bug

Facebook Connect went down hard tonight. HuffPo reports that their site was redirecting to a Facebook error page, even when people weren’t attempting to log in.

Yep. Busted third-party JavaScript brings portions of the Internet to its knees: huffingtonpost.com/2013/02/07/fac…

— Kent Brewster (@kentbrew) February 8, 2013

It makes me more comfortable with our decision to strip so many 3rd party javascripts from GigaOM during our last redesign.

I just switched to a new server and found myself struggling with empty HTTP headers and broken or partial images. The problem is the memcache extension for PHP and WordPress MU/WordPress multisite’s need to reinstantiate the wp-cache after determining the correct blog for a given request.

Versions of the memcache extension prior to 3.0 go wrong somehow and it shows up when you try to do an HTTP HEAD request on a page (the result is empty) or enable X-SendFile support for WP MU/MS’ file handling (all the files and images in the media library will break). Upgrading to the the 3.x version (in beta since 2007) fixes the problem.

You may have to uninstall the old version before installing the beta, and installing the beta via PECL requires adding “-beta” to the package name. Here are the commands:

pecl uninstall memcache
pecl install memcache-beta

You’ve got to both respect Automattic for releasing their internal code as open source while also giving them a break for not assuring that it works for anybody else. One of their projects, the WordPress.com Video Server is a sophisticated WordPress plugin that handles video transcoding and offers a bit of a YouTube in a box solution for WordPress.

The bug I found is that the code assumes WPMU is running in subdomain mode, rather than subdirectory mode. This assumption causes an ajax request to go to the wrong URL and return the wrong data. The patch simply applies WordPress’ plugins_url() function that debuted in 2.6.

WordPress is built around the Loop, and all the cool kids are using multiple loops on the same page to show the main post and feature other posts. The problem is: WordPress doesn’t properly reset the $pages global for each post. If the post in main loop (or default query) is paged, then all the other posts will show the same paged content as in the main post. I started a ticket and submitted a patch, but in the meantime you might have to unset( $GLOBALS['pages'] ) in your custom loops just before calling the_post().

It may be a sign that none of the core WordPress developers much likes or uses Microsoft Office, but the core code hasn’t been updated to recognize the Office 2007 file extensions like .docx, .pptx, or .xlsx. It’s no criticism, wouldn’t have discovered it if a user hadn’t complained, and I stewed a bit before deciding it was a bug.

It’s now ticket #8194 in the WordPress.org Trac. It only affects my MU users now, though, and the same patch works there.

You’ll find a few mentions of doc, ppt, and xls in wp-includes/functions.php (around lines 1613-1615 and 1651-1654 in 2.6.3). Once corrected, those lines should look something like this:

Update: Sweet! It’s been accepted and should be in the WP2.7 release.

I just submitted a trac ticket about this:

The update_post_meta() and delete_post_meta() functions don’t know how to deal with post revision IDs. add_post_meta() does, it uses the following block of code to make sure the passed $post_id is a real post, not a revision:

This is important because the global $post_id when a post is being saved is for the revision, not the real post. If you pass that to these functions, update_post_meta() and delete_post_meta() will try to update or delete the records related to the revision. When update_post_meta() fails to find any records for the post_id you submitted, it calls add_post_meta(). Each attempt at updating the meta will actually create a new row in the table.

For now I’m checking with wp_is_post_revision() before making any calls to the post_meta functions.

Updated: Mark Jaquith committed the patch less than two hours after I submitted it! I love WP.

After an upgrade to MySQL 5.0.51b on RHEL 5 I started seeing curious results in a fairly common query. Here’s a simplified version:

What I expected was to get a handful of post ID numbers sorted in descending order by the post_date_gmt. Instead, I got a list of post IDs sorted in ascending order by the ID number. Something like this:

After some fiddling I discovered that the GROUP BY clause was causing a problem. So this query works:

…and outputs the results I expected:

The GROUP BY clause may be unnecessary, though it was originally written in to accommodate conditions where a JOIN (which is often added when the query is dynamically generated) causes MySQL to return multiple rows representing the same record ID.

Still, isn’t this behavior weird? It’s certainly different from previous versions.

PHP iCalendar solves a couple problems I’m working on, but I needed a solution to fix the duration display for Gcal-managed ICS calendars.

As it turns out, a fix can be found in the forums, and the trick is to insert the following code in functions/ical_parser.php.

case 'DURATION':
	if (($first_duration == TRUE) && (!stristr($field, '=DURATION'))) {
		ereg ('^P([0-9]{1,2}[W])?([0-9]{1,2}[D])?([T]{0,1})?([0-9]{1,2}[H])?([0-9]{1,2}[M])?([0-9]{1,}[S])?', $data, $duration); 
	
		$weeks = str_replace('W', '', $duration[1]);
		$days = str_replace('D', '', $duration[2]);
		$hours = str_replace('H', '', $duration[4]);
		$minutes = str_replace('M', '', $duration[5]);
		$seconds = str_replace('S', '', $duration[6]);
		
		// Convert seconds to hours, minutes, and seconds
		if ($seconds > 60) {
			$rem_seconds = $seconds % 60;
			$minutes = $minutes + (($seconds - $rem_seconds) / 60);
			$seconds = $rem_seconds;
		}
		if ($minutes > 60) {
			$rem_minutes = $minutes % 60;
			$hours = $hours + (($minutes - $rem_minutes) / 60);
			$minutes = $rem_minutes;
		}
		
		$the_duration = ($weeks * 60 * 60 * 24 * 7) + ($days * 60 * 60 * 24) + ($hours * 60 * 60) + ($minutes * 60) + ($seconds);
		$first_duration = FALSE;
	}
	break;

Hopefully this gets worked into the baseline with the next release.