plaes.org | Be free!

While browsing the Wikipedia on Sunday evening I managed to stumble upon a neurological disorder called Attention-Deficit/Hyperactivity Disorder or ADHS for short. Yeah, I actually read this kind of stuff because it interests me Anyway, upon reading this page, it seems to be the one that is bugging me, because 18 of these 20 symptoms there seem to be present (well, at least that’s how I feel)…

On Saturday, I was forced to hack Beagle a bit and mainly because the revbumped version in portage didn’t compile for me. It was quite weird thing because almost nothing had changed inbetween the old version and revbumped one, so it took a bit time to figure out what was wrong. But I managed to fix it. Nothing still beats open source software and even Sun’s people have agreed to that.
Monday brought few surprises - first I got my dog-tags:

And secondly - who could have thought that visiting the Movial office for Monday meeting would end up having a box containing OMAP development board. Getting it up wasn’t that easy. First, I was missing power, because the dev-board’s adapter had american pinout. So I spent ~2 hours searching from different hardware stores in Helsinki center, until I gave up and bought some cable with plug and made the american->finnish converter myself. Time for first boot:

Beautiful, isn’t it?

It was, but I spent ~2 hours trying to figure out how to get some output on the serial console… I tried changing various settings… and in the end I never got the output.. And on next day, I got the missing cable from Markku and Tuomas and I was finally able to see the kernel messages

So far I have update the scratchbox ebuilds for Gentoo, and I’m currently trying to figure out how to create the Gentoo toolchain and devkit for it. Anyway, Gentoo ebuild stuff can be downloaded from here: scratchbox-1.0.6.ebuild and files/scratchbox.rc.

<DISCLAIMER>
Gentoo is not broken by design. I just break it in purpose to learn about the system and of course - in order to fix things, you (or someone) have to break it first
</DISCLAIMER>
Yesterday I decided to have some fun and decided to unmask D-Bus 1.0 and well.. upgrade.
Upgrade was quite easy, but revdep-rebuild reported that I had ~80 b0rked packages including most of the GNOME and of course the best one being Openoffice.org
But before I started recompiling all these I decided to check how D-Bus-1.0 proof the portage is. As it was masked package and involves quite a lot of changes - previously D-Bus package contained most of its bindings and the core library in one package, but now the bindings have been separated into their respective packages. As expected most of the packages which depended on dbus-glib were not updated yet, so I started filing bugs. Fortunately I didn’t piss anyone off while doing this and developers seemed to be quite happy to see some initiative and in the end I had reported 15 bugs for packages which didn’t have dbus-glib dependency.
When I left home for work, I started emerging Openoffice and decided to take a look how things were going… So I ssh-d to my machine, but before I did look into emerge status, I decided to see whether Gaim (the text-mode version) works with new dbus. It didn’t so I decided to investigate it with strace:

open("/var/lib/dbus/machine-id", O_RDONLY) = -1 ENOENT (No such file or directory)
fstat64(2, {st_mode=S_IFREG|0644, st_size=0, ...}) = 0
mmap2(NULL, 131072, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7a42000
write(2, "process 16846: D-Bus library app"..., 221) = 221
rt_sigprocmask(SIG_UNBLOCK, [ABRT], NULL,  = 0
write(2, "  D-Bus not built with -rdynamic"..., 64) = 64
tgkill(16846, 16846, SIGABRT)           = 0
--- SIGABRT (Aborted) @ 0 (0) ---
+++ killed by SIGABRT +++
Process 16846 detached

Hm.. So /var/lib/dbus/machine-id doesn’t exist… Now what? Then an idea struck - maybe I should restart dbus service! (I assume this is so far the stupidest idea in this month;)).

amd@sl007 /var/lib/dbus $ sudo /etc/init.d/dbus restart
 * Caching service dependencies ...                                                        [ ok ]
 * Stopping avahi-dnsconfd ...                                                             [ ok ]
 * Stopping avahi-daemon ...                                                               [ ok ]
 * Stopping NetworkManager ...
Read from remote host xxx.xxx.xx: No route to host
Connection to xxx.xxx.xx closed.

Yay! I had just locked myself out from my box… and of course, why does NetworkManager bring down the interface when it has been stopped.

Well, Hans wasn’t certainly the first one to run whole GNOME session under Valgrind, but he was certainly the first one who created a set of scripts which make this kind of testing more fun. I have previously tried to valgrind the whole desktop myself and it usually involved lots of bizarre hacks + starting the X with .xinit.
I created an ebuild for Gentoo, but unfortunately had to repackage Hans’s package because Gentoo doesn’t handle rpm files by default. Also SUSE’s gnome-session binary seems to be renamed to gnome, so I had to add a workaround for this too…
As Hans mentions, at least one gigabyte of RAM is required. Evolution did work and whole desktop was quite usable (albeit a bit slow), but neither Epiphany or Firefox started, and some panel applets didn’t load either.

Ah.. almost forgot.. To get usable output for debugging purposes you should have at least following specified in /etc/make.conf

FEATURES="splitdebug"
CFLAGS="-g -ggdb"

That’s at least what I have.

Today there was another user in #gentoo-desktop who managed to run into circular dependencies problem. While trying to help to fix this issue, it was suggested that someone should write a graphviz translator to visualize these kinds of situations. So I took the task myself and well the result was quite hilarious. And its only 12 lines of Ruby code (mainly because all the hard work is done by Graphviz itself)
[Update]
Well, it took some thinking, but the minimised graph is here:

Check the code too.

I just found out that plaes AT openoffice org can be used for contacting me
All because of the “hard” work for getting OpenOffice.org to build with gcc-4.2.x compiler. Should be possible since 2.0.4 got out. Will try tonight…
[Update]
Gah.. It didn’t build.. Some stupid problems with STLPort (which always has problems)