Introduction

I’m a messy guy when it comes to my /etc/portage/package.accept_keywords or my /etc/portage/package.use conf files. To that end, I created a short script using sed to get things cleaned up. I will admit that I could have avoided using a script at all if I asserted more control over what is put in those files. But if you’re lazy like me and were checking out the deployment script I cobbled together, you might find this useful. Hopefully, my laziness is of benefit to you.

Disclaimer

I’m a strong proponent of testing before implementing. Therefore, I do not take any responsibility for your use of this script on your systems and the consequences that will ensue (good or bad). Please do not just run this on a Gentoo system without first backing up your files. Remember the golden rules of IT:

1) Have a backup plan
2) Have a data backup

Follow these rules, and you will successfully recover most of the time.

Code Me

#!/bin/bash
##Name: pkgconfcleaner
##Author: Cephas0
##License? No Brah!

sed -i 's+#.*++g' /etc/portage/package.use
sed -i '/^\s*$/d' /etc/portage/package.use

sed -i 's+#.*++g' /etc/portage/package.accept_keywords
sed -i '/^\s*$/d' /etc/portage/package.accept_keywords

What does it do?

This is a rather simple thing. The first string finds all of the commented lines in your conf file of choice (hard coded here) and removes them. Then the second line removes all of the leftover space. It looks gnarly but it’s pretty simple in its objective.

I don’t know anything about scripts. How do I get started?

Using a text editor like “nano”, create a file named “pkgconfcleaner” in the /usr/local/bin folder so that the script will be in the path. Next copy and paste the code into the editor. Ok, so here’s a testing hint. If you remove the “-i” from the sed commands, it will not implement the code, but it will print out what the file would look like if it did. Neat, huh? The last thing you need to do to make this work is to run the “chmod” command to make the file executable.

chmod a+x /usr/local/bin/pkgconfcleaner

Something like the code above should be sufficient. (It might not work for you if you are not root or using “sudo”. If you got an error use “sudo” and try again. If it just failed, don’t type it out again, use “sudo !!”)

Wrap up

That’s a wrap on this post. While scripting this short script to fix my laziness may not have been the best allocation of my time, it does have other applicable uses you may find helpful.

#!/bin/bash

read -p "What pkg are you building?: " pkg

##Code attribution for the code below
##https://www.ostechnix.com/download-packages-dependencies-locally-ubuntu/

for i in $(apt-cache depends $pkg | grep -E 'Depends|Recommends|Suggests' | cut -d ':' -f 2,3 | sed -e s/'<'/''/ -e s/'>'/''/); do sudo apt-get download $i 2>>errors.txt; done

This post is about something I tried when I was working on an offline Debian upgrade project. While it didn’t ultimately provide the solution to that project, it did open up a wonderful possibility. To kick this post off, we must have a talk about dependencies, and since that can become mind numbing quickly, I’m only going to gloss over that topic. We’ll talk about what this script does, how to use it, and then turn you loose.

Dependencies _> The Underworld

Dependencies are what the majority of packages or projects rely on to work. Think of it like a base foundation that many people contribute to. This is usually in the form of “lib” or library packages. Other developers will use this pre-written code in their projects, and that’s the end of it right? Not really. Actually, a single project can use dozens to hundreds of dependencies all stacked upon one another like a pyramid of code. This can quickly become a large security issue as the more a system has installed the more dependencies it relies upon. It is at that point that the system’s security becomes more and more dependent (no pun intended) upon every dependency. In other words, the weakest link in any program is the amount of dependencies it uses as much as a chain is only as strong as its weakest link. 

So there’s some of the ugly; let’s talk about the bad for a second. Let’s say you’ve gotten entangled in a project that needs some offline packages installed. Where do you start?

The Journey

For me I started at the online Debian package repository. I needed to download Java for another project. Needless to say you quickly find that you need at least four packages right off the bat.

openjdk-8-jre.deb openjdk-8-jre-headless.deb openjdk-8-jdk-headless.deb openjdk-8-jdk.deb

Yikes! Each package has even more dependencies. And those have even more dependencies. Wouldn’t it be nice if you could just get all the packages and the dependencies without the downloads?

The Solution

I was getting desperate for a solution. Downloading package after package after package is the worst. I have a life and better things to do. Enter salvation in the form of ingenius scripting from OSTechNix. Simply make a folder of the package you wish to download and get cracking.

Here’s the code again below for reference. We’ll step through it.

#!/bin/bash

read -p "What pkg are you building?: " pkg

##Code attribution for the code below
##https://www.ostechnix.com/download-packages-dependencies-locally-ubuntu/

for i in $(apt-cache depends $pkg | grep -E 'Depends|Recommends|Suggests' | cut -d ':' -f 2,3 | sed -e s/'<'/''/ -e s/'>'/''/); do sudo apt-get download $i 2>>errors.txt; done

The Code

I’m going to assume you have made your directory and you are ready to proceed to the next step. If you want you can copy the script above and put it in your /usr/local/bin which will make your script available in your system paths. Make sure it’s executable. I usually run my scripts as root on test systems, so for your system you may wish to use “sudo” in front of whatever you named this script.

read -p "What pkg are you building?: " pkg

This is the first line I added, and it offers some bonuses. You can put as many different packages as you want, spaced out of course. It’s a simple input line for bash with the variable at the end. As you can see, we use that later.

for i in $(apt-cache depends $pkg | grep -E 'Depends|Recommends|Suggests' | cut -d ':' -f 2,3 | sed -e s/'<'/''/ -e s/'>'/''/) 
     do 
        sudo apt-get download $i 2>>errors.txt
     done

I’m going to skip over the code attribution because I think that’s rather self-documenting. The rest of this code starts with a standard for loop. What follows next is a calling of the apt-cache command and the depends command for the package ($pkg, told you we’d use it later) you want to download. Then we pipe to grep, doing a little cutting, run sed (which does some awesome clean up), and then we finally get to downloading the packages.

Wrapping it up

Before you start running this script, make sure you’re in the actual folder you created. Otherwise you could end up with a lot of deb packages everywhere. Not to worry if you did though. Here’s some shortcode to get things cleaned up. We’ll assume you’re in the /tmp/ folder, and you ran, for example, the java packages I listed out earlier. What a mess!

cd /tmp/
##gotta get in the tmp directory first right?
##remember the java folder (package folder) I made?
mv *.deb /tmp/java

And boom. You’re all good. Hope it helps.

Introduction

So you wanna make your life easier when emerging a gentoo package? Good news, so did I. I’m going to start this off with, as with many things I do, it’s not always the best way. I would say the backtracking option is very likely something you can and should take out. Read up on that and see if it makes sense for you. For me, this is something I cobbled together. It has some weak points and places it can be improved. It gets the job done though, and that’s enough for me. That said, I’ve felt like if I continue to just keep this knowledge until it’s perfect, no one can benefit from any of it. Therefore, it is incumbent on you to take responsibility for testing, improving, and revising anything I release. Furthermore, if you want to make recommendations with code examples in the comments, I’m all for it. Help each other out, and together we can make things better. Alright, with that, let’s delve into the code and start making our lives a little easier.

Gentoo? Dear Lord Man. For serial.

If you detected a Southpark vibe there, good on you. Gentoo is not for the faint of heart. It will, however, force you to be better than what you are. You will learn a lot about compiling, systems, kernels, and multiple other things. I started with Calculate Linux (a gentoo based Russian distro). If you’re completely new to linux, don’t let the steep climb throw you. Start with Calculate and learn some of the basics as well as the pitfalls. For my master project, I built out my first Gentoo system. I won’t cover that here, but in an upcoming post, I will go over what and how I did my first build. It took a month because I started with a Dell laptop and had issues with the wifi broadcom drivers. Fun times. Let’s get back to the deployment script though.

Nailed it. Code me.

#!/bin/bash
##Author: Cephas0
##Distro: Gentoo
##Knowledge level needed: Copy and Paste?
##License Brah? No...?

array=($(ls | sed 's/\(Manifest\|metadata.*\|.ebuild\|files\)//g'));

for index in ${!array[@]};do echo "$index ${array[index]}";done

read -p "Enter the pkg to be deployed: " num
pkg=${array[$num]}
current=($(pwd | sed 's+/usr/portage/++g'| sed 's|\(.*\)/.*|\1|'))
emerge -av --autounmask-continue --backtrack=200 =$pkg; dispatch-conf; 
exists=($(find /var/db/pkg | grep $pkg))
if [ -d "$exists" ]; then
   echo "Package has been deployed."
   exit
else 
   emerge -v =$pkg
   echo "Package has been deployed."
fi

Put the alcoholic beverage down and step away from the mouse

What’s all of that Greek?!?! No worries – we’ll go line-by-line through it. For now, if you want to kick it off, put it in your /usr/local/bin which will put it in the system path. Small disclosure: on my personal systems, I run everything as root. It just makes it easier for me. If you’re a sudo person, no problems, just run things as sudo. Make sure you make the script executable.

Code Breakdown

The first line of code is below.

array=($(ls | sed 's/\(Manifest\|metadata.*\|.ebuild\|files\)//g'));

I’ve set up a variable called array. The key to this script is that it assumes you have already found the package you want to install and you are in the correct portage directory.

🔥 Spark Note From the Forge 🔥

But wait. How can I get there? Well let’s use an example. Let’s say I want to install “xclock”. How would I find it?

Well there are two ways I do it; although, full admission, there are more, shall we say, proper ways to get the job done. Regrettably, you poor soul, I’m not very proper.

You could change directory into /usr/portage and run:

find | grep -i xclock

Or you could run something like this:

find /usr/portage | grep -i xclock

And your output should look something like this:



Just make sure you change into the xclock directory before running the deployment script.

Ok, back to it. The first line runs the “ls” command and pipes it through “sed”. This proceeds to filter out unneeded extra things you typically find in the directory that won’t be needed for this process. To that end, we extract the manifest file, any metadata file, and cut off the .ebuild and files from the input that will be shuttled to the array variable.

for index in ${!array[@]};do echo "$index ${array[index]}";done

Ok, so the next line of code is basically a one-liner which is why it has the semicolons. You can put this in a more readable format, but I chose not to because for me it was fairly readable. Just something to keep in mind. We start with a for statement and begin to sort through the array. The nitty-gritty of the “!array[@]” here means that instead of just looping through the array itself and using only the elements in the array (which this code would do shown below)…

array[@]

…we want to go through the slices of the array and use the indices like (0,1,2,3,4).

🔥 Spark Note From the Forge 🔥

Indices? I’m not a programmer!

Don’t sweat it. There’s only one thing to remember: indices start at 0 always. That’s how the system begins its counts. It’s true for all programming, and it’s true here. Here’s one for you LOTR fans: if you think of the “0” as the One Ring, you can think of the rest of the numbers as rings bound to the One Ring which starts at 0.

Geez, what a nerd. Hey, get a room, nerd, with your weird numbers and…stuff!

This is going great. We’re over one of the biggest humps believe it or not. The rest of this line goes through each index number. Then we use “echo” to print out the index number and the array element for each package or element in the array. Simple.

read -p "Enter the pkg to be deployed: " num

If you’ve been reading my other blogs, this will be a familiar one. Skip over it if you have or you know what this is. Otherwise, stay tuned. All this one line does is pauses the script (-p) and takes user input (yours) and then puts that input into a variable (num). Why did I use the variable “num” though? Because the script expects you to put in the index number that was printed.

So if we were picking a new virtualbox to install and I wanted the latest package, I would enter the number “4” into the response prompt and press enter.

pkg=${array[$num]}

If you’ve done programming with arrays before, you’ll recognize this move. If not, no worries; I’ll cover it now. All we’re doing is using the array index to select the correct package from the array and putting only that package name into a new variable called “pkg”.

So if we were speaking about the example above and writing pseudo code, it would look somewhat like this:

pkg=${virtualbox-6.1.4[index number 4]}

Ok, so I feel pretty good about the progress we’ve made, and hopefully, if you didn’t know much about bash arrays or indices, you’re picking up some useful ideas for your next script.

current=($(pwd | sed 's+/usr/portage/++g'| sed 's|\(.*\)/.*|\1|'))

The next line is looking for what directory we’re in by basically using the pwd command, which is great, but dumps a lot of extra junk with it. Thankfully we have sed which allows us to clean things up. So if we’re using the virtualbox example above, we’re going to have a string looking like this from pwd:

/usr/portage/app-emulation/virtualbox

But what we really need is only the:

app-emulation

Enter the sed sphere. We pipe the output of pwd to sed and strip the front two tags. Great. That just leaves:

app-emulation/virtualbox

So how to get that last pesky section off? And that, my friends, is why we pipe in the next sed statement of glop. I’ll be completely honest here: I don’t fully remember what it does other than it will pick off the slash and what is after the slash. The how’s and why’s are on the internet and were in my head when I was working on this diligently.

emerge -av --autounmask-continue --backtrack=200 =$pkg; dispatch-conf;

This should look familiar, right? If not, what’s going on is a very questionable use of the emerge command in a script. Basically I’m running emerge, telling it to be verbose, ask, automatically unmask any package requirements (a.k.a. /etc/portage/package.accept_keywords), and backtrack. Why backtrack? The emerge man page puts it best:

–backtrack=COUNT
Specifies an integer number of times to backtrack if dependency calculation fails due to a conflict or an unsatisfied dependency
(default: ´10´).

I’ll leave it there for that. Next we run the “dispatch-conf” command so that we can automatically accept all changes. Side bar here: this will dump an enormous amount of junk into your package files like /etc/portage/package.accept_keywords. If you are strenuous about keeping the amount of comments down and your files clean then, by all means, you can add some code to the script and keep your package.accept_keywords cleaner than mine. I have another script that cleans up that mess for me, but that’s for another post.

exists=($(find /var/db/pkg | grep $pkg))

This line sets up a scenario for the remainder of the script. Namely, we check to see if the package was installed and we set the variable “exists”.

if [ -d "$exists" ]; then
   echo "Package has been deployed."
   exit
else 
   emerge -v =$pkg
   echo "Package has been deployed."
fi

We are in the home stretch now. We start out with an “if else” statement. We check to see if the directory exists (meaning a successful install), and if it does, we end the script and print a success message. If not, we strip all of the options out and automatically try to emerge the package. This is there because of the dispatch-conf option I called earlier.

Wrap up

That’s it for this post. Feels like a lot of ground got covered in a short amount of time. It’s not the most perfect script, but it does get the job done. I won’t promise you it will work all the time. You will have to contend with certain upgrades in a different way such as python and perl. Additionally, dependency conflicts are generally not going to be solved by this. I’ll cover some of these in other posts down the line. For now, enjoy the deployment script and have fun.

Welcome to Part 2

If you’ve been following along, you’ve gotten all of your offline files ready for deployment. If you missed that section you can go there now. In the sections below, we’ll discuss: the offline deployment, suggestions for running your offline deployment, and finally some fascinating ideas and projects I tried which have interesting potential but, unfortunately, did not work for me in this instance.

Disclaimer

This guide strictly deals with upgrading your system. It will not cover dist-upgrade although that is certainly something you can try and test. This information is provided as-is and, therefore, I take no responsibility for incidents with your equipment. I am a huge proponent of testing. Please ensure you know what you are doing before you attempt this.

Tools you need

1. WinSCP (If you’re using Windows and, for this, it’s almost, almost worth using Windows just to use this awesome, free tool)
2. Two systems. One should be online and the other is, of course, the offline one. They both should be very close build-wise. NOTE: If you want to test this out, I recommend changing the /etc/resolv.conf file on one of the systems. Comment: Remove everything in there and save it. This ensures apt will break without using the correct options and your test is as clean as it’s going to get without introducing USB flash drives.
3. Putty, or if you’re on Linux, SSH
4. Internet connection with both systems on the same network if you’re testing. Otherwise, you’ll just need internet for the online portion.

Copying over the Needed Files to the Offline System

Are you ready to get this done?

We’ll use WinSCP and transfer our files over to the “offline” system in its /tmp/ folder assuming it’s on your network and the only edits you made were to the /etc/resolv.conf file for testing. Otherwise, if you cannot reach your offline system with a network connection, you’ll have to use a flash drive. Mounting a flash drive is out of scope for this post, but the rest of the commands are relevant to your endeavor.

Assuming you’ve connected to your remote system with WinSCP, it’s time to copy some files over. This is where WinSCP shines because it saves so much time. We’re going to specifically copy over the:

1. Archives folder
2. Lists folder
3. Any additional packages (if its a .deb just put it in the archives folder once you’ve noted the full package name including its .deb extension) or scripts you may need.

We’re going to place these folders/files into the /tmp/ directory. Once everything is in the directory, assuming you kept the file names, we can get the actual update process started.

🔥 Spark Note From the Forge 🔥

Be a boss. Tar your files or zip them to make the transfer faster. Why? Because an archive, tar, or zip presents as a single file. The network won’t speed up and slow down as it finishes a file and starts a new one. Instead, as an archive or zip shows up as a single file the network keeps the pedal to the floor the entire way through. Want more mileage? Go full bore and use “xz” compression.

Moving files to the correct directories and cleanup

I kept the file names the same, so starting the in the /tmp/ directory I will run the following:

## Time to make the money. Clean out the archives section first. Make room on the tiny system.
## Move in the new archives and lists and clean up.
cd /tmp/;
apt-get clean;
mv archives/* /var/cache/apt/archives;
rm -rf archives;
rm -rf /var/lib/apt/lists/*;
mv lists/* /var/lib/apt/lists/;
rm -rf lists;

Can this be written more elegantly? Yes. But my intent is to fully show what I’m doing. Do I need semicolons at the end of these? No, but I tend to like to chain my commands together for situations where I can only paste one line into the terminal as it doesn’t make sense to write a script if I’m going to be on and off the system quickly. Believe me, I have found in the world of embedded systems that that happens. Like with Road Side Units (RSU) where you may be doing the same thing with very little variation dozens of times. You may love vi (one of the oldest unix/linux text editors), but me, not so much.

If everything has completed properly then we start pulling the triggers on things.

Running the Offline Update and Upgrades – Finally!

sed -i 's/jessie/stretch/g' /etc/apt/sources.list;
apt-get update --no-download;
apt-get upgrade -yf --no-download --ignore-missing;

For the upgrade, you can try this instead to keep the default options, but I have not had much success with it:

DEBIAN_FRONTEND=noninteractive apt-get upgrade -yf --no-download --ignore-missing -o "Dpkg::Options::=--force-confdef" -o "Dpkg::Options::=--force-confold";

So what does all of this do? If you remember from the previous blog, the sed command changed over “jessie” to “stretch” in one line. You can change the words to be what you need. For instance, you could make it look like this:

sed -i 's/stretch/buster/g' /etc/apt/sources.list;

But for this case, whatever you used on the online system must be used for the offline system. Is it really necessary? For documentation purposes at a minimum, yes. It’s a cheap, short, lazy way to let folks know what the offline system has. Don’t be that guy that leaves other folks wondering.

The next line simply updates the package lists in the system. This is important as we want the package database to be updated with the latest packages for our distributions archives we just moved from /tmp/ to the archives folder.

Once we get to the upgrade line, you’re in the home stretch. The system should be able to begin checking the archive folder for the necessary packages and will begin the upgrade process. I found there was no automation on my side as sometimes I needed glibc to be upgraded, for instance, which brought up a blue screen (updating grub will also bring up a blue screen requesting input). With that in mind, I’d plan to stick around and see things through.

But wait! What about my other offline packages that are not part of the main repository? Like that influxdb package, you pulled down. What happens there? Does apt-get upgrade or apt-get install work for that?

Not in my experience. What I did was throw those packages in the archives folder and then used dpkg to install it like so:

pushd /var/cache/apt/archives;dpkg -i influxdb_1.7.9-1_amd64.deb;popd

This is, again, not the slickest way you can write this, but, hopefully, it gives you an idea for a one-liner install.

Things I Tried that Failed

My main, initial challenge was that I had no internet on the remote boxes I needed to reach. I had a vpn connection and ssh. So naturally I tried to do a reverse ssh proxy which failed for me. I tried using port forwarding with ssh as well, and again, it didn’t come through. Trying to provide internet to a remote box over vpn was making me bang my head!

Time for plan b, and that turned out to be a project called sshuttle. It’s a neat project written in python that performs as a sublevel vpn, dns tunnel and more. It didn’t work out for me, but it might work out for you. Here’s what the project creator had to say on their github:

“Transparent proxy server that works as a poor man’s VPN. Forwards over ssh. Doesn’t require admin. Works with Linux and MacOS. Supports DNS tunneling.”

https://github.com/sshuttle/sshuttle

🔥 Spark Note From the Forge 🔥

What a bummer! I couldn’t get this python project installed right away. I’m so accustomed to working with linux scripts that it didn’t cross my mind that you might need to run a python project with a python installer! This project does need a separate project called “setuptools“.

Inside of that project, you will need to run an initial setup script as well. Another “gotchya” is you will need to match the version of setuptools that the creator of sshuttle is using, but here is a link you can use to get an idea and get started (https://pypi.org/project/setuptools/).

Well how do I install this project offline then? You will need to get all the project files, and then do what you’ve done with the archive files and copy them to /tmp/. Change directory into that file. From there, you can run the following code to install it on the user account you are logged into the remote system with:

python install setup.py –user

Well, so much for sshuttle then. I also gave a project called apt-offline a try. This was promising, but in the end, it also didn’t fulfill my needs. It did, however, set me down the path to understanding the Debian system better. While this project simplifies things for the average user, it still didn’t do everything I wanted. You can check out that project here if you want to give it a try: (https://github.com/rickysarraf/apt-offline).

That wraps it up for this post. Hopefully, you found the droids you were looking for and this has been useful.