Posted by George Wright
Fri, 10 Aug 2007 00:20:00 GMT
On Wednesday Collabora moved out of the old office which we had outgrown and into our new amazing 4-room office opposite King’s College, Cambridge and above the Fudge Kitchen on King’s Parade. The old office is a shoebox in comparison! We now have a large front office, a large back office, a smaller back office (which I think we’re letting out to Ept Computing) and a large conference room. This ought to improve productivity, and as an added bonus it’s situated right next to my college supreme convenience!
To respond to some of the comments I’ve seen about Alp’s and my WebKit work; yes, I will be trying to get it to work on the Nokia 770 as I personally own a 770 and not an N800 and we may well be making a public release when it’s more stable. Currently I’m investigating the upstream source code and hoping to start working on adding the features we need to WebKit itself and submitting patches which implement them. Alp already has commit access to WebKit and will be working upstream anyway.
Last week, Tobias Hunger blogged about my involvement with standardising Mission Control, which is the abstraction layer which sits between Telepathy’s Connection Managers and the client. To anyone who’s interested in the work, the preliminary specification is online. So far, we still need to work out a sensible API for the Channel Handlers and iron out a few problems with the current specification, but we’re getting pretty close to coming to an agreement about it all. I personally would be very happy at the idea of having a standard API for this which can be shared between KDE (Decibel) and GNOME (Nokia Mission Control).
Posted in KDE, Computing, Cambridge | 2 comments
Posted by George Wright
Tue, 07 Aug 2007 16:59:00 GMT
For the past few days Alp Toker and I at Collabora have been working on integrating WebKit with the abstract browser user interface on the Nokia N800. Thanks to Nokia’s Engine Abstraction Layer (EAL) we were able to compile (with much difficulty) WebKit for the N800 and create our own EAL which provided hooks for the Nokia browser interface to communicate with WebKit.
Some of you may wonder why we are even bothering with this project. Well, for a start, WebKit is significantly smaller than Gecko and renders pages significantly better than Opera does (in my opinion). It also has a built in SVG renderer and has an active community surrounding its development. And, of course, it has its roots in the KDE project. Additionally, it’s run as a serious open source project and they do things the right way for open source, which is fantastic.
So far the port has been fairly successful. After a little setback we were soon on the ball and managed to get a proof of concept EAL working with WebKit rendering Google’s homepage.
By now we’ve managed to implement a few more hooks thus making more browser functions work, such as changing URL, going forwards/backwards in history etc, but it still needs a bit of ironing out - although the basic implementation is there and surprisingly usable.
There’s still quite a lot of work to be done, such as improving the rendering performance (it’s pretty good, but it still feels a little slow), adding input focus signals to WebKit GTK (to let us know when to bring the on screen keyboard out), and context menu entries (for right clicking). There’s also a slight problem with the rendering such that native widget theming isn’t great due to Hildon’s use of non-scalable widgets but we’ll be working on that.
At the moment we’re still developing this on our own but Alp and I are hoping to make a public release when the port has stabilised a bit.
Posted in Computing | 43 comments
Posted by George Wright
Thu, 24 May 2007 00:00:00 GMT
This summer I’ve managed to land an internship at Collabora Ltd to work on freedesktop.org projects.
Collabora’s current projects include Telepathy, Farsight and Nice. I will be working on something related to one of those projects.
So far we’re still in the brainstorming phase of what to work on, but Rob is keen that I work on getting TCP-like stream properties (flow control, ordering guarantees, retransmissions etc) into Telepathy’s “Tubes”, so that applications can communicate using stream-like Tubes over whatever UDP connection ICE has NAT-traversed. This will effectively provide a generic method for applications utilising either D-Bus or their own protocol (over a SOCK_STREAM) to communicate through NAT routers whilst still speaking something like TCP.
Of course, I would like to get the community’s opinions on the projects and most of all I’d like suggestions on other things I can spend my time hacking on which will benefit the populace at large. Any criticisms (constructive or otherwise) are welcome. The work will almost certainly be in glib but if there’s any relevant work that needs doing in KDE, Collabora may consider letting me work on that.
Posted in KDE, Computing, Cambridge | 6 comments
Posted by George Wright
Mon, 26 Feb 2007 00:22:00 GMT
The second day of FOSDEM was good fun. The night before was not so fun. After having dinner with a bunch of KDE developers, I decided to go and try to find the hotel I was staying in. After two and a half hours of running around Brussels like a headless chicken carrying a few tonnes of computers on my back, I finally managed to get to the hotel we were staying in - the Hotel Continental.
For those who don’t know, the Hotel Continental was the cheapest hotel I could find using Google that was in Brussels. It’s literally right next to the Bruxelles-Midi station and a twin room costs 61.50 EUR per night. The downside? The room is terrible. The wardrobe door fell apart when we tried to open it, the room was tiny and there wasn’t a shower anywhere near the room. However, it wasn’t all terrible. The staff were exceptionally friendly and courteous, the beds were comfortable and the breakfast was quite nice, which made the whole experience much more pleasant than it could have been. If you don’t mind a cheap place to stay that’s nicer than a youth hostel I think the Hotel Continental is quite a nice place to stay.
As for the conference itself, I finally managed to meet Lorn Potter today. I’ve known him for years due to my interest/work in Opie and Qtopia but as he’s located in Australia I’ve never had the opportunity to meet him, so it was great to finally meet him today. The rest of the time was spent trolling on the Debian booth and being viciously attacked by Nattie of #debian-uk fame, who has now surpassed Jes Hall and Lauri Watts on my top ten list of people to hate for touching my hair.
Anyway - it was a great conference and I hope to see many of you again next year!
Posted in KDE, Computing | 6 comments
Posted by George Wright
Thu, 22 Feb 2007 06:53:00 GMT
Well, it looks like my ThinkPad T43p has decided to die. A few months ago the UltraNav touchpad/trackpoint combination thing decided to start being faulty, in that it would randomly stop working and then only start working again when it felt like it (a reboot didn’t fix the problem, and no mouse was detected by the kernel if I did reboot), which was a pain.
So today I rebooted the laptop and what am I greeted with upon next boot? The following error message splurted out over and over again:
translated ATA stat/err 0x51/40 to SCSI SK/ASC/ASCQ 0x3/11/04
Really not my idea of fun. A hard disk failure - the most annoying thing in the world. Luckily I managed to get the hard disk into a USB caddy and copy the entire home partition off it, but it seems the hard disk really is faulty as dd if=/dev/sdi of=/dev/zero on my desktop (where sdi is the disk) gets the following sent to dmesg:
[19135224.412000] sd 23:0:0:0: SCSI error: return code = 0x8000002
[19135224.412000] sdi: Current: sense key: Medium Error
[19135224.412000] Additional sense: Unrecovered read error
[19135224.412000] end_request: I/O error, dev sdi, sector 61507808
[19135224.412000] Buffer I/O error on device sdi, logical block 7688476
[19135322.864000] sd 23:0:0:0: SCSI error: return code = 0x8000002
[19135322.864000] sdi: Current: sense key: Medium Error
[19135322.864000] Additional sense: Unrecovered read error
[19135322.864000] end_request: I/O error, dev sdi, sector 61507816
[19135322.864000] Buffer I/O error on device sdi, logical block 7688477
Oh the joy. I’ve now contacted Lenovo/IBM to see what they can do for me under warranty, and upon reading the warranty documentation it seems the trackpad and disk drive are customer replacement units so hopefully I won’t even have to send it in! Hooray for sane warranty agreements! :)
Posted in Computing | 11 comments
Posted by George Wright
Mon, 19 Feb 2007 13:22:00 GMT
Thanks to the response of my previous blog about the NX client I’m now working on for the 770, I’ve come to the conclusion that the best way to allow GTK to interact with the client library I wrote would be to implement a D-Bus interface in addition to the standard C++ interface so that the library can work much more cleanly with other toolkits (and hopefully even other languages!). Unfortunately, this is going to be on hold for at least a few days while I sort out my university work… It’s going to be quite a cool project though!
For the first time this year, I’ve finally decided to go to FOSDEM. In previous years it’s always been during term time and as I was at a boarding school I could never go, but now that I’m at university I’ve decided to just take the plunge and go (besides, Alasdair Kergon told me that I absolutely had to come this year). If anyone I know is going give me a shout and I’ll see if I can find you amongst the large crowd of (possibly drunken) geeks! I’ll be leaving Waterloo on the 6:06am Eurostar to arrive in Brussels somewhere around 9:30am and returning on a Eurostar departing sometime around 9pm on Sunday.
Posted in KDE, Computing | no comments
Posted by George Wright
Sun, 18 Feb 2007 04:54:00 GMT
So recently I managed to acquire a Nokia 770 internet tablet (thanks to mjg59). It’s a really nifty device – the screen on it is unparalleled (except, obviously, for the N800 – whose screen is slightly better), and it has quite a nice Linux platform on it. The problem? It runs Hildon, which is GTK-based. The input methods are heavily integrated into the GTK on the device and using any other widget toolkit just doesn’t work (well, it does, but you’re going to need a physical keyboard).
In light of this, I decided to start hacking around with GTK. Its API is not as pretty as Qt in my opinion, but it’s not too shoddy. It just takes a bit of getting used to, although the really verbose function names and endless requirement to call them really annoys me (gtk_foo_bar_baz_lala_eggs_breakfast() anyone?).
Armed with NXClientLib which I developed for my Summer of Code project last year, I decided to try and port my NX client to Hildon. In typical style, I made the architectural decision last year to write the library in C++ and link to QtCore/QtXml simply because it made my life a whole lot easier (I would dread to think how much extra code I’d need to write if I didn’t have QString). Anyway, I’m currently in the process of gluing GTK and Qt together in perfect harmony (haha) so that NX can take over the world, but it’s proving to be less trivial than I thought it would be.
Some gratuitous screenshots:
The initial login interface
The window automagically changes the layout for the on-screen keyboard
I’m afraid that’s all there is to “Maemo NX” thus far. I’ve spent most of the time trying to learn how GTK/X11 works and also thinking how best to implement the whole thing. If anyone is an expert at X window reparenting in GTK I’d be most grateful if they could take a look at this code and let me know why it doesn’t reparent the X window whose ID is specified by int wid into the GTK window I create and then make the reparented window visible. It seems to reparent fine, as the window with id wid disappears, but there’s no sign of it in the GTK window. If I can’t get this to work I’ll simply use mozplugger-helper which does a fine job of reparenting a window, but the fewer dependencies I can have the happier I’ll be.
If anyone’s interested in the code so far, it can be found in the FreeNX subversion repository at http://svn.berlios.de/viewcvs/freenx/nxclientlib/maemonx/.
Posted in KDE, Computing | 17 comments
Posted by George Wright
Sun, 31 Dec 2006 02:26:00 GMT
For those of you that don’t know, I have an HTC Universal phone. I have two main priorities for this thing - firstly, to get Linux booted on it with the phone hardware working nicely, and secondly to get NX working on it (yes, I’m strange).
Today was the first day in ages that I’ve actually decided to do anything. Those of you that met me in Ireland at aKademy will know that I was working on bootstrapping Debian Sarge on it without much success. Well, today, I managed to bootstrap an almost fully working Etch system without too much difficulty (and by that I mean it only took me 14 hours to do..).
Anyway, it was installed by running
debootstrap --include=wireless-tools,hotplug --arch arm --foreign etch /etch http://ftp.uk.debian.org/debian
at first on my desktop, then copying the entire directory over to my 2GB SD card. I had to create /dev/mmcblk0p1 and /dev/mmcblk0p2 though, and pass the kernel parameters rootdelay=3 root=/dev/mmcblk0p2 init=/bin/bash for it to be usable. When I got to bash I could then run debootstrap --second-stage which finished the bootstrap process.
After all that I installed kdrive from the 720degrees project but ended up having to replace the Xfbdev binary with that from one of the pre-built Universal GPE tarballs as it didn’t seem to have support for VGA modes.
In any case, I’ve now got X11 to run semi nicely, but the lack of a working touch screen is a bit infuriating at the moment.
I’m still trying to work out how tslib works - it seems to be a bit of a weird library. I can create the device /dev/input/event0 and get output from it when I tap on the screen, but Xfbdev always moves the cursor to the top right if I try and specify that as the tslib device. If anyone knows about how tslib works I’d be grateful if they could share it with me as documentation on it seems to be rather sparse.
I intend to get more working soon, but for now, I’m pretty washed out after all the fiddling I’ve done today. On the todo is to get the phone hardware to work, write a telepathy backend for the phone hardware and experiment with various desktop environments or window managers to find which one is suitable (I have a pet hatred for qtopia/opie/gpe. They’re all rubbish.). I’d also like to get the wireless working more reliably (I’m having to use the usbnet at the moment) and also write a keymap for the keyboard that isn’t rubbish. Ah the joys of embedded Linux.
Posted in Computing | 14 comments
Posted by George Wright
Wed, 22 Nov 2006 01:31:00 GMT
As some of you may know, I have a Windows Mobile 5 based phone. It’s possibly got the worst software I’ve ever used in my life but the hardware’s really nice so I put up with it.
One of the major things I use the phone for is as a modem for Linux because I have an (effectively) unlimited data tarriff on my contract, so being able to use it with my laptop is incredibly useful.
Previously, I have been using the USB cable and using wvdial to treat the phone as a dial up modem to connect to the internet. This wasn’t without complications as Windows Mobile seems to require an initialisation string which no other phone uses. This is:
AT+CGDCONT=1,"IP","APN",""
Where ‘APN’ is the access point name for your data carrier. In my case (T-Mobile), it’s “general.t-mobile.uk”.
To connect via USB though, you first need to load the appropriate module to give you a serial interface (/dev/ttyUSB0) to communicate through. You’d think this was as simple as running ”modprobe ipaq”, but no - it’s more complicated.
First you have to modprobe ipaq then run lsusb -v. You then need to look up your phone’s entry and record the hex numbers for the vendorID and the productID then rmmod ipaq and load it again passing those identification numbers as parameters to the module, so in my case:
modprobe ipaq vendor=0x0bb4 product=0x00cf
This will then give you /dev/ttyUSB0 to play with. However, it’s still isn’t that simple from here. You need one more fiddle to get it working. You need to set “Stupid Mode” in wvdial to force pppd to start as soon as the modem connects. Thus, the final wvdial.conf is as follows:
[Dialer Defaults]
Modem = /dev/ttyUSB0
Baud = 115200
Init = AT+CGDCONT=1,"IP","APN",""
Phone = *99#
Username = USERNAME
Password = PASSWORD
New PPPD = yes
Stupid Mode = yes
So, after setting up Linux you just need to run the “Wireless Modem” program on the phone and start the softmodem.
Unfortunately, trying to get it to connect via bluetooth appears to be infinitely more difficult. Luckily, I finally managed to get it to connect today after a few hours of research and fiddling, and it now seems rather trivial.
Firstly, Edgy’s bluez packages appear to be sufficiently broken that it doesn’t bring up the bluetooth adapter in discoverable mode - thus resulting in severe breakage when trying to connect the phone. This is fixed by running the command:
hciconfig hci0 piscan
You then need to fiddle some values in /etc/bluetooth/hcid.conf to allow the phone to pair.
I set a default passkey using the passkey "xyz"; option in hcid.conf as I couldn’t be bothered to set up a helper application. However, it should work just fine if you do set up a helper application. That’s basically all that needs to be altered.
Now onto the phone.
Going into Settings -> Connections -> Bluetooth, make sure the phone is discoverable and pair it with the computer. Once it’s paired, the Linux box’s ID should appear in the list on the “Devices” tab. You then just need to go into the “Internet Connection Sharing” program and start the connection using “Bluetooth PAN”. Once that’s done, it’s fairly trivial. On the computer, run the command:
pand --connect BADDR -n
Where “BADDR” is the 6-element bluetooth address of your phone, which can be obtained using the command hcitool scan. After that, it should splurt out something like:
pand[8361]: Bluetooth PAN daemon version 3.7
pand[8361]: Connecting to BADDR
pand[8361]: bnep0 connected
This means that pand has brought up a new ethernet interface called bnep0 for you to play with. After this, it’s just standard networking. Run a dhcp client on bnep0 and it will give you an IP address and NAT you across to the phone.
Hooray - the phone is connected via bluetooth to the internet. I still haven’t worked out how to get it to act as an actual modem over bluetooth (taking AT commands etc), but to be honest, I fail to see any point if we have Bluetooth PAN instead.
Posted in Computing, Misc | 33 comments
Posted by George Wright
Sat, 18 Nov 2006 01:27:00 GMT
On Thursday, Google’s London office hosted the first “Open Source Jam” - an invitation-only event designed to bring together members of the free software community to hack on stuff at Google’s office.
I was the only person in KDE there as John Tapsell, in his infinite wisdom, forgot it was Thursday and didn’t show up. It’s a really nice office in Belgrave House a few minutes walk from Victoria underground station. They were very hospitable and there was free wireless internet for us - although it was very erratic and the lack of a DNS server didn’t help. To be fair I didn’t really use the wireless as it was such a short event that I ended up talking through the whole evening instead of geeking.
Among the people there were a couple of guys I already knew, namely Alasdair Kergon (agk) of LVM fame, and also Paul Nasrat, a Red Hat employee.
The employees themselves were really friendly - and ridiculously good at table football. The room we were in was some sort of a hybrid cafeteria and recreation room, sporting lots of drinks refrigerators, a pool table and a table football table.
All in all, it was quite a fun event and they’re thinking of making it a regular thing on a monthly basis. Unfortunately Cambridge is a bit far to come from for a three hour session but I’ll certainly endeavour to go a few more times.
Posted in Computing | 2 comments
