The CDDB hash function has a high rate of collisions for single track CDs. Using the freedb.org database, the RH6.2 source cd comes up as In the corridor of uncertainty by Flacco and the Sandman.

On the dia list, we have been discussing writing a new canvas for using in GTK/GNOME applications as a replacement of GnomeCanvas. It will take all the good points of the current Dia canvas (internal handling of multiple views, abstract rendering interface meaning only a single draw() routine is needed which can handle GDK, Libart, gnome-print, etc backends) with the benefits of the GnomeCanvas (based on GtkObject system, arbitrary transformations, etc). Depending on how things turn out, we may submit it for inclusion in GTK+. More info when we work out what we are going to do.

There is something weird about bug buddy or the gnome bug system. I am getting unrelated bugs from different people posted as followups for bug #3500. All the followups have been posted with bug-buddy 0.6 or 0.7. It is getting quite annoying (and yes I have reported the problen: #8144)

Also got a message from someone wanting help installing libglade, because he can't install Unreal Tournament without it :)

The GNOME mailing lists are really slow. There is almost three days between sending a message and it getting posted to the list.

They have an ask slashdot story on slashdot at the moment that seems to be asking the question:

There is some GPL'd code I really like and want to use in my proprietary application. I know that the GPL says that I should GPL my software if I want to use the code, but do I really have to? How about if I put it in a DLL? *

The comments range from incorrect opinions and suggestions about how to get round the GPL to sensible messages from people such as Bruce Perens. Slashdot is definitely not what it once was (both that a story like this would even be on the front page, and the comments people posted).

Today is my 21st birthday. It looks like I am getting some Lego Mindstorms stuff, which should be interesting.

The custom shape code in dia now has a proper distance algorithm. This should get of the bugs where sometimes clicking on one object causes another one close by (which happens to be implemented with a custom shape) to be selected.

Had my 21st birthday party yesterday (the actual birthday is tomorrow though). It was pretty good, although we ended up with the same amount of swan gold as we started with (you can't even give that stuff away). Maybe this was also because we had some good beer as well.

Thanks to everyone who came, as I had a great time.

That is annoying. If your browser has the diary page cached, you will overwrite your last diary entry. The form on the diary page has the entry number as a hidden field.

I have been looking through what was installed when I did the upgrade from RH6.1 to 6.2. It looks like it installed a lot of kde stuff during the upgrade without asking (there wasn't any kde stuff before). A bit annoying, but not too bad. What was annoying was when I tried to install a new gnome-core RPM I had just built, and I get the following message:

file /usr/share/gnome/apps/.directory from install of gnome-core-1.1.8-20000413 conflicts with file from package kdebase-1.1.2-33

This seems fairly odd. Why would a file in /usr/share/gnome/apps be part of kdebase? So I decided to check what actually is in kdebase:

There was another GNOME steering committee meeting last night. This one was one hour earlier, which was unexpected but nice (the last one started at midnight). This probably had something to do with the daylight saving (something we don't have to worry about over in western australia, as it confuses the cows (yes that was one of the reasons people voted no in the referendum -- the cows would get confused if they were milked at a different time)).

Seven days till my birthday. I checked in the starts of the new plugins API for dia. This code reduces the number of entry points in a plugin to two. It should also allow me to write a nice plugin manager dialog where you can turn off loading of plugins if you want (this will be more important as we get more heavy plugins such as the python plugin).

Went to see Loon X-Wing (of Beaverloop fame) playing at the uni at lunch time yesterday.

Have been setting up my system with the new hard drive. I almost have it to a working state and a delivery man gives me RH6.2. I guess I will see what improvements have been made.

Noticed that a new version of gtkglarea was released in february. This means that the gtkglarea wrapper in pygtk should work with a released version of gtkglarea (there were a few API changes a while back, but no released version containing them).

Have been upgrading my system for the past few days. I got a nice new 20GB hard drive. When I was installed it, everything was working fine, but part way through the install of Red Hat the system would turn itself off.

While checking all the things I had changed, and trying bios upgrades, I got to a stage where the system would turn off as soon as you powered it up. I eventually worked out that the power supply was pretty much dead (not overloaded -- I don't have that much in my case), so I moved everything into one of the spare boxes and everything (including the new hard drive) works fine. The new case has three fans, and is a bit noisy. I should probably unplug one of them (they are all hooked up to the motherboard, so I wonder if you can tell the motherboard to turn them off :-)

Tracked down the segfault when subclassing GtkObject wrappers in the new pygtk. I also reenabled object attributes on the wrappers defined in C without problems, so this problem looks like it is fixed. The wrapper rescue code really is evil.

I don't know what all the fuss is about the gnome steering commitee thing. The main job we are doing is doing a bit of release coordination, so things get done on time (for october gnome, Elliot did most of this work, but it is probably less stressful to have a group of people doing the work). The other stuff discussed was about starting the process of creating a gnome foundation (which should take over a lot of what steercom was set up for).

Got the new ExtensionClass pygtk to a state where you can play with it. It actually compiles into something that actually imports, which is a step forward. I started work on a defs file for gtk, which has all the GTK widgets and has the constructors set for most of them. It also specifies public fields of objects, but that is not fully implemented yet.

I also got single wrapper per GtkObject working, and added support for an ExtensionClass instance dictionary. The new GtkObject wrappers are pretty much just as functional as the old ones, but without all the weirdness found before (eg. _o, multiple wrappers, two layers of wrappers, etc). When things get closer to working, I should do some benchmarks or something. The new stuff definitely loads quicker, and probably runs faster and uses less memory. I posted instructions on how to test things out on pygtk-list.

Got back from GUADEC yesterday at 4pm. 17 hour flights are way too long. Good thing they had a SNES at every seat. They also had a phone at every seat, and it looked like you might have been able to plug in a modem. If there was somewhere to plug in a laptop, Singapore Airlines would be almost perfect for hackers :).

Here are some notes on what I did in paris:

Put together some slides for the workshop talks at GUADEC. I have put them up on the web here for anyone who wants to read them.

I put out new versions of libglade, pygtk and gnome-python yesterday. The new libglade version contains bonobo support from Michael Meeks. The other two were mostly bug fixes.

Got a message from jwz on dia-list. He had a few interesting ideas for dia. I don't know how soon we will be able to implement them.

Did a little more work on libglade. Should probably do a new release soon. Should also release new pygtk and gnome-python before going to GUADEC.

Today was very humid rather than being very hot.

Thought I would see what the Advogato trust web looked like with graphviz. I would not recommend anyone else doing this, as it takes longer than I wanted to wait. Running it on just the purple lines finished in reasonable time, but ghostview was not usable with such a large single page PS file.

Someone suggested adding support for graphviz. As graphviz is not open source (despite what they say they believe in their licence faq), we can't link it with dia. If I get around to doing this, it would probably involve modifying graphviz to export the data in an easily readable format, and getting the import filter to read this data through a pipe.

More work on ExtensionClass based pygtk. I now have an almost finished code generator that will convert files in Havoc's new defs format to ExtensionClass based wrappers. There are a few problems, but it is mostly working. Handling of plain functions is partly done as well.

I should apply some of the fixes people have sent in to the HEAD branch of pygtk. I should also look at fixing the gettext.py bug for catalogs created on big endian machines.

Have started using Sawmill as my window manager. It feels quite nice and is very configurable. I am sure I could easily waste hours getting it to act exactly how I want it to :)

Have started working on the new ExtensionClass based pygtk code. It is on the extension-class-branch of pygtk in CVS. Currently there is not much to see other than a program to convert header files to Havoc's new defs file format and a program to merge changes from one defs file into another (needed because the header parser is not perfect and there is some info that is just not included in C headers). There is also a bit of code in pygtk/gtk/ (I am rearanging some of the modules to be nicer to the module namespace, and friendlier to case insensitive platforms such as win32). Code generator comes next.

Haven't posted anything here for a while. We released a new version of dia. Some of the features are not complete, but for most things, it is more stable than the last version. I also updated the web site this time :)

Still looking into using ExtensionClass in pygtk/gnome-python. I wasn't able to find any info on subclassing ExtensionClasses from C and didn't get any response from Jim Fulton, so I went ahead and added the feature myself. I posted the patch to Jim and the zope-dev list for anyone else who may want it. I hope it gets included, as I don't want to end up maintaining a fork of ExtensionClass. This looks like it will greatly reduce the amount of hand coded stuff found in pygtk, which will be good.

The river is still toxic. There were hardly any boats out on the river today. It is not surprising that there weren't many small boats (you wouldn't want to risk falling in), but there didn't seem to be many larger boats either.

Thought a bit about buying my own set of scuba gear. This will be easier than sharing the sets with the other rovers down at Pelican Point.

As for dia, I posted something about some of James Cape's suggestions for the UI of the toolbox to the list. Not many people liked the idea of having seperate windows for each sheet. I can't say I like the idea that much.

I finished integrating Lars's properties patch into dia (after modifying it so that it uses the offsets code). Now you can group a number of lines and set their dash pattern (before, you would have to ungroup the lines and go to each individual properties dialog.

There are a few missing features though. Undo on changes to a group does not work correctly. I will have to implement a "compound ObjectChange" for this to work correctly. I should also make it so you can modify the properties of multiple objects without having to group them (just selecting them).

Went for a dive on saturday, around 7pm. Since the river is toxic at the moment (an algae bloom), we went to the wreck just off north mole. It seems that all the jellyfish from the river had gone out to the sea where we were diving. There were a few stinger jellyfish out there as well. They were particularly annoying, as you couldn't really see them until they were about half a metre from you. I got to try out my new dive light -- it makes a big difference to what you can see.

I hate the heat at this time of year. I got up to around 40 degrees celcius yesterday.

The properties code in dia is working a bit. I switched the flowchart box object over to using properties, and it seems to be working fine. I also added properties code to the group object. You can select two flowchart boxes, group them and set both their properties at once. There is still a bit of work to be done w.r.t. undo with the group object -- I think I will have to implement a composite ObjectChange to handle that. It will also be necessary to add change notification to the group properties dialog so that we only set the properties the user changes (this might be a good thing in any case).