cdn_file_url_alter's path blacklist by default has admin*. AdvAgg CDN tests on status page not working as a result.

The above event / description is based on using Wim Leers CDN module project;

http://drupal.org/project/cdn

Notes: The second php code listing (above) is applied in D6's settings.php

See https://bugs.launchpad.net/pressflow/+bug/597718
for content relating to how this new file_create_url() should be integrated into latest Pressflow.

[Changes:] After some research on this D6 file_create_url() use, I find that Drupal's (D6) method is flagged as 'buggy', and that Pressflow has 'corrected' the use (@return) of file_create_url(). Duh *beware* D6 coders...

Hence, Pressflow will not process parameters the 'OLD' D6 way. The returned value @return is not what it was under Drupal D6.

In the current case of 'advagg', this effects the value (URI expected) passed to drupal_http_request().

In my testing, this 'change' effects; Pressflow 6.20 to Pressflow 6.22

Adding Wim Leers 'comment' related to this: -> http://drupal.org/node/839282#comment-3358288

Refer #4

Yes, I will outline the fix:

Simply, we need to feed advagg a full URI, as the current file_create_url() fails to do this...

The following edits allow this:

Step 1: edit advagg.install and change thus (the + lines)

  // Test CDN module; JS & CSS paths.
  if (module_exists('cdn')) {
    // Make sure CDN module is on.
    $status = variable_get(CDN_STATUS_VARIABLE, CDN_DISABLED);
    if (($status == CDN_ENABLED || ($status == CDN_TESTING && user_access(CDN_PERM_ACCESS_TESTING))) && !variable_get(CDN_THEME_LAYER_FALLBACK_VARIABLE, FALSE)) {

+      $url_cdn_css = variable_get('my_cdn_css_base_url', '') . file_create_url($css_path . $filepath . '.css');   // Append CDN CSS Base
-      $url_cdn_css = file_create_url($css_path . $filepath . '.css');  // remove
      $parts_css = @parse_url($url_cdn_css);
      // Do not test CDN CSS if the hosts are the same.
      if (strcmp($parts_css['host'], $_SERVER['HTTP_HOST']) == 0) {
        $parts_css = FALSE;
      }

+      $url_cdn_js = variable_get('my_cdn_js_base_url', '') . file_create_url($js_path . $filepath . '.js');       // Append CDN JS Base
-      $url_cdn_js = file_create_url($js_path . $filepath . '.js');       // remove
      $parts_js = @parse_url($url_cdn_js);
      // Do not test CDN JS if the hosts are the same.
      if (strcmp($parts_js['host'], $_SERVER['HTTP_HOST']) == 0) {
        $parts_js = FALSE;
      }

      if (!empty($parts_css)) {
        // Send request and also time it.
        timer_start(__FUNCTION__ . 'cdn_css');
        $data_cdn_css = drupal_http_request($url_cdn_css);
        $time_cdn_css = timer_stop(__FUNCTION__ . 'cdn_css');
      }
      if (!empty($parts_js)) {
        // Send request and also time it.
        timer_start(__FUNCTION__ . 'cdn_js');
        $data_cdn_js = drupal_http_request($url_cdn_js);
        $time_cdn_js = timer_stop(__FUNCTION__ . 'cdn_js');
      }
      $mode = variable_get(CDN_MODE_VARIABLE, CDN_MODE_BASIC);
    }
  }

Step 2: Edit advagg.admin.inc and change thus (the + lines)

  timer_start(__FUNCTION__ . 'local');
  $data_local = drupal_http_request($url, $headers);
  $data_local->timer = timer_stop(__FUNCTION__ . 'local');
  list($data, $rows) = advagg_form_print_r($data_local);
  $form['async']['normal'] = array(
    '#type'           => 'textarea',
    '#title'          => check_plain($url),
    '#default_value'  => $data,
    '#rows'           => min($rows+1, 50),
  );

  if (module_exists('cdn')) {
+    $url_cdn = variable_get('my_cdn_css_base_url', '') . file_create_url($filepath);        // Append CDN Base
-    $url_cdn = file_create_url($filepath);        // remove

    // Send request and also time it.
    timer_start(__FUNCTION__ . 'cdn');
    $data_cdn = drupal_http_request($url_cdn);
    $data_cdn->timer = timer_stop(__FUNCTION__ . 'cdn');
    list($data, $rows) = advagg_form_print_r($data_cdn);
    $form['async']['cdn'] = array(
      '#type'           => 'textarea',
      '#title'          => check_plain($url_cdn),
      '#default_value'  => $data,
      '#rows'           => min($rows+1, 50),
    );
  }

  return $form;
}

Step 3: add the following to your active settings.php:

if (file_exists('sites/all/modules/cdn/cdn.module')) {         // if Drupal Module CDN exists
  $my_cdn_css_url = 'http://yourcdnbase.com';     // base CDN url to add to CSS
  $my_cdn_js_url  = 'http://yourcdnbase.com';     // base CDN url to add to JS
  $conf['my_cdn_css_base_url'] = $my_cdn_css_url;
  $conf['my_cdn_js_base_url']  = $my_cdn_js_url;
}

Now the URI will be available for advagg checks.

Notes: The references to 'http://yourcdnbase.com' *need* to be changed for your cdn environment.
The 'yourcdnbase.com' [string] represents the missing 'URI' that file_create_url() fails to return (it returns only a relative path - instead of the URI to the missing 404 file location). We need a 'correct URI' to sent to the advagg drupal_http_request() **This is where the 'missing schema' error -1002 result comes from.** Note: we add the missing 'schema' (http or https).

See above code edits and previous notes on Drupal versus Pressflow use of this function.

Helpful Note: The error status showing:

[error] => missing schema
[code] => -1002

is basically telling us that a incorrect 'URI' was attempted / used. 'schema' refers to the 'http/https' (or lack of it - in this case) to the next called function: drupal_http_request()

Lets compare the actual [current] core for Drupal 6 vs Pressflow for file_create_url() (file.inc in /includes)..

1) Pressflow 6.x: file_create_url()
Notes: This comes from the latest Pressflow 6.22 core release

/**
 * Create the download path to a file.
 *
 * There are two kinds of local files:
 * - "created files", i.e. those in the files directory (which is stored in
 *   the file_directory_path variable and can be retrieved using
 *   file_directory_path()). These are files that have either been uploaded by
 *   users or were generated automatically (for example through CSS
 *   aggregation).
 * - "shipped files", i.e. those outside of the files directory, which ship as
 *   part of Drupal core or contributed modules or themes.
 *
 * @param $path
 *   A string containing the Drupal path (i.e. path relative to the Drupal
 *   root directory) of the file to generate the URL for.
 * @return
 *   A string containing a URL that can be used to access the file.
 */
function file_create_url($path) {
  // Clean up Windows paths.
  $old_path = $path = str_replace('\\', '/', $path);

  drupal_alter('file_url', $path);

  // Return path if it was altered by any module, or if it already is a
  // root-relative or a protocol-relative URL.
  if ($path != $old_path || drupal_substr($path, 0, 1) == '/' || drupal_substr($path, 0, 2) == '//') {
    return $path;
  }

  // Shipped files.
  if (preg_match("#^(/?)(misc|modules|sites|themes)/#", $path) && (strpos($path, file_directory_path() . '/') !== 0)) {
    return base_path() . $path;
  }
  // Created files.
  else {
    switch (variable_get('file_downloads', FILE_DOWNLOADS_PUBLIC)) {
      case FILE_DOWNLOADS_PUBLIC:
        return base_path() . $path;
      case FILE_DOWNLOADS_PRIVATE:
        // Strip file_directory_path from $path. Private downloads' URLs are
        // rewritten to be served relatively to system/files (which is a menu
        // callback that streams the file) instead of relatively to the file
        // directory path.
        if (strpos($path, file_directory_path() . '/') === 0) {
          $path = trim(substr($path, strlen(file_directory_path())), '\\/');
        }
        return url('system/files/' . $path, array('absolute' => TRUE));
    }
  }
}

2) Drupal 6.x: file_create_url()

function file_create_url($path) {
  // Strip file_directory_path from $path. We only include relative paths in urls.
  if (strpos($path, file_directory_path() . '/') === 0) {
    $path = trim(substr($path, strlen(file_directory_path())), '\\/');
  }
  switch (variable_get('file_downloads', FILE_DOWNLOADS_PUBLIC)) {
    case FILE_DOWNLOADS_PUBLIC:
      return $GLOBALS['base_url'] . '/' . file_directory_path() . '/' . str_replace('\\', '/', $path);
    case FILE_DOWNLOADS_PRIVATE:
      return url('system/files/' . $path, array('absolute' => TRUE));
  }
}

* With Pressflow, the base_path() call in file_create_url() comes back as "/". *

Peter

@mikeytown2

Hmm, interesting .....

Wim Leers (via cdn module), defines this 'default'; define('CDN_EXCEPTION_FILE_PATH_BLACKLIST_DEFAULT', "*.js\n*/image_captcha/*");

I had (previously) set the CDN blacklist to ONLY "*/image_captcha/*

I note that the patch (#13 above) only deals with the CDN blacklist 'default'??

@mikeytown2, I will test new advagg.git 'patch' - later today!

Refer #13

@mikeytown2
The only advagg 'hack' I should need is the single (1) for 'prefixing' my special CDN URI case-> 'http://small.gdlcdn.com/802C5F/cdn1' as I choose to not use domain name references (like small.gdlcdn.com.

I like avoiding the millisecond lag of using CDN 'CNAMES'. I have bench-marked it to be very small with Edgecast CDN use - about 10-18ms :)

So this 'custom' prefix / method is not a coding 'concern' for advagg...

Refer #17

@David Straus, I have outlined the Pressflow advagg method at #5.

See http://drupal.org/node/1172012#comment-4532666

It works great! (Spent a lot of time testing this)

Notes: I have *not yet* tested the new advagg.git patch submitted @ #13 -> http://drupal.org/node/1172012#comment-4541260

Update: Just tested #13 -> http://drupal.org/node/1172012#comment-4541260
It works!

*Solved*

Just tested #13 -> http://drupal.org/node/1172012#comment-4541260

The latest *.dev advagg.git solves this issue with Pressflow!

Good News for Pressflow advagg users!

Report Status shows:

Adv CSS/JS Agg - Asynchronous Mode	Already set to TRUE
Adv CSS/JS Agg - CDN Async Mode	JS CDN
Requests through the CDN are getting back to advagg for JS files.
Adv CSS/JS Agg - CDN Async Mode	CSS CDN
Requests through the CDN are getting back to advagg for CSS files.
Adv CSS/JS Agg - Socket Timeout	Already set to 1 seconds. Raw timer: 62.95
AdvAgg JS Compress - Callback	The callback is working correctly.

Thanks mikeytown2 :)

Refer #21 -> http://drupal.org/node/1172012#comment-4542992
@mikeytown2

Would be good if you did explain how the above 'issue' come to exist.

I was one of at least two Pressflow users that noticed this 'advagg bug'. Did you get any 'bug reports' from the many Drupal users (not using Pressflow)?

See this Pressflow bug report link for related 'issues' effected by the CDN module:
https://bugs.launchpad.net/pressflow/+bug/729783 (David Strauss wrote on 2011-05-31)