What causes "auto-created mountpoint is not empty: {}" in astrid-runtime/astrid?

Inspect the directory and move out any files you need, then re-run the unmount/cleanup.

auto-created mountpoint is not empty: {}

Error message

auto-created mountpoint is not empty: {}

What it means

When tearing down a mountpoint that this provider auto-created, cleanup_created_mountpoint() refuses to remove the directory if it is no longer empty after unmount — something wrote files into it (or a failed mount left debris), and deleting a non-empty directory could destroy user data.

Source

Thrown at crates/astrid-storage-provider-fskit/src/main.rs:418

        errors.push(format!("registry: {error:#}"));
    }
    if errors.is_empty()
        && let Err(error) = cleanup_created_mountpoint(mountpoint, auto_created)
    {
        errors.push(format!("cleanup: {error:#}"));
    }
    rollback_outcome(true, &errors)
}

fn cleanup_created_mountpoint(mountpoint: &Path, auto_created: bool) -> Result<()> {
    if !auto_created {
        return Ok(());
    }
    validate_unmounted_mountpoint(mountpoint)?;
    let mut entries = std::fs::read_dir(mountpoint)
        .with_context(|| format!("read auto-created mountpoint {}", mountpoint.display()))?;
    if entries.next().is_some() {
        bail!(
            "auto-created mountpoint is not empty: {}",
            mountpoint.display()
        );
    }
    std::fs::remove_dir(mountpoint)
        .with_context(|| format!("remove auto-created mountpoint {}", mountpoint.display()))
}

fn prepare_mountpoint(
    requested: Option<PathBuf>,
    view: &astrid_core::storage_provider::StorageProviderViewV1,
) -> Result<(PathBuf, bool)> {
    let mountpoint =
        requested.map_or_else(|| default_mountpoint(std::env::var_os("HOME"), view), Ok)?;
    validate_mountpoint_layout(&mountpoint)?;
    let existed = match std::fs::symlink_metadata(&mountpoint) {
        Ok(_) => true,
        Err(error) if error.kind() == std::io::ErrorKind::NotFound => false,

View on GitHub (pinned to affd8760f4)

Solutions

Inspect the directory and move out any files you need, then re-run the unmount/cleanup.
Verify the fs was actually unmounted — if a hidden native mount still holds files, unmount it first.
Remove the directory manually only after confirming its contents are not needed.
Avoid pointing mounts at directories other processes write to; use a dedicated default mountpoint.

Defensive patterns

Strategy: validation

Validate before calling

fn mountpoint_is_empty(path: &Path) -> std::io::Result<bool> {
    Ok(std::fs::read_dir(path)?.next().is_none())
}
// call before unmount/cleanup; if false, surface files to the user instead of removing

Try / catch

if let Err(e) = cleanup_created_mountpoint(&mountpoint, auto_created) {
    if e.to_string().contains("auto-created mountpoint is not empty") {
        // leave directory intact; notify user to inspect and remove manually
    }
}

Prevention

Never store user data directly in the mountpoint root that the provider auto-creates.
Avoid concurrent unmounts of the same path (use the registry as a lock).
Check for leftover daemon artifacts (journals, .DS_Store) before cleanup.
If you need the directory contents, move them out before unmounting.

When it happens

Trigger: After unmount or rollback, validate_unmounted_mountpoint passed but read_dir still finds entries (entries.next().is_some()); triggered from mount, unmount, and rollback_after_native_failure for auto-created mountpoints only.

Common situations: User or another process placed files in ~/Astrid/<principal> while the fs was mounted or right after unmount; a crashed fskit daemon left journal files inside the mountpoint; race between two concurrent unmounts.

Related errors

AI-assisted analysis of astrid-runtime/astrid@affd8760f4 (2026-09-09). Data as JSON: /api/errors/09e04c6d3c825ea3. Report an issue: GitHub.