Happy Birthday, Debian!

Image taken from https://bits.debian.org/2018/08/debian-is-25.html, published under the MIT License (see: https://www.debian.org/license)

On Aug 16^th, 1993 the 20 year old Ian Murdock released a message on comp.os.linux.development (that’s a Usenet Newsgroup – Usenet was the way to connect to and interact on the Internet before there was the World Wide Web that you are using now (that’s why you have to put www infront of most websites), and Newsgroups where the forums/message boards of the Usenet) announcing his new Linux distribution called Debian (named after his then girlfriend/later wife Debra and himself Ian).

Before distributions, Linux only consisted of the kernel itself that was in source code – so you had to download that onto your host system, compile it, and by hand from your host system set up your machine with all the components needed to run it as a Linux system. This was really complicated, and so distributions arose, allowing the user to simply install the system from a bootable device:

Debian will contain a installation procedure that doesn’t need to be  babysat; simply install the basedisk, copy the distribution disks to the harddrive, answer some question about what packages you want or don’t want installed, and let the machine install the release while you do more interesting things

For the elder generations, 25 years might not seem as much. But watching back, this actually makes Debian one of the oldest Linux distributions and probably the oldest one that is still actively used; in comparison, the Linux kernel itself is just 2 years older! Praised for its stability it is the choice of system administrators.

Debian’s package management system is – as far as I could research it – the oldest and therefore first one, and it is still actively used and ported to different systems even today: .dep packages were ported to the UNIX System V via OpenSolaris, a port to BSD via the UNIX-like FreeBSD, which also macOS is derived from, and speaking of Apple – Fink brought Debian’s package management system to macOS and Cydia to iOS. According to Murdock himself, package management is “the single biggest advancement Linux has brought to the industry”. He himself later worked for Docker, which is – if you will – in a way an even bigger package managing software in the sense that it not only packages the software itself but also the entire running system.

It also is the basis of countless derivates, from which the most famous or important are probably Knoppix, Grml, Kali Linux, Raspbian and of course the most important of all: Ubuntu, that itself is the basis of countless additional derivates (the Wikipedia has a great graph showing all Linux distributions and how they a related to each other – Debian makes up for more than a third of that space). And Distrowatch’s Top 10 Distributions list (which actually contains 11 distros) lists 3 Debian-based distros, including Debian itself.

So it’s more than fair to say, that Debian had a great impact on the Linux world, helped shape it to what it is today. I myself therefore wish the Debian system and it’s developing team a very happy birthday, and hope that it will see the next 25 years with same prosperity!

And if you are puzzled as to why it is I talk about Debian – well I myself used that distribution for a year in my early Linux days. I started out with S.u.S.E. Linux 6.3 at arround 1998-2000. I got fed up with it after they introduced YaST2 for SuSE 8.0 and switched to Debian 3.0 aka Woody in 2004. However, we weren’t meant to be for each other. It was incredible outdated, which made me really laugh at this passage of the Newsgroup post:

Debian will contain the most up-to-date of everything.

While other distributions of that times where enjoying Kernel version 2.6, Debian came with 2.2 in the stable branch. It was also missing a lot of software – which at that time where Linux software was still a rare thing, meant a lot, especially in the everyday personal desktop PC context. After mixing stable, testing and unstable packages and dreading every update, because it meant another weekend trying to repair the system, I once more switched – to Gentoo.

However, I was always fond of Debian, their goals and their Manifesto. Debian has always been my goto distribution for running servers. And lately – well, in April I switched from Gentoo to Ubuntu as my main driver. For mainly two reasons:

1. With the last months and years Gentoo has become more and more unstable. While in my beginning years I could simply run emerge -avuND world && shutdown, go to bed and have an up-to-date system the next morning, nowadays most times this fails and I need to spend hours and days fixing it. That was my main reason to leave Debian long time ago – now it’s the reason for me leaving a distribution that has been my main driver for 14 years. That’s hard, yes. But also somewhat exciting.
2. I need certain software that unfortunately is just developed for Ubuntu and does not work under other Linux distributions – either at all, or only with limitations. I am really frustrated about that, because developing software only for one certain distribution is not at all the Linux way and shouldn’t be rewarded – however if you need the software…. This list of software includes (without being complete):
   • • ROS: Supported platforms are Ubuntu and sometimes Debian. On Gentoo installation was tricky but possible, however some packages weren’t ported. The maintainer is fast to react, but as I really needed one package really fast, the switch was inevitable.
     • Rock: Like ROS but with real-time support from Orocos, developed by the DFKI. Like ROS it’s Ubuntu only, even if the website says something different. I tried the installation under Gentoo, Arch and macOS and blatantly failed. During my time at the DFKI I did not meet anyone using anything else than Ubuntu, and even on Ubuntu installation was buggy as hell.
     • Unreal Engine: While under Linux you always have to build it from source, and it is not Ubuntu-only (and provides installation pointers for CentOS, Fedora, Arch and Mint), UE4 runs more smothly on Ubuntu and is easier to build than it was under Gentoo.

While Ubuntu wouldn’t have been my first choice I have to say that I am pretty happy with it. It’s different for sure and I still have a number of smaller issues, but we get along. On the longrun however I might also be looking into Debian once again.

PS: On a rather sad note – in 2015 for publicly unknown reasons the founder and inventor of Debian (an American who was actually born in Konstanz, Germany!) killed himself on rather mysterious circumstances after having something that could be called a nervous breakdown at the age of 42, leaving behind three children. May he rest in peace. Your legacy will live on!

Automount a specific USB drive

Actually this is a straight forward thing, however since it has been a while I had to google it myself, and was astonished about how many non-working solutions I found, besides solutions that simply mount every USB according to their label. So here is the straight forward solution to mount a specific USB device to a specific location on your Unix hierarchical file system, using udev. It assumes that you have a running version of udev, and the udev tools. If not, please consult the distribution specific documentation on the Linux distribution of your choice. This might include recompiling your Kernel, as udev will need the following settings:

General setup --->
[*] Configure standard kernel features (expert users) --->
[ ] Enable deprecated sysfs features to support old userspace tools
[*] Enable signalfd() system call
Enable the block layer --->
[*] Block layer SG support v4
Networking support --->
Networking options --->
<*> Unix domain sockets
Device Drivers --->
Generic Driver Options --->
() path to uevent helper
[*] Maintain a devtmpfs filesystem to mount at /dev
< > ATA/ATAPI/MFM/RLL support (DEPRECATED) --->
File systems --->
[*] Inotify support for userspace
Pseudo filesystems --->
[*] /proc file system support
[*] sysfs file system support

As for Gentoo Linux, the other things you will want to do, is to add “udev” to your USE-flags (by adding it into your /etc/portage/make.conf), get udev installed (calling emerge -avuD sys-fs/udev), and add udev to your sysinit runlevel (rc-update add udev sysinit).

Now to the fun part. First of all you need to get some information about the device you are interested in. There are a number of ways, like using udev monitor, etc. Most of them, to me however, are too messy. If you have no idea about your device and still need to figure things out, blkid -o list will show you a nice table of all devices, their device file, file system type, label, mount point and UUID – everything you need. For me, I know I have a stick with the label “Public” on an OS X with file system type exFat, and now I inserted it into a dual boot Linux with a number of partitions:

ancalagon ~ # blkid -o list
device                            fs_type      label         mount point                           UUID
---------------------------------------------------------------------------------------------------------------------------------------
/dev/sdb1                         vfat                       /boot                                 58FB-332D
/dev/sdb2                         swap                       [SWAP]                                73db158f-0e19-4d17-8c88-a8b0c1dff1f3
/dev/sdb3                         ext4                       /home                                 355af6d8-6f03-4a98-9a45-edafc3ccedde
/dev/sdb4                         ext4                       /                                     63be67f3-5c7c-48ea-a8b3-58dff9da1737
/dev/sda1                         ntfs         Wiederherstellung (not mounted)                     562065062064EF05
/dev/sda2                         vfat                       (not mounted)                         6265-B138
/dev/sda4                         ntfs                       (not mounted)                         6C58731C5872E46C
/dev/sdc1                         exfat        Private       /media/private                        56B6-CE90
/dev/sdd1                         exfat        Public        (not mounted)                         56BE-6477
/dev/sda3                                                    (not mounted)

If you want more information, with the device file you can get it with:

ancalagon ~ # udevadm info /dev/sdd1

I want the stick to be mounted at /media/public, so I need to create a rule file; on Gentoo it lies under /etc/udev/rules.d/90-local-usb.rules. Actually the name is totally arbitrary, except for the number at the beginning, and the extension that always has to be .rules. The number should be something high, because we want udev to first run all other rules (e.g. the ones that assign the device to a device file) before running ours. 90 is a good value for that.

So in my case, this is what I added:

SUBSYSTEMS=="usb", ENV{ID_FS_UUID}=="56BE-6477", ACTION=="add", RUN+="/usr/bin/logger --tag udev Mounting public", RUN+="/bin/mount -o umask=0077,nosuid,uid=1000,gid=1001 '%E{DEVNAME}' /media/public"

We need to provide the system or subsystem, which for an USB device is usb. The UUID comes from blkid and identifies the device. The action triggers when to run the command. In our case, when a new USB device is added and it has the UUID we want. And finally the mount command. I’ve added another command such that there is a log entry but thta is no need. And as I want it to be accessible as user, I added uid and gid accordingly. If you need to find out your user and group id, just run:

ancalagon ~ # id -u 
ancalagon ~ # id -g 

And that’s it. If you want to see if the rule triggers, just run

ancalagon ~ # tail -f /var/log/everything/current

It should output:

[udev] Mounting public

somewhere. And you can <em>simulate</em> the USB event with udevadm, by triggering the rule you just wrote (although this is rather interesting for more general rules that should fit more than just one device). This is how it’s done:

ancalagon ~ # udevadm trigger --action="add" --property-match=ID_FS_UUID="56BE-6477"

dyld: shared cached file was build against a different libSystem.dylib, ignoring cache

After my last update of OS X 10.8.5 my terminals were flooded with above message – like twenty-ish lines when I started and up to six times after each console command I issued, making it impossible to work with the console anymore.

The error message also told me to run update_dyld_shared_cache and restart my machine. That’s what I wanted to do, but when I issued the command this is what my computer presented me with:

pygospa@lalaith ~ % update_dyld_shared_cache
update_dyld_shared_cache failed: you must be root to run this tool

So I switched to my administrator account which who is in sudoers and typed in the command again:

eru@lalaith ~ % sudo update_dyld_shared_cache
 update_dyld_shared_cache: for arch i386, can't put /usr/lib/libtidy.A.dylib in shared cache because it is not owned by root
 update_dyld_shared_cache: for arch i386, can't put /usr/lib/libncurses.5.4.dylib in shared cache because it is not owned by root
 update_dyld_shared_cache: for arch i386, can't put /usr/lib/libutil.dylib in
 (...)

There where at least 50 of those lines, and as it seemed nothing had been done, so I checked the permissions and as it turned out, all files in /usr where in possession of root except for /usr/lib, which belonged to the administrator. Huh, that’s strange. Anyway, let’s run it with the rights of the administrator then:

pygospa@lalaith ~ % update_dyld_shared_cache
update_dyld_shared_cache failed: you must be root to run this tool

Okay, I actually don’t like changing owners on a Mac, as it might break something, so my first try was to just make the files writeable for everyone. So I switched to /usr/lib, entered a chmod 777 * and got a bunch of permission denied responses:

eru@lalaith /usr/lib % chmod 777 *
chmod: Unable to change file mode on cron: Operation not permitted
chmod: Unable to change file mode on libCoreStorage.dylib: Operation not permitted
chmod: Unable to change file mode on libIOKit.A.dylib: Operation not permitted
(...)

Coming from Linux this is really strange, as eru was in possession of all those files… I tried making the directory writeable for everyone – with the same effect. Okay, let’s switch to root – root can do anything.

You’d think…

root@lalaith # l /usr/lib
drwxr-xr-x 316 eru wheel 10744 15 Sep 18:33 lib/
root@lalaith # chown root /usr/lib
chown: /usr/lib: Operation not permitted

Okay. Now I was really puzzled. What the frack is up with this strange behaviour? I did, what any good Windows User does, when he encounters a problem. I rebooted. Same old, same old. Next stop: Google-Town.

Tried sudo update_dyld_shared_cache -force and sudo update_dyld_shared_cache -root / -force, sudo update_dyld_shared_cache -verify

(…)If you suspect the shared cache is somehow corrupt, you can run: sudo update_dyld_shared_cache -verify which re-creates the cache and compares the result with your actual cache file.

and a bunch of other stuff. Nope. Okay, Apple Support Communities, hosted under discussions.apple.com – sometimes they have pretty good answers. And yep, there was someone with the same problem, and there was an answer marked as “solving my questions”, and it was short, and it just said… what?!

You have a missing or damaged system file. Back up all your data, then boot from your installation DVD and reinstall the OS. Your data will be undisturbed. After rebooting, run Software Update to bring everything up to date. You may have to run it more than once.

You’re kidding. I don’t have time for this, seriously! Fracking Mac, fracking Apple, to hell with this load of crap! I was really angry.

Fortunately I don’t follow instructions straight away, and after a lot of head aching googling and cursing the day, I came across Safe Mode.

Starting up into Save Mode does several things:

• (…)
• Mac OS X v10.5.6 or later: A Safe Boot deletes the dynamic loader shared cache at (/var/db/dyld/). A cache with issues may cause a blue screen on startup, particularly after a Software Update. Restarting normally recreates this cache.
• (…)

Sounds related – and yeah, it solved my problem! Starting into Save Mode is pretty easy: Power down, hold the “Shift”-Key, Power – and keep holding the “Shift”-Key for quite a while – untill the Apple logo appears.

It’ll then take a while – at least on a recent, souped-up Macbook Air it was enough time to brew a new cup of coffee. But it actually came up, I logged in, and: Hey! In Safe Mode the annoying message is gone! Reboot, and yeah, in Normal Mode everything’s back to normal.

The ADM-3A leaving it’s mark

Did you ever ask yourself, why on *NIX-Systems your home directory has the shortcut tilde (~)? Or why on the text editor vim the cursor could not only be moved by the arrow keys but also via H, J, K and L? Why not W, A, S and D, which today is famous as it is used by many games? Well I often did ask myself, but never actually tried to find out why. I had my explanation for H, J, K, L, as they lie on the home row of the keyboard, thus allowing fast movement.

One might think that H, J, K and L an idiosyncrasy of vi/vim, but when you look carefully you find other software that use the same keys for moving: Rouge, Hack and NetHack – the predecessors of Diablo use HJKL. Also the C Shell (csh) and it’s improved and today still popular version TENEX C-Shell (tcsh) can be controlled by H, J, K and L. The most recent tools are the web interfaces from Gmail and Google Labs – as well as the browser Pentadactyl. Of course, for the later tools it’s more convenience than a historical cause. But regarding vim, by accident I just now found out why these keys are used – and why the tilde is the shortcut for the home directory on *NIX systems.