I earlier wrote about I was experimenting with PowerShell, so I had to install Windows in a virtual machine. It's not a big problem, we at the uni get a license to use it for free till the end of our studies (or something like that), so I started to search for what emulator should I use.

Of course my favorite is qemu, but till I don't have kvm-enabled hw and kvm support isn't fully merged from kvm, it's slow.

Given that I had to use some other emulator, I thought I gave vmware a try, they offer a 30-day trial and I wanted to see what's new in their recent versions - I think I last tried it about 2 years ago. It seems the automagic gui screwed up something and there is no more vmware-config.pl, so I gave it up.

Last, I tried Virtualbox. Man, it's fast! Of course it's bloated as well (external kernel module, guest addons, etc) - but if I forget about these issues, it really gaves me the same experience vmware gave me when I last tried it. Impressive.

As a side note, git-1.6.2 is out, sadly I haven't got much time for it - only 5 patches of mine are in the changelog. But at least now the pdf version of the user manual is supported out of the box.

Now that I know a little bit more about Erlang I checked how to configure vhosts with ejabberd. It turns out it's quite trivial once you understand the syntax of the config file, which is just eval'd, so it has this horrible (for an outsider) syntax. ;)

Actually configuring vhosts are just about adding the new host to the DNS and adding it to /etc/ejabberd/ejabberd.cfg, like:

{hosts, ["mydomain.org", "another.com"]}.

And that's all you need. Simple, isn't it? :)

So we had this bug for a while, and the solution was just using an other dir, so the user later can merge the contents of the two dirs.

Sigh, testing the setup still takes a lot of time. For example to test the above bug, you need to do a base install, which still takes a lot of time. Hopefully sooner or later we can switch to kvm (as the old kvm-incompatible hw gets unused) and then this can be faster.

I just found a nice sentence in a commit-digest, released after KDE 4.2 was out:

Work on porting KControl and Konversation to KDE 4.

Wow. Let's say I just use kdebase from KDE 3. Now, kcontrol is part of kdebase. So it's nice we already have kmail and other fancy stuff, but we still doesn't have kcontrol? Brr...

(Yes, there is a system settings, and that won't let you configure your printer, for example.)

etckeeper recently started to depend on tempname which is like mktemp, but we don't have it in Frugalware.

I recently sent this patch to Joey (the etckeeper maintainer) to add support for mktemp.

The full name is Oracle Enterprise Manager Console. You can download the 10g version from here. You can find a user/pass on bugmenot. After unzipping, just running '/path/toclient/runInstaller -ignoreSysPrereqs' (I only tried with an absolute path) should be enough. It can be installed a user, later it'll prepare some scripts to run as root.

I always replaced $HOME/orafoo with $HOME/.orafoo, when the installer asked. If you get 'no such file or directory' errors, you can just edit the relevant script and fix the path - I needed to fix /bin/rpm, /bin/awk and /bin/sed. (All of them are under /usr/bin.) A hack for the lazy ones: just symlink these to /bin, then no modification of the scripts is needed.

To your ~/.bash_login:

export LIBXCB_ALLOW_SLOPPY_LOCK=true
export ORACLE_HOME=~/.oracle/product/10.2.0/client_1/

There was a recent interview with the developers of UHU-Linux, which was the de-facto "Hungarian Linux" a few years ago here at Hungary.

The guys had a pretty closed approach and finally it seems they make money from products based on their distro - releasing something useful for the users for free is definitely not the number one task on their todo.

There were a couple of comments to the article, a few of them was about Frugalware, given that we are still the only distro where the origin is Hunagary and it's actively used outside Hungary as well.

In fact I just did a quick search, listing the last commit dates of each devel. Two of the Hungarian ones are pretty inactive. Both of them do other administrative tasks. (Buildserver / bugtracker hosting, etc.) But the rest is quite active, which was a surprise. I had the impression that lately we got some new French devels and once they got their git accounts, they got a bit less-motivated but this is simply not true. All of them (and this is true for the other non-Hungarian devels as well) was more or less active in the last month.

I think this is pretty straight-forward, even if the ohloh stat says the commit number (compared to last year) is sightly decreasing. This is a typical example: the commit number is not everything. I think this time last year we might be a bit more active overally, but that was because several devels (including me) spent a horrible amount of their free times on Frugalware, which is not something you can do for several years. Today we have a much stable devel base, increasing package maintainer community and yes, I'm happy with our current status. :)

I already wrote once about Pootle forcing us to use UTF-8 and not respecting the po headers (where charset can be set), but there is a new chapter in this story. So in case the charset is wrong, the generated mo file is still fine, as long as you have utf8 support in your system. Of course in the setup, where we have an "as much as possible" stripped down system, this is not true.

A contributor reported this recently, and the symptom was that "sometimes" netconfig is localized, sometimes not. It turned out a bit later that it's localized when it's running in the target system's chroot (when the install of the packages is finished and the config tools are invoked), but not before the installation. Of course the previous solution worked here pretty fine as well: I just had to convert the remaining translations to utf8 manually, and then add code to convert from utf8 to the native encoding in the pre-package stage (currently named 'make prepare') of frugalwareutils.

Hand converting is boring. And sometimes it's completely messed up, for example you need an 'iconv -f utf8 -t latin1' twice to get the correct accents. ;-)

So I got the followings today at the uni on a Windows box:

    ping alaplab.
    Pinging alaplab.mit.bme.hu [152.66.252.18] with 32 bytes of data:
    Reply from 152.66.252.18: bytes=32 time<1ms TTL=127
    Reply from 152.66.252.18: bytes=32 time<1ms TTL=127
    Reply from 152.66.252.18: bytes=32 time<1ms TTL=127
    Reply from 152.66.252.18: bytes=32 time<1ms TTL=127
    Ping statistics for 152.66.252.18:
        Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
    Approximate round trip times in milli-seconds:
        Minimum = 0ms, Maximum = 0ms, Average = 0ms

Now I wonder if they just ignore the dot at the end of the hostname or WTF.

We had this old #3331 bug, and it turns out that Fedora had an s390 patch that fixed it. Of course the patch name is misleading since it fixed something on x86 as well, so it was the last thing I thought to try... :-)

I got a reply reply to my libbtctl patch I sent back in last year:

On Sat, 2008-12-27 at 18:32 +0100, Miklos Vajna wrote: > Hi, > > I recently tried to build libbtctl with mono-2.0 and the build fails > because the .pub signatures are no longer supported. > > A similar commit in evolution-sharp's svn: > > http://svn.gnome.org/viewvc/evolution-sharp?view=revision&revision=191 > > Here is a patch to do the same for libbtctl: > > http://ftp.frugalware.org/pub/frugalware/frugalware-current/source/gnome-extra/libbtctl/mono2.patch > > Of course you can generate your own new key with 'sn -k 1024 > libbtctl.snk' if you don't want to take the one I use right now.

Because the patch wasn't in bugzilla, I missed it for the last ever release of libbtctl.

Use bugzilla in the future.

Cheers

Needless to say, there is no libbtctl component in the GNOME bugzilla.

Yay for bureaucracy!