Docker squashing is something that has been around for awhile. For those not in the know, when you create a Docker image it layers each of the commands in the Dockerfile into their own filesystem overlay. The aggregate of these overlays are what you containers starts with.
Now, when you have vary large files getting downloaded into the build process of an image, you can find yourself with a very bloated image. Squashing is generally the solution here.
Sometimes I find myself having to do development in corporate environments where you aren't given administrative privileges to the machine at your desk. I know that many would make certain hand gestures at such a situation and walk away. But sometimes the problems that the corporate overlords are paying you to solve is just to interesting to walk away. In that case, you find ways to work with the cards you've been dealt.
For years now I've stuck a Visual Studio Code shortcut in my Startup folder (Win+r -> shell:startup). This way, when I login to my Windows machine, VSCode simply will pop up with my previous workspace state. When you work with multiple different Remote-SSH host windows all within a virtual environment that resets when you logout you need more than just a shortcut.
I've been running this blog since the beginning of 2020 in one manner or another. The largest gap and travesty that I've left unattended is the fact that there is no chat capability built in. About 6 months ago I did try to implement my own thing with a DigitalOcean droplet and a postgres database. It didn't work well and involved nasty iframes in my React app. I just didn't have the time to invest.
With the now fully released Docusaurus 2.x (the engine this blog is currently using), I decided to see if there were any new options for comment integration. With minimal forum reading I came across a GitHub app called Giscus, and it was exactly what I wanted!
Ever find yourself attempting to look up some esoteric part datasheet or source code library that only has documentation in a foreign language that you don't know? I do!
Often you can just drop the PDF into Google Translate and all will be well enough. But if you want more control, more fine grained translating, or just less tracking of your behavor in google, you can do it via Python.
Bottom line, ifconfig has finally reached a point of scarcity that I feel I need to switch to the heir apparent iproute2. So often do I get on an isolated network that doesn't have basic tools like ifconfig, route, or vconfig and because of the environment, my workaround apt install net-tools isn't always available.
When building from source code in Linux (or any *nix) based system, you'll often run into a web of other dependencies that you need to cherry pick from the internet. If you are required to use an older version of a package, its wise to use dependencies that were released just before the source code your are attempting to download. Even if the API for the dependency is still technically compatible, there are also many other factors that can ruin your day.
Static builds, technically linking of binaries in such a way that there is no dynamic loading of external libraries to start the process. A more broad way that I prefer to think of static applications are ones that have no external userspace dependencies. One may argue that my definition of a static application could be a docker container. Not true! A docker container depends on docker (i.e. an external userspace dependency).
Periodically I become a bit obsessed with static building and linking of various tools. These tools can be invaluable with performing troubleshooting of niche systems where you can't depend on the system's ability to accomplish what you need to analyze or debug issues. The catch is that classical unix philosophy has encouraged maximization of shared libraries and dynamic linking. This creates tight couplings between applications and their distributions. Linus has spoken about this coupling and the weaknesses that it brings into the Linux ecosystem.
There are quite a few projects that attempt to mitigate this issues. Let's explore some of these ...
When working with Raspberry Pi, about 99.999% of the information on the internet is focused on Linux support or other higher level interfaces like the RPi.GPIO python package (which depends on Linux itself). When doing, what I'll snootily refer to as, actual embedded development there usually isn't some first class linux driver to do all the dirty work for you. I ran into this very issue when attempting to drive the RPi4 I2C controller from Uboot without an out of the box U-boot driver for the BCM2711.
As part of the Embedded Systems Analysis material development I was recently doing some work with a Raspberry Pi I had sitting on my desk. I noticed that the USB-C power connection had fallen out of the device. Naturally I got under my desk to fetch the squirrel-y cable and plug it back in. I got the cable all the way to the RaspberryPi before I realized that the connector that was supposed to be attached to the RPi was still on the cable. _The whole USB-C assembly had cleanly popped of the RPi4.
In my day to day I'm always working on several projects. While the contributions or time spent on the projects is small, the amount of resources they take up is immense. All to often these days, whenever you want to build a thing from source, you may as well budget at least 20GB for a single project. This of course includes toolchains, dependencies, build cache, and so forth. This causes me a good deal of hardship because I can easily find myself filling up my limited SSD space. Today I decided to take some time to consolidate with the expectation that I was going to buy a HDD for archival purposes. Turns out that wasn't needed for some very strange (new to me) reasons.