FAIL.

Yeah using url() with cron.php/update.php doesn't work since the test bot runs in a subdirectory apparently. I had the same problem with tests and drupalGet() with adding a permission to run update.php.

Wow, lots of bugfixes. And, nice ASCII art.

+    $edit = array(
+      '1[search content]' => 1,
+      // @todo Comments are added to search index without checking first whether
+      //   anonymous users are allowed to access comments.
+      '1[access comments]' => 1,
+      // @todo Without this permission, "Login or register to post comments" is
+      //   added to the search index.  Scratch that, comment.module did not even
+      //   add anything to search results.  Aforementioned link text must be
+      //   added via node links.
+      '1[post comments]' => 1,
+    );

Should be using
DRUPAL_ANONYMOUS_RID . '[search content]' => 1, etc
Side node, I should really get back on #300993: User roles and permissions API so we don't have to use code like this in tests...

+    // Create a node.
+    $edit = array(
+      'title' => $this->randomName(),
+    );
+    $this->drupalPost('node/add/page', $edit, t('Save'));
+    // Check that the node exists in the database.

Why not use drupalCreateNode() since you're not doing anything special?

And yes, the ASCII art is funny/cool...but seriously...

Since there were no steps posted to reproduce the bugs identified here (grumble, mutter), I had to reverse-engineer the test to figure it out. I came across some things when I did.

+    // Enable comments for 'page' nodes.
+    variable_set('comment_page', COMMENT_NODE_OPEN);
+    variable_set('comment_preview_page', COMMENT_PREVIEW_OPTIONAL);

There's already enough going on in this test; let's not muddy the waters more by adding comments to Pages as well; just use article. If we don't have coverage for comments on Pages, let's put it in comment.test in a separate issue, since it seems unrelated to these bugs here.

--- modules/filter/filter.test	25 Apr 2009 18:01:10 -0000	1.19
+++ modules/filter/filter.test	28 Apr 2009 02:27:59 -0000

Speaking of test locations, I realize that the bugs ultimately lie in the filter system, but the test is clearly an integration test of the search functionality, and should be relocated to search.test IMO. I can't forsee anyone ever finding this test in filter.test once all of the underlying bugs are fixed.

+    // Enable check_plain() for 'Filtered HTML' text format.
+    $edit = array(
+      'filters[filter/0]' => 1,
+      'filters[filter/1]' => 1,
+      'filters[filter/2]' => 1,
+      'filters[filter/3]' => 1,
+      'filters[filter/4]' => 1,
+    );

Why do we need to specify 1 for each of them, if we're only interested in "Escape all HTML"? Again, let's please keep the test focused only to what it needs to do to exhibit the buggy behaviour. Unless this is required, in which case, please clarify in the comment. Also, please add the "Escape all HTML" keywords to the comment as well.

+    /**
+     *                     ,____
+     *                     |---.\
+     *             ___     |    `
+     *            / .-\  ./=)
+     *           |  |"|_/\/|
+     *           ;  |-;| /_|
+     *          / \_| |/ \ |
+     *         /      \/\( |
+     *         |   /  |` ) |
+     *         /   \ _/    |
+     *        /--._/  \    |
+     *        `/|)    |    /
+     *          /     |   |
+     *        .'      |   |
+     * jgs   /         \  |
+     *      (_.-.__.__./  /
+     * @see http://drupal.org/cvs?commit=79939
+     */

:P

+    $this->assertText($title, t('Comment body text found in search results.'));

Forgive my ignorance, but how does this prove anything? Unless I'm really tired, you're using the same text for both the title of the node and the body of the comment, so this test will pass even without this patch.

Fixed on all counts

this time for real

Looks good. Committed to CVS. Thanks for the patch, and the reviews.

Committed. Thanks!