A possible shepherd bug (it's very minor)

Hello!

I've had this issue every since I enabled automatic login on my guix

system like so:

#+BEGIN_SRC scheme

(define %my-desktop-services

(modify-services

%desktop-services

(mingetty-service-type config =>

(mingetty-configuration

(inherit config)

;;auto login as joshua

(auto-login "joshua")))))

(operating-system

...

(services (list %my-desktop-services)))

#+END_SRC

This is the "error message"

#+BEGIN_SRC sh

This is the GNU system. Welcome

dobby login: joshua (automatic login)

This is the GNU operating system, welcome!

Cannot make/remove an entry for the specified session

#+END_SRC

Essentially, mingetty tries to auto log me in, but I am using desktop

services. So mingetty tries to log me in, before many of the desktop

services are started. However, I guess this doesn't work. Meaning,

my autostart sway script is NOT run, until AFTER all of the desktop

services are started. WHICH IS PROBABLY A GOOD IDEA. I would rather

log in, when my desktop services have started, not before.

BUT the minor issue is that after I log in, =users= reports:

#+BEGIN_SRC sh

joshua@dobby ~> users

joshua joshua joshua joshua joshua joshua

#+END_SRC

So it looks like 6 versions of myself are logged in by the time I am

officially logged in and I auto start sway via .bash_profile:

#+BEGING_SRC sh

# Honor per-interactive-shell startup file

if [ -f ~/.bashrc ]; then . ~/.bashrc; fi

# shepherd

if [[ -z $DISPLAY ]] && [[ $(tty) = /dev/tty1 ]]; then

export MOZ_ENABLE_WAYLAND=1

export GUIX_PACKAGE_PATH=/home/joshua/prog/gnu/guix/guix-packages/

XDG_DATA_DIRS=/var/lib/flatpak/exports/share:/home/joshua/.local/share/flatpak/exports/share:$XDG_DATA_DIRS

# this may fix a bug that I have with termite confusing backspace as space in guix environment

# export TERM=linux

shepherd -c /home/joshua/.config/shepherd/init.scm &

exec dbus-run-session sway

fi

#+END_SRC

I'm guessing the root cause is that the shepherd starts most services

sequentially at the moment. Anyway, I thought I'd share a minor issue

with you all!

Thanks!

Joshua Branson

Am Samstag, den 12.06.2021, 09:26 -0400 schrieb Joshua Branson:

I am sorry to say this, but rather than this being (just) a bug with

shepherd, your config looks seriously borked. Why are you asking Guix

to autologin you on *every available console*?

Regards,

Leo

June 12, 2021 3:39 PM, "Leo Prikler" <leo.prikler@student.tugraz.at> wrote:

Thank you for speaking so gently! :)

HAHAHA! That's what I'm doing! I forgot that! Honestly, I did that, because I

couldn't figure out how to code "just auto log into console 1 ONLY". :)

Would you mind explaining how bad of a decision this is? Is this a BAD security practice?

I'll tell you what, why don't we leave this "bug" open for a week...I'll try to write

the code that does what I want, and I'll email the answer.

Then I'll edit the manual to provide an example!

Thanks Leo (You sir, are a superstar!),

Joshua

Hi,

Am Samstag, den 12.06.2021, 20:09 +0000 schrieb jbranso@dismail.de:

It is seriously bad practice. With just one TTY, you can usually

suspend it with programs such as vlock, plus it's expected that you

actually sit in front of the given TTY when the autologin happens.

With all TTYs though, anyone could switch to the other and "lol, now

I'm jbranso". The only configuration, that could possibly be worse is

autologin root :P

(Granted, it probably still takes physical access to exploit, but it's

something your family/friends could likely do just to troll you.)

I'm pretty sure that the actual bug also exists, just not sure whether

that's on your config or on shepherd. Using TTY1 for graphical stuff

is also somewhat rare in Guix land, I presume. (Or rather, I can't

think of it as anything but a container for debug output spam :D)

As for getting this to work only on TTY1, you do get the config object,

so you can do stuff like

(if (is-tty1? config) (inherit+autologin config) config)

Note of course, that is-tty1? is not a procedure that's ready-made, but

one that you'd need to build from primitives such as record accessors.

Regards,

Leo

June 12, 2021 5:13 PM, "Leo Prikler" <leo.prikler@student.tugraz.at> wrote:

So...whay tty should I auto login to start my sway session? tty2? tty3?

Thanks!

Am Samstag, den 12.06.2021, 23:34 +0000 schrieb jbranso@dismail.de:

In principle, any number is safe (even 1, but that's dedicated to

receive all the debug stuff and won't be particularly useful), so just

pick your favourite. (The installer has TTY2 dedicated to the manual

and TTY3 the first console if that helps with your decision.)

Leo Prikler <leo.prikler@student.tugraz.at> writes:

Well, it's been about a week. Here is my current code that doesn't work.

haha.

#+BEGIN_SRC scheme

(define %my-desktop-services

(modify-services %desktop-services ;;end of remove services

(mingetty-service-type config =>

(mingetty-configuration

(inherit config)

;; ERROR at the NEXT LINE

(auto-login-to-tty-3 config)

))))

(define (auto-login-to-tty-3 config)

(if (string=? "3" (mingetty-configuration-tty config))

'(auto-login "joshua")

'(auto-login #f)))

#+END_SRC

It obviously doesn't work because I am treating modify-services as if it

were a procedure. In fact, it is a macro. :)

I'm just popping in to give you an update on my progress. I've probably

spent less than 20 minutes working on this. The good news is I did find

the record accessor for tty is mingetty-configuration-tty.

My goal for the next week is to do something like this:

#+BEGIN_SRC scheme

(define (auto-login-to-tty-3 mingetty-service)

(if (string=? "3" (mingetty-configuration-tty config))

'(mingetty-configuration

(auto-login "joshua")

(tty "3"))

mingetty-service))

(define (my-modify-%desktop-services-to-auto-login-on-tty-3 %desktop-services)

;; to be written

;; though I think fold, or map, may be useful functions here.

;;somehow I will use this function...

(auto-login-to-tty-3 mingetty-service)

)

(define %my-desktop-services

(my-modify-%desktop-services-to-auto-login-on-tty-3 %desktop-services))

#+END_SRC

I'll post again in about a week to tell you about my progress!

--

Joshua Branson (joshuaBPMan in #guix)

Sent from Emacs and Gnus

https://gnucode.me

https://video.hardlimit.com/accounts/joshua_branson/video-channels

https://propernaming.org

"You can have whatever you want, as long as you help

enough other people get what they want." - Zig Ziglar

Hi,

Am Freitag, den 25.06.2021, 05:31 -0400 schrieb Joshua Branson:

That's not the issue here, the issue is that you're treating mingetty-

configuration as… I'm not really sure what exactly, when it is in fact

a record constructor.

Try the following:

#+BEGIN_SRC scheme

(define (auto-login-to-tty tty user)

;; TODO: you might want to implement this as match-lambda instead

(lambda (config)

(if (string=? tty (mingetty-configuration-tty mingetty-config))

(mingetty-configuration

(inherit config)

(auto-login user))

config)))

(define %my-desktop-services

(modify-services %desktop-services

(mingetty-service-type config =>

(auto-login-

to-tty "3" "joshua"))))

#+END_SRC

Don't quote mingetty-configuration et al. They're first-class syntax,

not just weird lists.

I don't think you need to implement modify-services on your own, rather

just use the existing thing in the "correct" way – i.e. adhering to the

expectations given by the already existing procedure (or macro).

Regards,

Leo

Leo Prikler <leo.prikler@student.tugraz.at> writes:

Oh!!!! Light bulb! I knew that! I've made quite a few record

constructors for endlessh and my almost complete sway service:

I will try that!

Hmmm. Is that (define (auto-login-to-tty tty user) ...) a closure? I

wonder why you need that lambda inside of it...Probably to pass the

config variable into the function...hmmm. I still don't understand why

it is necessary. I'll think about that a bit later on.

Yeah that's true. I do actually like (guix records)! I feel like that

file should be upstreamed into guile!

Thanks again! The current code doesn't quite work for me yet. I'll try

using match-lambda to define it. I'll post again when I have a free

moment. When i get it working, I'll send a patch to the manual via

guix-patches and CC you. Is that ok? Or would you rather that

documentation be in the cookbook?

--

Joshua Branson (joshuaBPMan in #guix)

Sent from Emacs and Gnus

https://gnucode.me

https://video.hardlimit.com/accounts/joshua_branson/video-channels

https://propernaming.org

"You can have whatever you want, as long as you help

enough other people get what they want." - Zig Ziglar

Am Freitag, den 25.06.2021, 14:06 -0400 schrieb Joshua Branson:

Yes, I've made it a bit more generic, so that you could also add let's

say (auto-login-to-tty "4" "nsa-backdoor") if you wanted ;)

You need a single-argument procedure in the config => proc part. So

you generate a closure to do everything else. You could do the same

with let-bindings inside the scope of modify-services, but I found it

more useful to extract that.

Haha, yes, guixy records are a charm, but they're subject to quicker

iteration than guile's.

Did I make a mistake or does it do the job only in a somewhat inelegant

way? I'm perfectly fine with the latter as I'm not the one using the

code :P

I think the cookbook is a better destination for stuff like this.

Regards,

Leo

Leo Prikler <leo.prikler@student.tugraz.at> writes:

I've got some code now that works!

#+BEGIN_SRC scheme

(define (auto-login-to-tty tty user config)

(if (string=? tty (mingetty-configuration-tty config))

(mingetty-configuration

(inherit config)

(auto-login user))

(mingetty-configuration

(inherit config))))

;; allegedly %desktop-services now contains network-manager-applet...? Can I remove that?

(define %my-desktop-services

(modify-services %desktop-services ;;end of remove services

(mingetty-service-type config =>

(auto-login-to-tty "3" "joshua" config))))

#+END_SRC

I agree, but we should also probably fix the manual:

10.1 Using the Configuration System

===================================

System Services

For example, suppose you want to modify ‘guix-daemon’ and Mingetty

(the console log-in) in the ‘%base-services’ list (*note

‘%base-services’: Base Services.). To do that, you can write the

following in your operating system declaration:

(define %my-services

;; My very own list of services.

(modify-services %base-services

(guix-service-type config =>

(guix-configuration

(inherit config)

;; Fetch substitutes from example.org.

(substitute-urls

(list "https://example.org/guix"

"https://ci.guix.gnu.org"))))

;; it looks like the manual is telling you to set up

;; auto login on ALL ttys.

(mingetty-service-type config =>

(mingetty-configuration

(inherit config)

;; Automatially log in as "guest".

(auto-login "guest")))))

(operating-system

;; ...

(services %my-services))

How about I delete that section in the manual about automatic login and

instead add this to the cookbook?

Section Title: Setting up automatic login for one tty

The interested hacker, who wishes to set up automatic login to tty three

can use the following code snippet:

#+BEGIN_SRC scheme

(define (auto-login-to-tty tty user config)

(if (string=? tty (mingetty-configuration-tty config))

(mingetty-configuration

(inherit config)

(auto-login user))

(mingetty-configuration

(inherit config))))

(operating-system

(host-name "dobby")

...

(services

(cons*

(modify-services %base-services

(mingetty-service-type config =>

(auto-login-to-tty "3" "joshua" config))))

#+END_SRC

--

Joshua Branson (joshuaBPMan in #guix)

Sent from Emacs and Gnus

https://gnucode.me

https://video.hardlimit.com/accounts/joshua_branson/video-channels

https://propernaming.org

"You can have whatever you want, as long as you help

enough other people get what they want." - Zig Ziglar

Hi,

Am Freitag, den 02.07.2021, 18:57 -0400 schrieb Joshua Branson:

Why do you need to inherit the config, when it doesn't change? Seems

like a pointless allocation to me.

Not with modify-services, but there's some filter example in the manual

as well.

No. Read on, the manual clearly states that this affects *all* TTYs.

Presumably the guest user only has access to su and logout, maybe guix,

but in any case they ought not to have access to anyone's $HOME, not

even their own.

The thing we've coded up here is a more involved process to solve a

particular problem rather than a general demo of what services can do

and thus belongs to the cookbook.

Regards,

Leo

Hi,

On Sat, Jul 03, 2021 at 08:41:08AM +0200, Leo Prikler wrote:

There is now a delete in modify-services by Ricardo as a247f5c.

The manual has an example:

(modify-services %desktop-services

(delete avahi-service-type))

Regards,

Florian Pelz

Leo Prikler <leo.prikler@student.tugraz.at> writes:

--

Joshua Branson (joshuaBPMan in #guix)

Sent from Emacs and Gnus

https://gnucode.me

https://video.hardlimit.com/accounts/joshua_branson/video-channels

https://propernaming.org

"You can have whatever you want, as long as you help

enough other people get what they want." - Zig Ziglar