Working at Red Hat

About two years ago, I was at a conference with a load of GNOME people. I mentioned over drinks to two friendly Red Hat hackers, that I had an idea about a packaging framework. It was just an idea, as I was working for a large defence company, and had precious little time to maintain gnome-power-manager, let alone start anything new.

Nearly a year ago, I was hired by Red Hat to work on power management stuff. Now, my boss is one of those cool bosses that gives you quite a bit of ‘space’ and with his blessing I started to hack on PackageKit. 18 months later PackageKit is feature-complete, and the defacto standard across a dozen or so distributions.

Every week or so, I put up a new screenshot on this blog of cool stuff I’m working on. Every week people critique my ideas, and I go away to fix them up so the next version is that little bit better. Every week a few people say thanks, and tell me I’m doing some cool stuff, which is nice.

I don’t want people to think that PK or DK-p are Red Hat projects, but the simple economics is that they pay me to hack on cool projects. Whilst working at Red Hat I’m working alongside the very best people in the industry, in an environment that rewards innovation and thinking a little bit different.

What I’m trying to say is, most things you see on this blog are possible because of Red Hat. Sometimes I feel Red Hat doesn’t get the credit it deserves. Red Hat Rocks.

[CVE-2008-4311] DBus 1.2.6

Please don’t yet update DBus for CVE-2008-4311. It’s known to break PackageKit, cups, ConsoleKit, DeviceKit, DeviceKit-power, gdm, and system-config-services. There’s a partial bugfix that has been pushed so PackageKit tools still run (without the GetTid or SetLocale errors), but as introspection is still broken they’ll be odd little warnings and errors for other stuff.

I do understand how important this update is, but given this wasn’t a root login vulnerability, or anything crazy like that, I’m surprised it didn’t sit in updates-testing for a few days to fix up all the other system daemons. The worst bit is that it’s broken automatic updates for thousands of people.

I’m planning to spend this morning closing duplicate bugzillas. Fun.

Command not found

I’ve been quietly admiring the Debian BASH command not found functionality for a while. For those unfamiliar with the functionality, it suggests installing packages if you typed a command that is not installed. It also seemed to be pretty helpful suggesting replacements for typos.

When Rahul opened a bug to include this in Fedora, I figured this is just the sort of thing a framework like PackageKit needs to hook into. Roman Rakus put a simple patch in Fedora for bash, and now we’ll try to get it upstream. Then it will work on all distros that support PackageKit.

I spent this afternoon hacking a simple implementation. It’s up to the distro whether it should be installed by default on a base install. It’s also pretty configurable, so if you just want it to launch lshal everytime you type lshla, you can have that. If you want it to ask you before it installs shed to provide /usr/bin/shed, then that’s possible too. The simple configuration file is available here for review.

There’s also a screencast available if you want to see all the hotness.

PackageKit and Pango are now friends

PackageKit in git master has a new trick up its sleeve. Thanks to Behdad, It now optionally installs a gtk-module that integrates with Pango to detect when a document needs a new font installed to render correctly. No code changes are needed to any application using Pango, as all the magic is done in the gtk-module. When used with a session interface like KPackageKit or gnome-packagekit, the following UI is shown:

Font installer
Font installer

I really want a way of being able to say “don’t ask about this language again” if not found so to not piss off the user. I also know my grammar is pretty shitty. All comments and recommendations gratefully received. Thanks.