CAS-ifying Roundcube 0.4.2

I've been testing Roundcube 0.4.2 as a replacement for our now-decrepit Squirrelmail 1.4.17 Webmail. I've been extremely impressed so far. It does natively or there is a plugin for almost everything I can think of. It acts so much like an application that I found myself angry that I couldn't dynamically resize the Subject/From/Date columns in the folder view. It had tricked me so much into thinking it was a native app that when the illusion broke I was actually confused.

Only one plugin has given me trouble so far, and that's the CAS plugin. For those who aren't familiar with CAS, it is an open-source Central Authentication System that many campuses are using these days for their portals. The benefits to us for using CAS for Roundcube in addition to our portal would be that users only have to sign in once and they can access either Roundcube or the portal without needing to authn again (as long as their cookie/session on the CAS server is still good). We've been trying to convert as many open-source login pages as we can to use CAS.

The problem is perhaps best described here:

http://code.google.com/p/rc-cas-plugin/issues/detail?id=1

Basically even if you do everything right, you will get a redirect loop because somewhere between the version of Roundcube this plugin was written for and 0.4.2, the code to kill/regenerate the PHP session changed. I made some comments in that ticket that include what I was able to do in order to actually log in (comment out a call to kill_session()) but that is not a very clean solution, and it doesn't solve the problem of users logging in from a different browser or IP. Really, this plugin needs work.

Hopefully this helps someone or helps remind me what the problem was should I find the time to work on fixing this plugin.

Z-Push Woes

Z-Push (http://z-push.sourceforge.net/) is an interesting project. It is a PHP web app that lets you serve different backends for mail/contacts to clients that support ActiveSync. The most common use for it seems to be to give devices with zero or minimal IMAP support (but decent ActiveSync support) access to their mail by letting them connect to that IMAP through ActiveSync.

The first thing that bothered me was it REALLY wants to run under mod_php in Apache. People have been able to get it to run in fastCGI by adding the following to compat.php:

if (!function_exists("apache_request_headers")) {
    function apache_request_headers() {
        $headers = array();
        foreach ($_SERVER as $key => $value) {
            if (substr($key, 0, 5) != 'HTTP_') {
                continue;
            }
            $headername = strtr(ucwords(strtolower(strtr(substr($key, 5), '_', ' '))), ' ', '-');
            $headers[$headername] = $value;
        }

        return $headers;
    }

(Taken from http://z-push.sourceforge.net/phpbb/viewtopic.php?f=2&t=36#p1349)
Sadly, the z-push developers didn't accept that patch.

The second problem I ran in to was z-push was segfaulting in PHP's imap_search function. I looked at the raw logs and saw that the client was getting a good response from the search request, but PHP still segfaulted. Some investigation led me to suspect the SE_FREE flag, per this PHP bug report: http://bugs.php.net/48619

Basically I changed the following line in backend/imap.php:

$search = @imap_search($this->_mbox, "SINCE ". date("d-M-Y", $cutoffdate));

to look like:

$search = @imap_search($this->_mbox, "SINCE ". date("d-M-Y", $cutoffdate), ! SE_FREE);

and things started working. So if your PHP is newer than 5.2.10 you hopefully don't have to deal with this. But if not, hopefully this helps you.

Update Fink For 10.6

Lots of people may have figured this out, but if you're getting the following error when trying to install a package with Fink in Mac OS 10.6:

Use of uninitialized value $darwin_osx in numeric eq (==) at /sw/lib/perl5/Fink/Services.pm line 1404, line 1.

Use of uninitialized value $darwin_osx in concatenation (.) or string at /sw/lib/perl5/Fink/Services.pm line 1404, line 1.

Argument "10.6 does not match the expected value of . Please run `..." isn't numeric in exit at /sw/lib/perl5/Fink/Services.pm line 1404, line 1.

 You can change line 1448 of /sw/lib/perl5/Fink/Services.pm look like this:

my %darwin_osx = ('1' => '10.0', '5' => '10.1', '6' => '10.2', '7' => '     10.3', '8' => '10.4', '9' => '10.5', '10' => '10.6');

Basically I added '10' => '10.6' after '9' => '10.5'. And I was able to install things again.

Windows Media Player 12 UPNP Client Sucks.

I was impressed when Windows Media Player 12 saw my Mediatomb UPNP server I use to stream video to a Playstation 3. "Now I don't have to use iTunes/mt-daapd to listen to my music library," I thought.

Then I browsed my music. I'm not exactly sure who is at fault, but either way Media Player is pretty stupid in that it has no way I could find to clear it's cache of what it thinks the UPNP server is offering. The only way to start fresh I've found was to go to C:\Users\%you\AppData\Local\Microsoft\Media Player and delete the files that start with "_HME_home" with Media Player not running. Next time you start Media Player, it should re-index all files.

What I can't figure out it why it presents so many duplicate entries unless I remove the artist/album containers that Mediatomb creates.

Netapp + Rquotad + NFS + RHEL5

Those mounting volumes via NFS on a Netapp filer wanting to use rquotad on a volume with qtrees may notice that, under some conditions, things do not work.

If you are using qtrees, you can still get rquota to work by mounting each qtree individually, each with the "nosharecache" option. This forces each mounted qtree to have different st_dev values so that remote quota calls are enabled for all mounted qtrees, instead of just the first one. Without "nosharecache", all mounts from a volume on a Netapp filer will have the same st_dev value and any rquota calls will only query the first mounted qtree.

See http://linux.die.net/man/5/nfs for info on the "nosharecache" . . . there are some caveats to using it.

Create Maildir Folders

mkdir -m 700 $MAILBOX/{,tmp,new,cur}

Substitute $MAILBOX for what you want to call your Maildir folder.

(Taken from http://plasmasturm.org/log/81/)