I just updated my server to Debian Lenny, which was released on February 14th.

So far, the only issue seems to have been my WordPress installation. I was getting errors in the Apache error log that the wordpress.wp_terms table didn’t exist in the database. I had to manually browse to /wp-admin/upgrade.php to do a database upgrade. After this, everything worked as normal.

So, if you come across a WordPress blog with no posts, no categories, and saying “Nothing to display for this category”, be kind and go /wp-admin/upgrade.php for them and run the database upgrade.

Thanks to this post on Erik Osterman’s blog for pointing out the issue.

It’s been a while since I posted anything up here, so I’m going to try and cover a few topics now. I started a new job developing payments routing software with ACI Worldwide in February, so I’ve been busy enough over the last few weeks.

To start with, I’ll mention http://photos.jmadden.eu, which I’ve just set up. This is a Plogger install that I can fill up with our photos. At the moment, all that’s up there are photos of Lou’s hen night in Killarney this year.

That leads me nicely on to my stag weekend, which was this past weekend (4th, 5th & 6th of April). I had a brilliant weekend, really enjoyed it & it was brilliant to have such a great bunch of friends around for the few days celebration! Thanks to everyone who made it. When I get a few photos in from it, I’ll put them up too.

While on the topic of photos, I also grabbed PloggerPress to integrate Plogger with WordPress, so I could have pages & posts of just photos. It works perfectly, but I’ve made a few changes to make it work a bit better for what I want. Firstly, I found a minor bug in it that I notified the developer about this evening. On line 150 of plogger-press.php, change

$url=substr($path, $number);

to

$url=preg_replace(“/[ |(|)]/”, “_”, substr($path, $number));

Plogger replaces certain characters (the one’s I noticed for the above change were ‘ ‘, ‘(‘ and ‘)’) with underscores. This ended up with some of the thumbnails for photos not being shown because the link for the ‘img src’ was incorrect.

As well as this, I added an ‘add_filter(“wp_head”, “my_head_func”);’ so I could alter some styling. I put a ‘class=”image”‘ into the ‘<img>’ tags so I could style them (borders, padding etc.) just to look a bit better.

Lastly, I’ve had very little time to work on SvnFs lately. I’ve done a small bit of work on getting write support done, but if you want it any time soon, I’d advise grabbing it from http://www.jmadden.eu/svnrepos/svnfs/trunk. It’s available anonymously, so

svn co http://www.jmadden.eu/svnrepos/svnfs/trunk

That’s it for now. We’re only three and a half weeks away from the wedding, so I doubt there’ll be many more updates before that. I’ll put up the stag photos when I get them, but that’ll probably be it!!

Having played with various mencoder options at different times, and with no success, I had a look at converting a standard avi file into an iPod recognisable mp4 again today.

Luckily this time I came across a post to one of the Mplayer mailing lists from someone having similar issues, and a link to “HOWTO batch encode video for iPod under Linux”.

He’s written a tool called podencoder which will encode DVDs and files into iPod mp4 format. After downloading that shell script to a directory in my $PATH, renaming it to podencoder and setting it executable, running

podencoder file.avi

produced file.avi.mp4 which iTunes recognised and was synced with the iPod.

On a related note, why can’t iTunes report back to you when it can’t import a certain file? Even if it gave no reason, just saying “I’m not importing that, now piss off” would be better than nothing.

Anyway, having searched for this type of thing before and coming up blank, maybe another post and link to that site will push it further up the search ratings.

…or at least according to Microsoft’s latest patch count!

I read this article, titled “Microsoft: Vista Least-Flawed OS” this morning. Microsoft counted the number of patches required to a variety of different operating systems, and claim that, because Vista required the least number of patches in its first year, it must be more secure than everything else out there.

This, of course, completely ignores the criticality and exploitability of the vulnerabilities in question, as pointed out by Rich Mogul in the above article.

Also, Microsoft have always been in the habit of counting every single bug that shows up in the GNU/Linux distros they’re using for comparison. This is pointless, as the distros contain thousands of packages, some of which are core requirements (like the kernel, shells, core utils etc.) and the rest are optional extras (eg. OpenOffice.org, XOrg etc.).

I use Debian Linux on my desktop, my laptop and a few servers. I use Debian Stable on the servers, and the number of security updates to these servers is quite low and infrequent. As well as this, in the recent past I can only recall one instance where an update required downtime to complete. I installed server specific packages, omitting a GUI and GUI applications for the servers, since they’re not required. On my desktop and laptop, I use the Gnome desktop environment, but KDE is also available as an option. I generally install OpenOffice.org as the office suite, but again, there are other packages provided to choose from for this. Basically, what I’m trying to illustrate is that Linux distros provide much more choice than Windows does, and include packages that Windows doesn’t (eg. office suite). Microsoft try to hold this against the distros by counting bugs identified in every single package, and not taking into account the number of affected users.

These security analyses that Microsoft do aren’t comparing like with like, and shouldn’t be considered as anything more than marketing fluff.

This is primarily to inform visitors to this website that I’m available and looking for contract work, either remote or on-site in the Limerick region.

I specialise in C and C++ software development on Unix platforms. I have extensive experience with scripting languages, such as Perl and Bash, on these platforms also. I’ve worked on everything from small open source projects to large scale high-availability software. I’ve also done quite a bit of web development using multiple web technologies such as PHP, Javascript, CSS and database backends.

As well as this, I have experience with server installation and administration of various elements of information technology infrastructure, such as web servers (Apache), database servers (MySQL, PostgreSQL, Sybase), email servers (Postfix), DNS (Bind) and directory servers (OpenLDAP).

My CV is available online. Any and all reasonable offers or requests will be considered and treated professionally and confidentially. Contact me at john@jmadden.eu.

In my Comments posting earlier this year, I said I had disabled comments on articles on this website to avoid having to deal with spam on another front.

As it turns out, just disabling comments within WordPress wasn’t enough. I was getting hit with more and more spam, all being held for moderation. This was getting to be an annoyance, so I looked into it.

Using grep on the Apache logs, and searching for the IP address that the spammers hit my site from, I found everything was being POSTed to wp-trackback.php. Simply changing the permissions on this file to 0600 (it was installed from a Debian package, so is root owned) has stopped the onslaught of spam, and doesn’t seem to have affected anything else.

I’ve just uploaded svnfs-0.4.tgz.

The major change in this version is to remove the requirement for a list of the repository at mount time, in favour of doing a list of the required directory when it’s needed. This slows down commands and tab completion, with the trade off of giving quicker mount times and always giving the latest contents of the repository.

As well as this, a few bugs were fixed.

• listing of files and directories in the root of the mounted filesystem
• segfault caused by an off-by-one in _svnclient_list_func()
• don’t return subdirectory files in svnfs_readdir()

As usual, any comments or patches welcome to john+svnfs@jmadden.eu.

I’ve updated SvnFs further to fix RHEL build problems (thanks to Niall Donegan for his help testing the numerous proposed fixes).

Version 0.3 and upwards require version 1.4 or higher of the Subversion development libraries to compile and run. I haven’t yet updated the configure script to check for this, so this is likely the problem if you get compile errors similar to

svnclient.c:48: error: syntax error before ‘*’ token
svnclient.c: In function `_svnclient_list_func’:
svnclient.c:53: error: `path’ undeclared (first use in this function)
svnclient.c:53: error: (Each undeclared identifier is reported only once
svnclient.c:53: error: for each function it appears in.)
svnclient.c:56: error: `dirent’ undeclared (first use in this function)
svnclient.c: In function `svnclient_readdir’:
svnclient.c:81: error: `SVN_DIRENT_KIND’ undeclared (first use in this
function)
svnclient.c:81: error: `SVN_DIRENT_SIZE’ undeclared (first use in this
function)
svnclient.c:102: warning: initialization makes pointer from integer
without a cast
svnclient.c: In function `svnclient_read’:
svnclient.c:134: warning: assignment makes pointer from integer without
a cast
make[2]: *** [svnclient.o] Error 1
make[2]: Leaving directory `/root/svnfs/src’
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/root/svnfs’
make: *** [all] Error 2

I’ve decided to work with the latest API, so I won’t be “fixing” these issues to work with older deprecated API functions.

As usual, this release is available from the SvnFs page.

I’ve uploaded the latest version of SvnFs. It’s available from the SvnFs page.

This provides read-only support for Subversion repositories. There are still some issues that need to be ironed out (design issues rather than bugs).

1. Doing a recursive list of the repository at mount time is slow
2. Reading large files is slow, due to lack of caching

Other than that, from my testing, it works pretty reliably.

I did come across an awkward bug though, which I now can’t reproduce! Basically, between the time I created the Subversion repository for SvnFs and now I’ve switched from en_IE@euro as my locale to en_IE.UTF8. Something in there caused Subversion to get mixed up and incorrectly report the filesize in the svn_dirent_t structure filled by svn_client_list(). This meant that the filesystem thought the files were smaller than they actually were, and was causing short reads. I worked around it by removing and re-adding all the files that were showing incorrect sizes, just for my own testing.

I’m still working, albeit slowly, on SvnFs. I’ve implemented open() and read() functions, but there are still bugs to be ironed out, so I’m not releasing it yet.

However, I’ve finally gotten around to setting up anonymous access to the code, so the latest development version can be gotten from the Subversion repository with

svn co http://www.jmadden.eu/svnrepos/svnfs/trunk

Testing reports and patches graciously accepted!

The two major bugs to be fixed are:

1. Reading large files causes a segmentation fault
2. The directory listing of a repository lists files that have been deleted or removed

Once those two are fixed, I think v0.3 should be ready for release, and should offer a complete readonly filesystem of a Subversion repository.

After that, I’ll need to move some code around. Doing a full listing of a Subversion repository at mount time is very slow, and probably not needed – particularly for large repositories. So that will have to change to be more scalable.