NotOnServer only shows files if parent directory of missing file exists

I suspect the culprit might be this line in Drupal\auditfiles\Auditor\AuditFilesNotOnServer:

$target = $this->fileSystem->realpath($result->uri);

This is converting a URI found in the file_managed table to a realpath, but it appears to return false if the file parent directory doesn't exist when it needs to return the full path to the missing file.

For context, here's the entire method:

  public function getReferences(): \Generator {
    $maximumRecords = $this->auditFilesConfig->getReportOptionsMaximumRecords();
    $query = $this->connection->select('file_managed', 'fm');
    $query
      ->orderBy('changed', 'DESC')
      ->range(0, $maximumRecords)
      ->fields('fm', ['fid', 'uri']);
    /** @var array<object{uri: string, fid: string}> $results */
    $results = $query->execute()->fetchAll();
    foreach ($results as $result) {
      $target = $this->fileSystem->realpath($result->uri);
      if ($target !== FALSE && !file_exists($target)) {
        yield FileEntityReference::create((int) $result->fid);
      }
    }
  }

Ref: https://www.php.net/manual/en/function.realpath.php

Note:

The running script must have executable permissions on all directories in the hierarchy, otherwise realpath() will return false.

If a parent directory doesn't exist then realpath will return false, excluding it from this module's file_exists check and therefore not showing it in the "not on server report".

Hi there,

I was able to reproduce the issue.
I fixed merge conflicts and updated the MR

+1

@jonathanshaw I'm not sure what you mean, can you explain the exact scenario and how to replicate the problem?

Actually I was too hasty. The problem with stream wrappers that don't support realpath already exists and this doesn't worsen it.

But it could fix it:

$base_dir = $this->fileSystem->realpath($scheme . '://');
if ($base_dir) {
  $target = $base_dir . '/' . $path;
}
else {
  $target = $result->uri;
}

I'm not sure why we're using realpath here, but we should at least have a fallback when it's not available. My use case is s3fs.

I've provided a fallback to manually convert paths to absolute, copying what s3fs decided to do in #3264640: S3fs does not handle paths with './' or '..' the same as core. → . Sorry for somewhat hijacking this issue, hopefully you agree these concerns are closely entangled.