Hibernate was working out of the box (well the going to sleep part, at least), but resuming took almost as long as a cold boot. Another thing was that my wireless was broken after a resume. On my happiness scale (range: 0-10) this scored a 3.

To fix the wireless I thought I could just unload the ath9k module before suspending and loading it on resume. For this I added a modules script in the following directory /etc/pm/sleep.d:

#!/bin/sh
# Action script to remove/add modules

PATH=/sbin:/usr/sbin:/bin:/usr/bin

# pm-action(8) - <action> <suspend method>
#
case "${1}" in
    suspend|hibernate)
        modprobe -r ath9k
        ;;
    resume|thaw)
        modprobe ath9k
        ;;
esac

After this my wireless was working like a charm. On my happiness scale (range: 0-10) this scored a 5.

So next up: fix suspend. What was the problem? I could suspend my laptop, but it came right back to live after a few seconds of suspending. Not very useful.

After some googling around it hit me, USB wake ups! But how to do silence the wake ups from your USB devices? The following works for me:

# disable wakeup events when suspending from USB
echo disabled > /sys/bus/usb/devices/usb1/power/wakeup
echo disabled > /sys/bus/usb/devices/usb2/power/wakeup
echo disabled > /sys/bus/usb/devices/usb3/power/wakeup
echo disabled > /sys/bus/usb/devices/usb4/power/wakeup

Now both hibernate and suspend are working (happniess = 8).

This is a bugfix release, with the following ChangeLog:

2 Mar 2010: 1.1.1 Miek Gieben rdup@miek.nl

• tests: Skip gpg test when gpg is not found (Tom)
• rdup: Rename strmode() to rdup_strmode for FreeBSD (Charlie Kester)
• rdup: Dont descend when toplevel directory is a symlink (same behavoir as Gnu tar) (Reported by Charlie Kester)
• rdup: reap our childeren with waitpid() (Oliver Dain)
• build: Building fix on rpm distros (SuSE/Fedora) (Sven Hartrumpf)
• build: add --enable-debug to built debug build

Quick download and project page

This means I'm going to be offline for some time.

Here we go.

The following lists sums up my needs and troubles:

• BTRFS for my main partitions (only / and /boot will be ext4);
• install Ubuntu Lucid (Alpha3);
• Lucid's installer does not support BTRFS;
• there is no CDROM drive in sight.

The attack plan:

• use unetbootin to create bootable (usb) hardisk image;
• format the external disk as vfat! (not ext3 will not work when booting);
• install minimal Ubuntu image on RAID1 ext4 partitions;
• leave 2 disks with a large partition for BTRFS;
• install Ubuntu 10.04;
• reboot;
• configure BTRFS;
• move /var, /usr, n' stuff over to BTRFS.

0th problem: system will not boot from external disk

First I needed to make my system boot from the external USB hardisk, after fiddling in the BIOS, this was fixed.

Back to installing!

1st problem: external hardisk was not recognized as a CDROM

Fixed by adding: cdrom-detect/try-usb=true to the kernel boot parameters as told here

If you get "Incorrect CD-ROM detected" error on detection stage, reboot, press F6 and then ESC to go to manual boot line editing, and add the option 'cdrom-detect/try-usb=true'. On Ubuntu 9.10 server edition the install menu will be shown right after reboot. Chose "Help" and then press F6. At the boot prompt type "install cdrom-detect/try-usb=true" and hit enter.

Back to installing!

2nd problem: installer files missing

The installer could not copy some files. Luckily you can switch terminals in the installer with Alt+F2 which drops you in a busybox shell. Go to /var/log/ and check the log files. In there I found that it was borking on, some deb called:

linux/fs-secondary-modules-2.6.32-14-generic....udeb

Which wasn't there. I did have:

linux/fs-secondary-modules-2.6.32-14-generic....ude

(No closing 'b'!!). Lets fix that by linking the correct name:

# ln -s .....ude ....udeb
operation not permitted

Of course, it is read-only mounted and a VFAT partition which does not support symbolic links. Plan B (copy it):

#mount -o rw,remount /cdrom
cp fs-secondary-modules-2.6.32-14-g.....ude \
fs-secondary-modules-2.6.32-14-g....udeb

Okay, that worked.

Back to installing!

3rd problem: /dev/md1 is used as swap

This system already had a RAID1 configuration on it, so I had to get rid of that first. Also see this link.

I could not delete that raid partition because it is in use as swap by the installer. WTF? Again the Alt+F2:

swapoff /dev/md1        # md1 was the culprit

I also tend to find fdisk much easier to partition the disks so I also did this from the command prompt - then let the installer rescan the newly partitioned disks. I went for the following layout (for 2 disks):

/dev/sda1       /dev/sdb1   combined to form md0 (boot)
/dev/sda2       /dev/sdb2   combined to form md1 (root)
/dev/sda3       /dev/sdb3   combined to form md2 (swap)
/dev/sda4       /dev/sdb4   not used, for BTRFS

Back to installing!

4th problem: grub did not install

Grub failed to install on /dev/sda and /dev/sdb. I tried grub-legacy and grub2 both failed.

When the installer asks "Install in master boot record", say "no" and specify only /dev/sda. It performs a grub-install /dev/sda which fails for some reason. Time to try it ourself. First where is grub-install located, turns out: /target/usr/sbin/grub-install

Now some more foo, drop to the shell again, and chroot to your new install

# chroot /target

This is with grub2:

grub-probe: error: no mapping exists for `md0`

See this. Next I tried:

grub-install --modules=raid /dev/sda
you are attempting a cross-disk install, but the filesystem
containing /boot/grub does not support UUIDs

Hmm, see here for a bug report.

Still in the chroot I did the following:

mount /proc
grub-install --modules=raid /dev/sda
Installation finished. No error reported.

Okay, that worked. No go back to the installer and choose "Continue without a bootloader", this will finish up installation.

Back to installing!

5th problem: does not boot

GRUB Loading stage1.5.

GRUB Loading, please wait...
Error 2

Hmm, back to the BIOS and enable my disks again - they were disabled so that I could boot from the external USB hard disk. Reboot and... I'm dropped in a grub2 shell. Back to reading the docs.

Next:

grub> ls
(hd0) (hd0,4) .... (md0) (md1) (md2)    # yes! md devices
grub> insmod linux              # needed?
grub> insmod raid
grub> set root=(md1)
grub> linux (md0)/vmlinuz-2.6.32-14-generic root=/dev/md1
grub> initrd (md0)/initrd.img-2.6.32-14-generic
grub> boot

The kernel names came from this announcement. So I knew it had to be something like vmlinuz-$version-generic.

6th problem: failed to mount root file system

Somehow I managed to miss type the above grub commands.

ITS ALIVE BOOTS!

I still need to figure out how to make this permanent. Looks like you only have to give update-grub as root when you are logged in. This creates a /boot/grub/grub.cfg to will boot your system. (Turns out that config file was missing).

The BTRFS part comes in a separate blog - this is already too long.

But for really nice looking output I wanted the following:

• Fonts in images/graphics should match the font of the main document
• TrueType fonts, instead of the standard TeX fonts. To avoid the "Heeh, that looks like a LaTeX document"-reactions;
• UTF-8 support and international character support;
• Easy integration of images (SVG based);
• Automatic building.

I'm using the following tools for this:

• Inkscape and pdfcrop for graphics generation;
• make for building;
• VI for editing (with syntax highlighting);
• evince as PDF viewer;
• xelatex for building the pdf.

General setup

As said earlier, I'm a fan of the memoir class as a replacement of the standard LaTeX classes. So I'm using it here again. When you are using xelatex (better UTF-8 support) you should also include the package xltxtra and xunicode.

Fonts

The package for TrueType font support is fontspec, to keep the math fonts working I had to include the package as \usepackage[cm-default]{fontspec}.

Note: fontspec mandates xelatex.

To use special characters I also found the pifont package. In my current working document I use the Liberation Fonts and Courier New:

\defaultfontfeatures{Scale=MatchLowercase}
\setmainfont[SmallCapsFont={* Caps}]{Liberation Serif}
\setsansfont{Liberation Sans}
\setmonofont[SmallCapsFont={Courier New}]{Courier New}
\fontsize{10}{10}

LaTeX will make TrueType fonts slightly larger than its core fonts, the Scale=MatchLowercase tries to remedy that.

Graphics

I'm using Inkspace to draw my graphics, an added bonus for using TrueType fonts in my LaTeX documents is that I can use the same fonts in the Inkspace drawings. So any text I add will be of the font Liberation Serif at 10 points.

Further more Inkscape has the nice feature that it can be used on the command line

$ inkscape --export-pdf=drawing.pdf drawing.svg

generates the pdf file. To make it fit in the document we need to crop the resulting pdf

$ pdfcrop drawing.pdf output.pdf && mv output.pdf drawing.pdf

And now we are left with a nice, cropped, pdf. See this link on how to include graphics in your LaTeX document, I just use:

\begin{figure}[h!b]
\includegraphics{drawing.pdf}
\caption{Fake caption line.}
\end{figure

Which will probably be extended somewhat to allow for cross references (search for \label and \ref).

Building

To help in the sometimes tedious building of LaTeX documents, I use make. For instance creating the pdf graphic files from the svg source files is done with the following Makefile:

pdfs := $(patsubst %.svg,%.pdf,$(wildcard *.svg))

%.pdf: %.svg
    inkscape --export-pdf=$@ $<
    pdfcrop $@ output.pdf && mv output.pdf $@

all: ${pdfs}

clean:  
    rm -f *.pdf

The only thing I need to do is to create a file with Inkscape, save it and run make.

For the document itself I'm using a similar, small Makefile.

.PHONY: fig

all:    fig go.pdf

go.pdf: go.tex go-setup.tex chapter*.tex ex*.tex src/*.go 
    xelatex go.tex && xelatex go.tex

fig:    fig/*.svg
    ( cd fig; make all )

clean:
    rm -f go.lol go.aux go.log map.log go.pdf
    ( cd fig; make clean )

Putting it all together

Next to the above Makefiles I have the following in the preamble:

\documentclass[a4paper,openany]{memoir}
\usepackage[cm-default]{fontspec}% provides font selecting commands
\usepackage{xunicode}% provides unicode character macros
\usepackage{xltxtra} % provides some fixes/extras
\usepackage[answerdelayed,lastexercise]{exercise}
\usepackage{pifont}
\usepackage{caption}
\usepackage{alltt}
\usepackage{url}
\usepackage{listings}
\usepackage{color}
\usepackage{graphicx}

\defaultfontfeatures{Scale=MatchLowercase}
\setmainfont[SmallCapsFont={* Caps}]{Liberation Serif}
\setsansfont{Liberation Sans}
\setmonofont[SmallCapsFont={Courier New}]{Courier New}
\fontsize{10}{10}
\graphicspath{{fig/}} % set default import pat

% among other \newcommands the follow environment for
% indented alltt (verbatim) text
\newenvironment{display}{%
  \def\FrameCommand{\hskip\parindent}%
  \MakeFramed {\advance\hsize-\width \FrameRestore}%
  \small
  \begin{alltt}
  }%
{\end{alltt}\endMakeFramed}

My master text go.tex used \input to include all other pieces of text, so there I have:

\chapter{Intro}
\input{chapter-intro.tex}

This keeps the main document file nice and clean and you can easily re-arrange entire chapters. After any significant change I just run make and view the resulting pdf file with evince, which is an awesome pdf viewer btw. It looks something like this:

The very-much-work-in-progress pdf can be seen here.

Maar nu lijkt het toch bijna zover te zijn. Ben benieuwd wanneer we echt overstappen en er een officiele DNSSEC root key wordt gedistribueerd.

% dig . @l.root-servers.net dnskey 

; <<>> DiG 9.5.1-P2.1 <<>> . @l.root-servers.net dnskey
;; global options:  printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 64132
;; flags: qr aa rd; QUERY: 1, ANSWER: 2, AUTHORITY: 0, ADDITIONAL: 0
;; WARNING: recursion requested but not available

;; QUESTION SECTION:
;.              IN  DNSKEY

;; ANSWER SECTION:
.           86400   IN  DNSKEY  256 3 8 AwEAAa1Lh++++++++++++++++THIS/IS/AN/INVALID/KEY/AND/SHOULD
/NOT/BE/USED/CONTACT/ROOTSIGN/AT/ICANN/DOT/ORG/FOR/MOR E/INFORMATION+++++++++++++++++++++++++++++++++++++++++++ +++++++8
.           86400   IN  DNSKEY  257 3 8 AwEAAawBe++++++++++++++++THIS/IS/AN/INVALID/KEY/AND/SHOULD
/NOT/BE/USED/CONTACT/ROOTSIGN/AT/ICANN/DOT/ORG/FOR/MOR E/INFORMATION+++++++++++++++++++++++++++++++++++++++++++ 
++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 
++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++8=

;; Query time: 119 msec
;; SERVER: 199.7.83.42#53(199.7.83.42)
;; WHEN: Thu Jan 28 08:43:00 2010
;; MSG SIZE  rcvd: 439

It should rock! :-) Further details can be found on the temporary project page.

Dit is nog lang niet af, maar het begin is er. Er moet nog

• eens goed doorgelezen worden;
• beter lopend verhaal van worden gemaakt;
• spelling controle is ook geen overbodig luxe;
• vergeleken worden met de laatste Engelse versie.

Heb je zin om te helpen, laat het gewoon even weten, als het allemaal teveel wordt open ik wel een account op github.

Dit natuurlijk allemaal onder het mom van; release early, release often. :-)

You can get the text version here.

Things do

• Finish translating the last few paragraphs;
• Proof read;
• Spell check;
• Update it to the latest Go release;
• Get this included in Go?

These are the release notes for the soon to be released rdup 1.1.0.

This is the first release of the new rdup branch, version 1.1.0.

Why this new branch?

I (and because of user requests) wanted to implement some extra features which needed (extensive) code modifications. I did not want to mess with the current stable release, hence a 1.1.x branch. However 1.1.x will obsolete 1.0.x soonish.

What is the difference with 1.0.x? (aka new features)

1. there is now only one default output format which always includes the files' contents;
2. the rdup output is now more comparable with the tar archive format, with the big difference of being able to delete files.
3. the new output includes: user- and groupnames and the modification timestamp of the file. They can now be correctly set on remote systems. (I.e the uid 'bert' stays 'bert' even if the remote side using a different uid numbers for 'bert');
4. more efficient, files are only read once. In rdup 1.0.x both rdup and rdup-tr read files, in 1.1.x only rdup reads files.

With this change the following new features were added

• rdup now has a -a switch: restore a-time when reading files;
• rdup-up now sets the m-time on files it creates, this is the default;
• rdup-up sets the user- and groupname more correct, see point 3. above;
• rdup-tr can now be run on remote systems, because it receives the files' content on standard input.

This work led to the following changes

• rdup has gotten the -P flag from rdup-up. This also makes the -a flag work even if you pipe the files' through external programs: they are only read once from disk;
• rdup-tr only processes from stdin to stdout;
• rdup-up only processes stdin;
• much more tests are added to the test suite (make check);
• code cleanups.

Some future work will of course remain, but I'm happy with how things turned out. A proper release will happen this week.

Those interested may check out the latest git code and play with that.