the open bugs when merged</agent_instructions>

Comments on the Issue (you are @copilot in this section)

• Fixes MountVolume test flakiness: Delete_MountVolume and ReparsePoints_MountVolume #125913

---

📍 Connect Copilot coding agent with Jira, Azure Boards or Linear to delegate work to Copilot in one click without leaving your project management tool.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-io
See info in area-owners.md if you want to be subscribed.

[danmoseley]

add comment that we have seen this IOException occasionlly (and if possible why it might be)

[Copilot]

Added in 7348e64. The comment explains that IOException can occur transiently when the volume is still being unmounted — the reparse point directory can be momentarily locked by the kernel while the mount is being torn down.

[danmoseley]

@copilot address feedback

[Copilot]

Pull request overview

Reduces flakiness in Windows-only NTFS MountVolume filesystem tests by replacing fixed sleeps with polling/retry logic around directory deletion and cleanup.

Changes:

• Delete_MountVolume.cs: replaces multiple fixed 300ms delays with a WaitForDirectoryGone polling helper (up to 10s), and switches a retry-loop delay to Thread.Sleep.
• ReparsePoints_MountVolume.cs: hardens cleanup by retrying Directory.Delete on transient IOException during unmount teardown.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/libraries/System.Runtime/tests/System.IO.FileSystem.Tests/Directory/Delete_MountVolume.cs	Introduces a polling helper and removes fixed-delay assumptions after deletes via mount points.
src/libraries/System.Runtime/tests/System.IO.FileSystem.Tests/Directory/ReparsePoints_MountVolume.cs	Adds targeted retry logic for transient unmount-related IOException during directory cleanup.