tkh’s blog

mlterm bug

Now it’s fixed in the upstream cvs tree. I’m happy to have the scrollbar behaving correctly. However, the debian maintainer has not responded to my report. Looks like he has been inactive for quite a some time. Isn’t a way to push him upload the fixed package?

mlterm

I’ve been using a terminal emulator called mlterm for a while. It’s a great terminal emulator supports anti-aliased truetype fonts, transparent background, and Unicode. However, I’ve been annoyed by an odd behavior of its scrollbar. When you drag the scrollbar, it warps rather than following the mouse cursor. It’s hard to express the exact behavior, but it’s really annoying. So I took time to debug mlterm and fixed it. Thanks to cscope’s nice cross-referencing, the process to locate the code fragment was really easy. I filed bug reports both to the upstream and to debian.

Here is the patch:

--- xwindow/x_scrollbar.c.orig 2005-02-10 08:15:07.000000000 -0700
+++ xwindow/x_scrollbar.c 2005-05-22 06:00:30.597051171 -0600
@@ -538,6 +538,11 @@

sb->is_pressing_down_button = 0 ;
}
+
+ if( sb->is_motion)
+ {
+ sb->is_motion = 0 ;
+ }
}


Star Wars

I’ve heard from people that the Episode 3 is much better than the previous two new installments. I still haven’t seen it, but I’ll see it sometime in next week.

Dual Core CPUs

Just read the TechReport’s review on the new dual core AMD Athlon 64 x2. It totally beats Intel’s Pentium chips in most benchmarks and the power consumption is very low. I’m convinced that I can switch from SMP to a dual core CPU without any concerns. Only downside is their prices. As described in the review, The highend chip is root-canal-without-anesthetic expensive!

Framebuffer compositing

Since people are having luck with installing the Ubuntu Xorg packages on top of debian sid, I tried it. After adding an apt-line for the Ubuntu archive, upgrading the XFree86 packages to Xorg ones went nicely. You need to install libgcc1 in the debian experimental, but I had already installed it. Then I restarted the X server and got the X screen, but there was no shadow around the border of windows. I googled and figured out that I need to run xcompmgr for the framebuffer composition (shadows and translucent windows). I searched both with aptitude and apt-search for xcompmgr, but there was no such package. As the last resort, I logined to ftp://archive.ubuntu.com and found that there are “universe” and “multiverse” archives in addition to the “main” archive. After adding them to the apt-line, I was able to install the xcompmgr package. First, I ran xcompmgr -c, and got shadows! Finally I saw the evolution in X11 since the introduction of antialiased fonts. Then I ran glxgears to make sure OpenGL apps run fine, but it just dumped core. After setting “AllowGLXWithComposite” to true in the Device section in xorg.conf, glxgears ran OK, but glxinfo killed the X server everytime I ran it. Bummer. I like shadows and transparent windows, but I can’t accept the unstable behavior of OpenGL apps. So, I ended up removing Xorg packages by hand with “dpkg -P –force-all” and downgraded to XFree86 packages.

The Time Traveler Convention

It will be held tomorrow at MIT. In short, “Technically, you would only need one time traveler convention.” It must be a really fun event. Here is the announcement:

The Time Traveler Convention

May 7, 2005, 10:00pm EDT (08 May 2005 02:00:00 UTC)
(events start at 8:00pm)

East Campus Courtyard, MIT
3 Ames St. Cambridge, MA 02142

42:21:36.025N, 71:05:16.332W
(42.360007,-071.087870 in decimal degrees)

Convention events now start at 8pm for present-time attendees. It will start with lectures and/or music, and at 10pm the moment of truth will arrive. No need to modify any existing publicity efforts, as time travelers can come at 8pm or 10pm.

We need you to help PUBLICIZE the event so that future time travelers will know about the convention and attend. This web page is insufficient; in less than a year it will be taken down when I graduate, and futhermore, the World Wide Web is unlikely to remain in its present form permanently. We need volunteers to publish the details of the convention in enduring forms, so that the time travelers of future millennia will be aware of the convention. This convention can never be forgotten! We need publicity in MAJOR outlets, not just Internet news. Think New York Times, Washington Post, books, that sort of thing. If you have any strings, please pull them.

Write the details down on a piece of acid-free paper, and slip them into obscure books in academic libraries! Carve them into a clay tablet! If you write for a newspaper, insert a few details about the convention! Tell your friends, so that word of the convention will be preserved in our oral history! A note: Time travel is a hard problem, and it may not be invented until long after MIT has faded into oblivion. Thus, we ask that you include the latitude/longitude information when you publicize the convention.

You can also make an absolute commitment to publicize the convention afterwards. In that case, bring a time capsule or whatever it may be to the party, and then bury it afterwards.

For further details, please check

http://web.mit.edu/adorai/timetraveler/

Chinese propaganda

Since I started contributing to Wikipedia, I’ve seen that so many Chinese cultured people bent and exaggerated historical events. Most of them lack an idea of approaching to the facts by citing authoritative sources. What I can do, as a Japanese citizen, is to fix them one by one. It’s tiring, but someone needs to do it, otherwise they will be established as facts.

Subversion

I migrated my CVS repository to a Subversion repository using cvs2svn. I had to ran cvs2svn a couple of times until I fully figured out what information I wanted to migrate. After successfully imported my source trees into Subversion, I tried to check out source trees. Local access using the file:// protocol URL worked well. Then I tried the file:// protocol from a remote machine via NFS, but svn command got frozen. Since the svn+ssh:// protocol worked well, I suspected that there is something to do with NFS. I read the online Subversion book. Bingo! The Berkeley DB specifically requires the filesystem to provide the POSIX file semantics (locks), and NFS doesn’t provide it. That’s why the file:// protocol via NFS resulted in deadlock. So, I learned that I have to use the svn+ssh:// protocol to access the repository, but I thought it was too much to use an SSH tunnel for LAN access. I read the Subversion book again and found that another backend DB called FSFS is much faster, efficient, and provides NFS access. Thanks to the Subversion’s nice backup mechanism, I was able to migrate the backend DB from Berkeley DB to FSFS with only a couple of commands. Here is the commands I used:

# svnadmin dump /home/svn > svnrepo.dump
# rm -rf /home/svn
# svnadmin create --fs-type FSFS /home/svn
# svnadmin load /home/svn < svnrepo.dump

Now everything works fine including NFS access.