Stanford provides a cluster of computers for use in CS classes known as myth that is equipped with gcc, gdb, Valgrind, etc. I’m using these computers as part of CS 107 this quarter. (If you aren’t a Stanford student taking CS 107, this post might be totally irrelevant to you, but maybe you can still get something out of it.) These machines can be a bit of a pain to use for someone who has never worked over SSH before, but there are a few things you can do to make life happier.

EDIT 6/8/20: This blog post is a bit out of date. I have written up some more recent tips here.

Avoid SSH timeouts

If you let an SSH session sit too long without activity (sometimes even as short as a few minutes), you’ll often get connection issues:

Write Failed: broken pipe

This is easy to solve by adding the following to ~/.ssh/config (you may need to create this file if it does not exist):

Host *
    ServerAliveInterval 60

This sends a “null packet” every 60 seconds telling the server “hi, I’m still here!” and reduces the likelihood of a session timeout.

Note: This tip works on Mac and Linux; on Windows, open PuTTY, go to Connection (in the left pane), and set “Seconds between keepalives” to 60.

Set up Kerberos/SSO (for mac users)

EDIT 6/8/20: This tip no longer works on more recent versions of Mac OS. I recommend setting up persistent connections instead.

Wouldn’t you rather type ssh myth and be ready to roll, instead of typing ssh [email protected], followed by your password, every time you want to connect? (I’m lazy, I know. ¯_(ツ)_/¯) Kerberos is a system that Stanford uses to manage authentication to its computers. If you set up Kerberos on OS X, you can type your credentials once and get an authenticated “ticket”; then, all SSH connections will automatically use that ticket instead of prompting you to enter your password every time. (Note for people familiar with SSH keys: Yes, SSH keys would allow us to easily log in too, but Stanford doesn’t support public-key authentication. ಠ~ಠ)

To set this up, open Keychain Access via Spotlight or Go > Utilities in Finder. Open the Ticket Viewer:

Click “Add Identity,” and enter your Stanford credentials:

Ticket Viewer should then show an authenticated ticket, expiring in 10 hours:

If you checked “Remember password in my keychain” while entering your credentials, OS X should automatically renew this ticket, and you shouldn’t need to enter your password again. If you ever change your password, you can come back here, click Remove Identity, and add it again.

Now that you have an authenticated Kerberos ticket, you need to configure SSH to send that ticket to the myth machines, proving that you have authenticated. Edit (or create) ~/.ssh/config, adding the following:

Host myth myth.stanford.edu
    HostName myth.stanford.edu
    User rebs
    GSSAPIKeyExchange yes
    GSSAPIAuthentication yes
    GSSAPIDelegateCredentials yes

This does a few things:

• The first line declares that this configuration should apply when we ssh myth or ssh myth.stanford.edu.
• The second line says that the server we want to connect to is at myth.stanford.edu. (This matters, because if we ssh myth, ssh doesn’t have a domain name to connect to and won’t know what to do.)
• The third line specifies your username. If your SUID is the same as your computer’s username, you don’t need this. However, if they differ and you don’t want to have to specify your SUID as ssh suid@myth every time, you can specify your username here.
• The last three lines configure SSH to forward Kerberos tickets as a means of authentication.

Now, you can ssh myth, and you should land at a shell prompt without needing to type anything extra. Yay lazy people ꒰･◡･꒱

Note: these instructions are for Mac, but I’m sure there’s also a way to do this in Windows or Linux.

Add the CS 107 tools directory to your $PATH

PATH is an environment variable that contains a list of folders separated by colons, perhaps like this:

$ echo $PATH
/usr/local/bin:/usr/local/sbin:/usr/bin:/bin:/usr/sbin:/sbin

Whenever you run a command, your shell checks that list of folders for a binary with that name. If you run cp with the above PATH, the shell will first check /usr/local/bin for a binary named cp – not there – then /usr/local/sbin – also not there – then /usr/bin – not there either :-/ – and then /bin – ah, found it!

CS 107 has a /afs/ir/class/cs107/tools directory with a few important executables. Instead of trying to remember that whole path and running /afs/ir/class/cs107/tools/sanitycheck custom_tests every time you want to test your program, you can add /afs/ir/class/cs107/tools to your PATH; then, you can just run sanitycheck custom_tests or submit and things will Just Work™.

To do this, log into myth and edit (or create) ~/.bashrc (or the configuration file for your shell of choice – but then you probably already know how to do this ʘ‿ʘ) and add the following at the end:

export PATH=$PATH:/afs/ir/class/cs107/tools/

Now try running sanitycheck or submit in an assignment directory, as you might run cd or ls. Voila!

Symlink the cs107 directory

Not only can I not remember where submit and sanitycheck are (re: above tip), I also have the hardest time remembering the path to the class samples and repos. Fortunately, it’s easy to create a symlink (symbolic link, like a shortcut or alias) from your home directory to the cs107 directory. Simply run ln -s /afs/ir/class/cs107/ ~/class (or replace ~/class with a path that you like better):

$ ln -s /afs/ir/class/cs107/ ~/class
$ ls class
ARCHIVE.README  cgi-bin  private_data  repos  samples  staff  tools  WWW
$ ls class/samples/lect4/
addr    addr.o        array_sizeof.c  Makefile  sizes.c  strsort    strsort.o      uninitialized.c
addr.c  array_sizeof  array_sizeof.o  sizes     sizes.o  strsort.c  uninitialized  uninitialized.o

Now, the cs107 folder is easily accessible from your home directory. (Note that this is very different from copying the folder to your home directory; a symbolic link is just an alias, so as the professor modifies the code in /afs/ir/class/cs107/samples/ during lecture, those changes will be reflected in ~/class/samples.)

Learn your editor

Really. CS 107 has guides for vim and emacs. If you use vim, take some time to work through vimtutor (just run the command). There are lots of vim tutorials online, and vimcasts is a cool resource.

EDIT 6/8/20: I have gotten much better at using Vim since I wrote this :) I have a better list of vim tips here.

There are lots of good resources that I won’t attempt to match, but here are some random vim tips:

• Press u to undo and ctrl+r to redo
• y<motion> will copy/yank, d<motion> will cut, and p will paste. For example, press dd to cut a line and p to paste it, or yw to copy a word. If you’re uncomfortable with motions, press v to enter visual mode, then move your cursor until you select what you want to copy or cut, then press d or y, and then paste where desired.

• You can configure vim to automatically indent things (add to ~/.vimrc):

  set tabstop=4     " Tab characters will take up 4 characters' width on the screen
  set expandtab     " Use spaces instead of actual tab characters to indent
  set shiftwidth=4  " If you press tab, 4 spaces will be inserted
  set smarttab      " Automatically indents things (e.g. indents a level when you open a code block)
  

• Enable relative line numbers! If I wanted to delete from the current line (line 6) to the line reading set incsearch, I have a few options. I could spam dd until all the desired lines are gone, but that’s kind of slow… I could count how many lines down set incsearch is, but that’s also very slow if you want to delete many lines up/down. Or, I could look at the relative line numbering to the left, see that the desired line is 5 lines down, and then hit d5j.

  

  To enable this, add the following to your ~/.vimrc:

  set number
  set relativenumber
  

• (On a mac, but maybe also on other systems) Enable mouse interaction! If you add set mouse=a to your vimrc, you can click around in the editor, scroll, drag to highlight, etc… In general, it is better/faster to use a keyboard to navigate and do things, but sometimes being able to use a mouse is nice.
• You can search in vim by typing / followed by a search term. To have search results highlighted as you type, add set incsearch to your vimrc.
• By default, backspace won’t work over line breaks (i.e. if you hold backspace to delete at a line, the cursor will stop at the beginning of the line instead of continuing to delete the preceding line, contrary to how most apps works). To make backspace work like most apps do, add set backspace=2 to your vimrc.

Optional: Set a custom shell

The default shell on myth is bash. It seems you can change your shell across all Stanford machines by emailing Research Computing, but short of doing that, you can add exec zsh (or whatever shell) to ~/.bash_profile. (Don’t add it to .bashrc, or else running bash from zsh will give you another zsh shell, and running bash somescript.sh will execute the script with zsh instead of bash.)

Depending on your situation, it might be nice to change your shell prompt so that you know you’re running on a different host. My zsh prompt is a strawberry on my local machine, but a trident on myth, and a space invader when the return code is non-zero :3

If you use iTerm, you can even set up profile switching so that e.g. your background changes colors when you ssh into myth.

Conclusion

Hopefully these tips help make myth a little nicer to work in. If you have any tips or suggestions of your own, leave them in the comments below!