It is always easy to blame someone else. I got confused by different examples. First example showed a file structure as locale/${language}/MyApp.po, which failed me when I was trying to integrate it into my program. This in combination with me playing with it cause the CMake generation fail in error basicaly saying The target with the name "MyApp" already exists.. That is because the basename is the same between multiple languages.

Second example is showing a different file folder structure locale/${language}.po. This one worked like a charm, but it looks ugly, I wanted the first variant and was keeping at it for hours to give up in the end.

A short note on my failure to realize that I should debug the CMake itself instead of blindly searching online for hours without a proper solution since there is not that much information online, which means it either others are much smarter than me or that this is underused module / feature in CMake.

Example Configuration

Example of the CMake configuration I used:

Messages.sh

#!/bin/sh

xgettext --keyword=gttr --from-code=utf-8 -p locale -o MyApp.pot `find . -name '*.cpp' -o -name '*.hpp'`

CmakeLists.txt

find_package(Gettext)

if (Gettext_FOUND)
	add_custom_target(
		ExtractTranslations
		ALL
		WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
		COMMENT "Extract translations from the source code."
		COMMAND Messages.sh
	)

	gettext_create_translations(
		locale/MyApp.pot
		ALL
		locale/sk_SK.po
	)
endif()

This year, the biggest sacrifice was most of the Qt5, becuase most of the programs I use switched to Qt6, so I decided to remove Qt5 packages which are not needed, this includes KDE Framworks 5 packages.

All went well and I removed around 200 packages, great outcome from this cleanup session.

KDevelop trouble

As I went to play with code in KDevelop I noteced that the shell integration provided by kdevelop! is not working and throws an error that qdbus binary is missing.

Error: The qdbus (or qdbus-qt5) utility is missing.

KDevelop is native Qt6 program, it should not depends on Qt5 binaries. This error was caused by removal of qt5-tools package, which provides this binary. It is not a compile time dependency and it looks like the Qt version this utility depends on does not realy matter, so everything worked, as long as the Qt5 package was installed.

I decided to look into the issue and checked the bug reports, issues, merge requests and the current / latest code. There was no mention of this problem, so I decided to fix it on my own.

QDbus

the binary for qdbus is named differently in different Linux distributions, that is why the original code is checking for multiple variants. I assume I need to do the same, so I tried to check few distributions on how this binary is named.

• qdbus-qt6 - Fedora
• qdbus6 - Artix, Arch, Manjaro, Ubuntu
• qdbus - General fallback

There were differences in Qt5 and as expected, there are differences in Qt6 as well. In my quick search I did not find any usage of qdbus, but I am including it in case it appears somewhere.

And here I thought that qdbus6 was generaly accepted name for Qt6 version of this tool, luckily I decided to check Fedora packages and found that situation is more varried that I originaly thought.

KDE Cli Tools

Package or repository “kde-cli-tools” provides tools on which the kdevelop! script depends, namely I found kioclinet5 and kde-open5. The situation here is simpler, the KF6 version provides a symlinks to KF5 versions while also providing the basic binaries without numbers. I decided to remove the numbers to not rely on fallbacks which tend to disappear.

Code

My changes can be found on KDevelop Invent page. I included a similar but shorted reasoning in the commit message explaining why I did ehat I did.

Conclusion

I like winter.

It has a dark side as well, after porting to Qt6 the WebInspector settings were not saved into the profile. It was a bummer. I tried to investigate it but came with nothing, I looked at it from a wrong angle and wasted a lot of time. The solution was simple, initialize the QWebEngineView with an active QWebEngineProfile and the settings will be remembered.

And here we come a full circle. After fixing saving and loading of WebInspector preferences by making sure the View (well, the Page which the View is using) is using the correct profile, the feature that it runs even with disabled javascript was broken. So with this knowledge I could construct a workaround and always enable JavaScript for the WebInspector. Other option to consider is to create a new profile specificaly for WebInspector with specific configuration for it. Both solutions are good enough, each with its own pros and cons.

Using the current profile results in:

• There is only one profile on disk.
• Most of the user settings are still respected (local storage, history, cookiesm etc.).
• The required settings (JavaScript) has to be force enabled, and maybe more, thiws has to be found over time.

Using a new profile results in:

• Guard against using this profile in browser itself.
• Different configuration of web profile settings ( data location, images, etc.).

I went with same profile solution, I think it is simpler to bake few exceptions compared to doubling the configuration. The time will tell how good this solution is.

While modern people can use smart phones for almost anything, I came to conclusion that while it is jack of all trades it rarely gets things right. I do not know if it is because I am old fashioned or just too slow to adapt. I am unable to accept using touchscreen devices for longer than few minutes (moments).

The main advantage in using specialized device is for me to have specialized functions such as special screen, better battery life, keyboard with special keys/buttons, and overall environment which allows me to focus on the given task (or just relaxing) without being disturbed by some annoying notifications (like most smart phones do when you use them a bit more). In other words I am able to separate tasks and focus only on them.

As a disadvantage I can only think of having to carry multiple devices when travelling. But with smart phones it is not that different either. In order to get these phones as thin as possible the batteries are relatively small and when you use the device it will quickly use all (or most) of its energy and I am forced to carry some bulky battery (power bank they call it, multiple of them at that). So even with smartphone I still need to take care of multiple devices, they just have a different purpose, which is to keep the main device running.

Reading “books”

I like reading, be it books, blog posts, web novels, documentation and more. To do this I have two e-book readers. One is a small PocketBook Basic 3 Lux (6-inch screen) and the second one is PockedBook InkPad X (10-inch screen).

My first ebook reader was PocketBook Basic. It was “cheap” compared to others, so it won me at the time. It still works, but it is slow compared to newer models. The reason why I changed it to newer model is because the button to switch to the next page is integrated in the plastic body (there is no separate button, the plastic is bended/moves to press the button) and the plastic cracked and cracked more until I decided it is time to get a new one.

PocketBook Basic

Both PocketBook Basic models use button only navigation/control. This is exactly what I desire, because with this setup I can touch the screen to track my thoughts as much as I want. If I make a mess, I can always clean the screen later. I consider touchscreen as a disability (or a disadvantage) for a device.

Due to the display size, the Basic version is only suitable for a text which can be rearranged to the screen size with current font settings. It is not suitable as a PDF reader. It provides me a great undisturbed experience while reading whatever crazy “book” I want.

The official cases are fine, probably not available for older model anymore and the one for the newer model seems to have fragile corners (if they did not fix it). I had issues with official case for Basic 3 Lux but the last one seems to be stronger than the other ones. Could be piece by piece?

PocketBook InkPad X

I bought this device because I desired something to read PDFs (be it specifications or stuff available only in PDF form). Most of the PDF documents are made to be printed on a standard A4 size paper (European guy here).

It has a touchscreen, which I only use for navigating the menu. Yes, the touchscreen is wasted on me. Luckily, there is still someone sensible in PocketBook and there are 4 buttons which can be used to kind of control the device (show menu, next page, previous page, other menu), so I mostly open the book and only press the navigation buttons and I have to be careful to not touch the screen or it will do some wild stuff.

It is a very useful device for reading documentation, just read and forget the world around. It is great to read the whole (or most of) specification when trying to implement something based on it.

The provided case for it is only for a protection while carying the device, which is sad. I resorted to buying a foldable case from AliExpress and it is great, a bit bulky but still awesome.

Overall

Overall, I like PocketBook devices. They provide a great platform to read whatever I throw at them, they work perfectly fine in offline mode, the battery life is great (depends on how much I read, can range from weeks to months).

Future

To make these e-readers even more usable I should at least try to KOReader, which is an alternative reading program for ebook readers. Based on reviews and what other people say about it, I would most probably love to use it.

Listening to music

I like music and I like to listen to music when travelling, specially when travelling by public transport (it is a necessity there), so I got myself a portable music player. After trying multiple low-end ones I went with something a bit better.

My requirements were (still are) as follows:

• 3.5mm Jack headphone connector
• Physical control buttons
• Good sound quality
• RockBox support

I went with HifiWalker H2, which is an awesome music player and I dual-boot the official firmware with RockBox. The main advantage in RockBox for me is a much better volume control (smaller steps) at a low volume. This is especially useful when using earphones where a change in volume could be like “hard to hear” vs “way too much”, with RockBox there are many more steps (Original 0-100 vs RockBox 0-255).

Mobile phone

I use a flip-flop style Nokia phone. It is the newer Nokia, so the design and many other things are subpar, but it is a great phone to sit in a dark cabinet and wait till it is needed (which is rare).

I rarely use mobile phone, so the best ability a mobile phone can have is to stay charged for a long time, so I do not need to care for it.

Ending thoughts

For anything complex I use personal computer, be it desktop of laptop. I love the versatility of it and how I can do all crazy stuff which I would never be able to do with any smartphone today.

This is asking for a part 2, which may come in few years since I had this article on hold for 4 years who knows when that will be.

CHANGELOG

• Reworked the bit offset from the left
  • I had a huge misunderstanding on how this works, so when I was converting it to the offset from right I did not account for the actual bit size which created a huge headache when trying to print the correct bit masks.
• Parse and print static function variables
  • Also print variables defined in lexical blocks (between “{}”)
  • The linkage_name of these variable is often manged, so print it.
• Add option to print the mangled name of the variable
  • This is useful in combination with parting static variables inside functions where apart from a nice name there is also linage_name available which specifies where the variable belongs in the code.
  • So far, I do not plan to write any demangler in the near future.
• Update licensing information
  • The original dwex code is under BSD
  • My code is from now on under GPLv3
• Play a little with README
  • I tried to rewrite it to look more like a real project.

All of the changes can be seen in the git logs.

Thoughts

The script is improving, gaining new features with each usage as I find new bugs or edge case with embedded toolchains.

Short imaginary TODO list is:

• Fix bugs which will popup.
• Refactor the code.
• Add some control for the output
• Print more stuff (I forgot about the reference type, again)

In the release of Falkon 25.04.0 is a feature to show the bookmark or history item address in the statusbar when the mouse hovers over the item (in menu, toolbar, sidebar). I thought I tested everything, but I did not. I did not test the folders in the bookmark toolbar. Before the release, there was a bug report was this exact issue, which I thought I resolved, but I only resolved a part of it.

So after the release was done, the bug was reopened and I managed to fix it fully this time (I hope), but the release was already made. I am not sure how should I approach this, so I pushed the fix and made an announcement on the mailing lists (distributions - for the distribution packagers to apply the patch, falkon - for users to know there is a potential issue. Even with this, there were instances where people and distribution packagers were not aware of the issue. I got a mail that this should also be on the main page as an article about the issue. I thought the mails are enough, since all people who can do something about this should be following one of the mailing lists.

Falkon 24.12

This release went mostly smoothly, but with the comming of Qt6.9 there was a change in the graphic backend which could be used. The implementation of Falkon “Use hardware acceleration” option sets the backend option for QtWebEngine/Chromium backend. This causes Falkon to crash with rendering ŕelated errors.

The fix for this was implemented during the Falkon 24.12 lifecycle but was never cherry picked for this release, because I thought it is not relevant, it will just work (right??). While I had other things to worry about, I totally missed this crucial detail and with a final release of Qt6.9 Falkon 24.12 started crashing when hardware acceleration was enabled.I did not report on this issue anywhere, I only mentioned a possible fix in the bug reports and that is how I found what the issue with Qt6.9 is as well.

This is one of the reasons why I do not like setting chromium options (flags) and rather opt to wait for official Qt API to bring some feature. I do not have will to play with all these flags. Users can set them on their own and thus the responsibility is on them.

Conclusion

So I kind of screwed up two major releases of Falkon and introduced multiple bugs and under reported them. Is this the price to pay for a progress?

Here are the changes and some explanation for them.

• Small optimization in address handling. The script was constantly converting the address from integer to hexadecimal string and back on every operation where address was needed. That is now gone and the conversion is done only at printing.
• Resolve constant addresses. This one is kind of simple and it bothers me how I did not consider such a simple thing.
• Experimental support for bit offsets in structs. I am not sure if I am doing it right here, it seems to work with my examples, but I am still not super convinced especially because I totally ignore the program data structure and endian-ness. Maybe I am here for a world of hurt.
• Maximum array size handling. This one is a nice one. When the array is empty the maximum (all bits are ‘1’) value is set for the array size. The script was than trying to expand this infinite amount of elements and it would either crash all eat all the memory. So, I for not investigated and set a configurable guard to limit the number of array elements.
• Apply data from specification to the DIE. The DIE may not contain all the required information like size or the address, this information is instead stored in a separate DIE made specificaly to specify and add some information to the original DIE. To add support for this, I went with a barbaric way and go over all DIEs in the elf file before running the type resolution on variables. There is a noticable slow down together with the next point.
• Variable “normalization” by address. I call it “normalization” but the real thing is that I filter the DIEs with “0” as an address. The problem which I encountered is that in many CUs (Compile Units) the same variable is defined but the address is either not present and the data from specifications should be used to get it, or the address is zero on purpose, or the address is directly only on one DIE and all other ones are without it. And I did not found a good way to get rid of them all apart from filtering variables with zero address. This adds some time to process once again and slows the thing down. Well, it can also speed it up since the printing is the slowest part of the program.

All of the changes can be seen in the git logs.

Maybe one day I will write a nice proper introduction and description of this little script. There is still a lot of work and feature to do.

I write this little note here because I took the survey from the development mailing list and found these options missing. How am I suppose to answer when the answer is not there and only supported answer is what is probably directly supported by Qt Company itself?

The major point was that the comment field at the last question had to be EMPTY to be able to send the whole survey, if I wrote whatever into the comment field, it would not send the survey since I did NOT fill the comment field.

And the last point on my mind is that I tried to send the mail to 2 different Qt mailing lists “interest” and “development” and got rejecting mail back “only subscribers can send messages here”. I gave up and made this post, enjoy.

Switches

I got the “Akko V3 Cream Blue Pro” tactile switches, they feel nice (for now after few minutes of using them). The switches are hot-swapable which is great for repairing or testing new and different switches.

Construction

The keyboard supports bluetooth, usb and 2.4 GHz radio connection (with a usb dongle). I only care about the usb connection, so anything else is a waste for me. To facilitate the wireless modes the keyboard contains 3000mAh lithium battery. I decided to see how hard it is to remove the battery. There is a guide to change the battery on other Akko keyboard on iFixIt website, I followed it and it was very similar (different keyboards, but same steps and things to worry about).

Akko gets full points from me when it comes to the construction and repairability in the future. While it takes some fighting to open the keyboard and the battery is glued in place, both usb connector and battery are internally connected via cables and connectors (this is awesome). So I disconnected the battery and taken it out, everything works fine and the only difference which I can see right now is that the battery status LED (oh, those LEDs are shiny) blinks each few seconds, before it was shining all the time. I placed a black tape over this misbehaving LED.

Backlight

The supplied key-caps do not have transparent letters, so the backlight is only around the keys. This is a missed opportunity in my eyes. Since I had my old keyboard which had transparent key-caps I swapped them, now it looks great.

Backlight Control

EDIT It turns out there is a keyboard shortcut Fn+\ (between Enter and Shift) which cycles through different colors, but this is not mentioned in the manual for this exact keyboard, but it is mentioned for some of the other keyboards Akko makes. This is very confusing and this keyboard shortcut should be in the manual (or is it a mercy of a programmer who added it and I should be happy it even exist?).

Key map

This mechanical keyboard with hot-swapable switches has ISO layout. It is hard to find a good keyboard with this layout, almost all of them have the ANSI layout.

There is a small drawback to the key layout which cannot be changed in their software, there is not context menu key and the right Super key cannot be remapped in the keyboard to act like a menu key. A sad thing this is.

Conclusion

It is a nice keyboard which is satisfyingly heavy and nice to use, at least after initial configuration and my little modifications.

NNow this keyboard is truly my own.

Falkon part

To implement this in AdBlock (which is an extension), I first modified the main Falkon code to move from the “createWindow” implementation to “newWindowRequested”. I used this chance to add an option to preferences to “Block automatic popups” which are all popups not originating from human interaction (or so I hope).

Next I went and implemented similar (more like same) plugin interface which is used for “acceptNavigationalRequest” function and which is also available for plugins. Having an example is awesome. Also Falkon has multiple APIs for extensions. The C++ is straghtforward normal code, Python I hope will catch it all as well (I truly just hope) and Qml needs special attention since it is manualy made API contrary to Python support which is mostly automatically generated (even when it produces a lot of warnings).

The results can be seen in the merge request 120.

AdBlock part

The AdBlock part of the implementation is relatively simple. I tried to reuse the original code for network interceptor, since it looks and behave the same way and done. Well, it was simple after the needed API was introduced. The implementation passes both Popup and Popup Exception testpages.

The results can be seen in the merge request 121.

Conclusion

As the popups are the most annoying part of advertisements and very important part of adblockers it is great that it is now possible to block them in Falkon as well.

This continues my lurking around Falkon AdBlock code. The number of supported features is slowly increasing (but it gets me a good headache while working on it).

I also updated the Falkon ABP Test Page results.

During the development of the AdBlock stuff I need to frequently interact with AdBlock Dialog window to modify the “Custom Rules” filter list. Since due to historical reasons and probably for the ease of use the dialog uses QTreeWidget and each item added to this list is mapped to the source QVector list (QList now) by an index number stored in the data section of the widget item.

I hope you already see where the source of my frustration and potential issues is (maybe relevant bug?). After the item was removed from the list, the underlying indexes in data section in other items where not updated which let to not deleting rules, deleting incorrect rules, crashing and maybe some other weird stuff.

SO I went to fix it with heavy guns, I wrote a simple model based on QAbstractListModel to handle the mapping between the items and underlying storage container. I chose list model because it is simpler to work with (yes, I lack experience with Qt models) and because I thought QListView would do a better job than the QTreeView. I was wrong, it turned out that QTreeView is much faster to display the data on the screen than the QListView, so I switched to it and I kept the simpler list idea (my head already hurts). to replace the original top level item, I enabled the header and use it for the same purpose. It does not look as nice as the previous version, but I hope it will be more stable and also faster to show.

The code can be found on KDE Invent.

Remove rule

Somehow I was exploring the ABP Test Pages again and when I saw the Remove rule I thought “this can be done” and went to play with it. It is the same as CSS hiding rules with an additional {remove:true} at the end. But it needs another JavaScript to work, so I am happy that the technology progressed and there is querySelectorAll(css) to select elements based on the CSS selectors. With a simple for look to remove all selected elements the deed was done.

This way, Falkon might have support for another rarely used AdBlock rule.

The code can be found on KDE Invent.

Ending thoughts

I like working on things which are rarely used.

I also updated the Falkon ABP Test Page results.

I always thought it works well enough (works for me) and with enough filters it even passes the Adblock Tester test page on 96 from 100 points.

I can talk about “what ifs” and complaining the whole day, but lets jump into the results.

Blocking and Hiding

Remove

Implemented in the Merge Request 119.

Inline CSS

Filter Options - Resource Types

Popup blocking implemented in the Merge Request 121.

Filter Options - Resource Type Exceptions

Filter Options - Constraints

Filter Options - Exceptions

Filter Options - Content Security Policy

Filter Options - Rewrite

Implemented in the Merge Request 116.

Snippets

An initial implementation of snippets can be found in Merge Request 29 and it needs a full review to be merged. I do not want to include complicated code which I do not understand because I will have to maintain it.

Circumvention

Service Worker

WebSocket

While trying to help a user in the IRC channel with an idea on how to make a NoScript-like plugin for Falkon I went over Adblock code to block network traffic and discovered that from Qt 6.4 there is a way to detect and WebSocket connections. Since I already saw it I though “it should not be that hard to add support for it” and it turned out to be a simple task. The already merged Merge Request can by found at invent.kde.org. The implementation also passes the tests at Adblock test pages WebSocket and WebSocket Exception sections.

Rewrite

Rewrite is a feature to redirect to other place. This implementation follows AdblockPlus in that it only redirects to few specific addresses and ignores the rest.

List of supported rewrites:

• blank-js empty JavaScript
• blank-css empty Cascading Style Sheet
• blank-html empty HTML document
• blank-text empty text document
• 1x1-transparent-gif 1x1 transparent GIF image
• 2x2-transparent-png 2x2 transparent PNG image
• 3x2-transparent-png 3x2 transparent PNG image
• 32x32-transparent-png 32x32 transparent PNG image
• blank-mp3 0.1s silent MP3 file
• blank-mp4 0.1s silent blank (black) 240p MP4 file

As a guide I used AdblockPlus Test Pages as well as the AdblockPlus documentation.

Redirects in adblockers

NOTE This section is my own understanding on the topic.

AdblockPlus

The original implementation of rewrite in AdblockPlus was too powerful and allowed to redirect anything to anything, which could been used to inject malicious code into the seemingly protected web browser and steal users data or worse. The implementation was later changed to only allow redirect to specific internal resources listed above. See rewrite at blog.adblockplus.org.

Brave Adblock

Brave Adblock allows more powerful redirect in essence it can redirect anything to anything but this feature is only active in the official Brave block lists. It is used to shrink the browser size and move a lot of their ads blocking scripts to the server to be downloaded only when needed. This all sounds nice and implementing this feature would mean locking it to only their official block lists. See the redirect-url merge request.

The Brave Adblock, at least its network blocking part is also used in other projects such as Qutebrowser, Angelfish and probably more KDE software. Is it worth it to switch to this instead of own Adblock implementation?

Ending thoughts

WebSockets is a great example of a feature which could have been implemented sooner if I was keeping a better eye on the changes and news in Qt library like the main developer of Qutebrowser does. I selfishly use his work for this. He puts the changes for each version in a separate issue, so it is easy to search and browse.

I will continue my random fixes for Falkon.

My reasons

There are multiple reasons why I consider it a great thing:

1. It forces me to think about what I am actually searching for and to try multiple keywords with similar meaning or with something related.
2. While searching I learn much more than I originally intended since I often read more than just a single search result.
3. It reduces my reliance on the internet services which can become inaccessible at any moment for many reasons. These reasons are alo many, is short: dead project, dead domain, changed address, paywalled, firewalled (CloudFlare), censorship, etc.
4. Since I do not have to go on internet and interact with the mess out there it also protects me from nonsense written by modern people (I hope these creatures will not understand how to cripple my docs) or big corporations (AI buzzword nonsense).

In short I am more independent and disconnected from weird world.

Tools and approach

The short list of tools I use to search docs:

• man - Probably the most known program to display manuals
• info - I have no idea what is the point of this program, some stuff is here as well
• wikiman - With downloaded snapshot of ArchWiki, an awesome TUI program
• Various -h, --help program arguments
• Various docs installed from repositories which live in their own docs folders and are in most case html pages. I use Qt-docs like this, connected to my IDE KDevelop.
• Since I often write programs I also like to have examples for the use of some libraries which sometimes can be installed from the repositories and at other times I download the source code or the docs from their website. I try to get some offline version.

I am probably missing some program which would integrate all this together (maybe wikiman when configured properly?). I remember there was at least one, but the one I remember was some crazy GUI in which I got lost and I forgot it. So far I mostly search manually for what I need.

Closing thoughts

Utilizing local manuals is a bit harder than asking modern advertising engine to provide the link to most paying site with enough keywords. Yes, I worded it awkwardly since the search results often bring me to some variant of StackOverflow, distro-wiki page or some weird AI generated garbage.

It requires effort to be independent of modern way of doing things.

Fantasy and dreams

Lets get to my dreamy part. The file format for the executables is called elf and the debug format mostly used with it is called dwarf, my fantasy comes from this little naming joke. I spent a lot of hour reading and rereading parts of the DWARF documentation to come up with a solution and good enough implementation and having dreams is a good motivation to do exactly that.

Implementation

As a language of choice I went with Python and the library pyelftools due to its qualities and portability. A very good starting point, inspiration and a great helping tool I used Dwarf Explorer (DWEX) by Seva Alekseyev. DWEX is an awesome tool to discover and roam around DWARF data, I can recommend it with confidence to anyone who wants to work with DWARF.

I used DWEX as a starting point to get pyelftools to read ELF file and its DWARF data, it provided a great jump at the beginning and I could focus on messing around with following the types and references. The whole thing is a bit of a mess and does multiple passes on the same data (I deem it inefficient) in an attempt to transfer DWARF references into a Python dictionary and than traverse it to make a flat version which I so desire.

Long story short, it was a headache but it works, there are some issues though. The main issue being that it uses a Godlt technique called recursion and will get into an infinite loop while forcing its way through references if the type contains a reference to either itself or different type which contain a reference to it. I intend to solve this issue in the future, but it is present at the moment.

There is one more thing to be aware of when trying to develop a tool like this, and that is that embedded world is a chaotic place and each vendor / company interprets and uses DWARF a tiny bit differently but since they also provide the whole toolchain you will not be aware of this mess until you try to write your own tool like this one. Stable DWARF version is a luxury, each vendor uses different and it is impossible to force their tools to generate a specific version. So the code of this little tool might need to be adjusted to work with multiple DWARF versions which in best case mean supporting multiple field names and in worst reworking whole routines, my headache.

Result

The result can be seen below. The only thing this script does is to print the whole list to stdout and let user handle the rest. It works best when paired with fzf but grep works fine as well. The whole idea was to get the whole name when searching for a member of some inner structure and I believe I managed to achieve this, somehow.

> python main.py ../Playground_C/build/playground_c
0x4030  my_stuff
0x4030  my_stuff.first
0x4030  my_stuff.first.number_1
0x4038  my_stuff.first.number_2
0x4040  my_stuff.second
0x4040  my_small_stuff
0x4040  my_small_stuff.one
0x4040  my_small_stuff.one[0]
...
0x4057  my_small_stuff.one[23]
0x4040  my_small_stuff.structure
0x4040  my_small_stuff.structure.first

Conclusion

The script can be found in my Gitea instance as Miscellaneous/elf_symbols where I also rumble in README about how I am tired of big proprietary tools.

PS: I believe this post is a mess, but well so be it. It is a good introduction to my fantasy world.

PS2: I wanted to write this half a year ago but I could not find a right words to do so but it was still stuck in my head…

Objective

The main objective is to port the add-ons to Qt6 / PySide6 and update version and author signature.

Progress

I managed to port the ToolbarTools and Unloader. Unloader probably can be rewritten in QML for better compatibility. I will leave that for later when I will have will to do that.

And than it had done me properly. I started porting TabCounter and it had kicked me properly (like it should, it is never simple). While porting it to Qt6 was as simple as the two previous add-ons, just changing “pyside2” to “pyside6” in an import statements and done. The problems which stuck with me started when I noticed that Falkon actually crashes at exit when TabCounter is enabled.

Debugging

So I started debugging TabCounter add-on like a pro. I added a lot of print statements all over the exit related code. All seemed to be executing correctly, so I fired QtCreator to get easy to use interface for debugger and after recompiling Falkon for debug symbols I found that it crashes in the general python library. So the problem needs to be in my add-on code, somewhere.

I continued to comment the code and attempted to isolate the code which causes the crash. It turned out the issue is caused by Qt connects to the Falkon code (navigation panel). It looks like the connects are not cleared when add-on is unloaded (is it truly fully unloaded, I wonder now) and I should disconnect the connections manually. Also, due to some reasons I do not know about I put in one connect a python lambda function which had to be removed to fully resolve the crash at exit.

Side effect

During the debugging and trying to blame other code rather than my own for the crashing issue I found a warning in the PyFalkon Shiboken compilation log.

qt.shiboken: (pyfalkon) template baseclass 'QCborStreamReader::StringResult<QByteArray>' of 'QCborStringResultByteArray' is not known

qt.shiboken: (pyfalkon) template baseclass 'QCborStreamReader::StringResult<QString>' of 'QCborStringResultString' is not known

Fix turned out to be a simple new include, so I prepared a Merge Request with potential fix and left it for later.

Conclusion

My energy was spent after finding the cause and I left it for later. The other add-on will be ported soon as well. After porting I will go over the store.falkon.org items and update them accordingly.

Notes

I will probably be adding these kinds of nonsense posts more often, it is fun to write them and share the headache.

Gitea

For a long time I had issues with Gitea being unresponsive in quicker requests. Now in retrospective I think it could be because the Mariadb was not configured properly to accept many connections at once but only the default number. Well, I solved it by changing the database type to SQLite and it runs nicely now.

Jenkins

I removed Jenkins as my back-end builder solution. For my purpose of running website build after I push new commits into the repository once in few weeks or months, it is a huge overkill. On top of that, it required frequent care after each update to work properly with Gitea (plugin updates) which I do not want to manage anymore.

I replaced stuff previously running under Jenkins by simple shell scripts which I start by cron during night. Works nicely.

Virtual Private Server (VPS)

Due to pricing changes it is a bit cheaper for me to get new more powerful VPS for cheaper than I have now. So I was manually moving and setting up stuff on this new VPS for last few days. It is nice to be free from all the dependencies required by Jenkins. I still want to configure some stuff on the server to have more fun, but it can wait for a while.

I also got scraped by Amazon bot and due to that I consider blocking all those bots by IP when it is known from where they operate.

SSH

In the past I used fail2ban to block weird IPs from accessing the server and trying to brute force into it, but due to some updates either in SSH or somewhere in the chain it stopped working properly. So I disabled it and I now use just SSH with some delay after failed attempt to login. Well, it is mostly to block the weird traffic since the SSH daemon is configured to login only with shh keys, no password access is allowed, so brute forcing this is truly a matter of luck.

Blog

The lack of posts is caused by my incompetence. I wanted to write an article about my little script which I made by I did not manage to come up with a good enough draft to turn into a blog post. And other ideas and stuff are also a bit limited and I was low on energy this whole year.

Falkon reports

I did not write the progress reports since I tried to work on stuff instead. Other thing is that Falkon website which is now also based on Hugo with KDE theme is very bad when it comes to long term support. It requires care and with my frequency of posts I am afraid that after each post I will need to update the KDE theme to match Hugo stuff and work properly. On other note, there is a nice collection of commits for Falkon which was merged in this year.

Conclusion

This year is crazy from many aspects and all I can say while not getting violent is a quote from Kirito from Sword Art Online Abridged:

“You know something? I really hate people! They’re selfish, ignorant, loud obnoxious pricks, with basically no redeeming qualities whatsoever. I mean really, look at all they’ve achieved! Genocide, global warming, reality TV, and just a never ending parade of failures and fuck ups! They are, without question, a complete write-off of a species, and how dare you make me care about them!”

Release of Falkon 23.08

Relatively a lot of stuff was changed/added/fixed for this release series.

The biggest thing I struggled with was session restore optimizations which is very slow and I manahed to get it faster by a bit and also introduce new bugs. Well nothing is for free. There is still a second part waiting and that is to delay putting icons into the tabs on the tabbar until they are realy needed. This operation actualy takes most of the time.

The rest are smaller parts and also stuff contributed by people who care about Falkon.

There should also be another release in this series 23.08.5 with more bugfixes. (It is already getting to the point I am fixing my own mistakes.)

Preparations for Falkon 24.02

The biggest change is a port to Qt6 which is possible thanks to people contributing to Falkon (and me having a headache). The Qt6 port is a minimal port, which means that no new features and some features are disabled.

Even the bare port like this takes a lot of energy since the version of KDE ExtraCmakeModules (ECM) was not increased since forever and after bumping it into a proper number it made the code unable to compile due to a lot of string issues. This was a headache since Clazy integration in KDevelop did not work properly so I could not use automatic fix. (yes, I could use the time to learn how to do it manually…)

Another point is that Qt6 is constantly releasing new version which break stuff, so there are some bugreports from not yet released Qt6 which I have no will to test.

Next plans

Version 24.02 is a special one and thus the feature freeze happened much sooner so no new features can be added there, which means the new features will land in some 24.04++++ release.

These are some of the requested stuff made possible thanks to Qt6 implementing it.

• Detection of captive portals
• New way to detect network availability status
• DNS Over Https (or so I think, since it pops in the terminal logs – doh - using system settings)

And probably many more since I did not take a complete look at what is new, the features will come over time.

Site Settings

This is a feature I am working on for a while with some breaks here and there. I plan to finish it within the new year of 2024. I am struggling with some system design and functionality, so after I sort these out I will go beg the KDE GUI team for help to help me design and improve the UI for this feature. Better later than never that I realized this.

PS: This thing touches a lot of essential parts of browser so I think it is important to take more time to properly make it.

SpeedDial

This page sucks a lot of power from the CPU. I have no idea whhy this little webpage is so demanding.

I tried to take a look at it and I slowly started reqriting it and it resulted in me fixinf a single typo. Yes, this is what you get after working most of the weekend on JavaScript code.

This thing is using JQuery and the JQuert Sortable thing to make the speeddial page interactive. I tried to search for a good replacement but I found nothing worth my attention. It is all either part of the BIG framework or abandoned code (mo update for a while).

Support for different devices

I am trying to use Falkon on PinePhone and PineTab2 to find the performance issues and the UI issues on these touch devices. And there is a lot of them since Falkon was never designed with Touch supoort in mind, the touch stuff is not even enabled in the code and I do not want to mess stuff by enabling it before properly testing it.

Fixing tests

All autotests should pass locally and also on GitLab CI, well they do not. It is not a surprise sicne I did not change the tests after changing the code, I totally forgot them. There are some interesting tests which hope that the browser is not fully initialized yet (QmlTabsApiTest) to do their stuff, that is simply not how it is supposed to be done. The problem with this approach is that it creates a seemingly random bug in the middle of the test.

I want to learn more about Falkon autotests and how they are done in order to write some tests for the Site Settings future feature.

PS: PyFalkon tests are broken for many years (they depend on a removed functionality) and I will leave this one for later. It is not shown on the Qt6 GitLab CI since it is missing PySide6 on the server, so it is fine, right?

Conclusion

While I am a bit more familiar with Falkon code-base there are still places I have no idea about, that is why I need to take more time to learn more about it. Also the same goes for Qt of whatever version, I work with Qt mostly only on Falkon and some smaller projects. That is another excuse why the changes take so much time. But is is a good exercise.

Overall, it looks good and I will continue looking over Falkon until someone better appears.

Yes, I am sure I forgot something.

Yes, I did not put any links to this post.

In recent times I played multiple minigames from KDE collection. These games are nice, small and can be played on almost anything these days. But no software can run away from bugs.

A short list of my adventure,

Development

History browser

When filter is not empty, do not delete the whole history under a date if the date entry is selected. This is especially cruel when trying to remove multiple entries and using keyboard shortcut Ctrl+A, Good Bey History. Now the entries unrelated to the filtering phrase should not be deleted. While I also deleted my whole history, this was reported at bugzilla.

Extensions

PyFalkon

Fix crash when working with bookmarks. This was reported on KDE Invent a while ago and I promised to integrate it into next release. Well, I missed few releases and here it is.

AdBlock

Cosmetic updates only.

Add support for the dark theme

Last time I missed AdBlock blocking page. I have few excuses why that is, I did not see this page in a long time and I did not look well enough into the code. There are few more random html files in the code base which I still did not look at.

Properly show the BLOCKED page

This feature was broken due to some Chromium or QtWebEngine security thing which does not allow to redirect a request to a data URL. The user was greeted only with ERR_UNSAFE_REDIRECT chromium page without any knowledge why. This temporary workaround works around this issue (I am already thinking of a bit nicer execution of it).

Clean up default subscription lists

The default subscription list is old and when I went over the links they did not work, so I deleted them. One was working fine, while still seemed outdated at least it could load with different domain (had problem with SSL and domain name).

KWallet

Change to use MAP format was merged. SO far nobody complained to me. I also have a small program which I put together from pieces of Falkon code to export the passwords from old binary based KWallet format. I think of putting it into either separate repository or into “contrib” folder.

Conclusion

Just few small changes which noone will notice or will think as “should be there” feature.

Development

GitLab FreeBSD pipeline

FreeBSD pipeline was failing for a while. It always failed on PyFalkon plugin and in qztools file. It looks like the “latest” FreeBSD something does not like “{}” as default option for QFlags type. To some extend I can understand it, since it runs some code generator and than builds the generated code. And if the code generator fails to resolve everything the compilation might fail.

On second look, it seems that Python support is disabled on FreeBSD package. It is a bit sad but the pipeline is still nice to have.

Custom path of QtWebEngine dictionaries

Falkon will now provide (in preferences) a list of dictionaries located in a folder specified by “QTWEBENGINE_DICTIONARIES_PATH” environmental variable. This functionality is built into QtWebEngine and it only needs to be instructed which language should be used for spellchecking. Now Falkon can do exactly that.

Zoom label in location bar

I thought this one is a good way to explore another part of Falkon code and that is the location bar. The zoom label was a tricky thing to make since I needed to learn how to manually draw it, so that it look kinda nice at least to me.

The functionality:

• Can be enabled in settings
• Is is hidden when the zoom level matches the configured default zoom level
• The zoom level is reset by clicking on the label (described in tooltip)

A bit of inspiration was the one in Firefox, but my skills and artistic spirit are lacking, it is what it is.

State: Waiting for a second look.

Merge requests

Add an option for GPU acceleration

This one got merged, had to do it manually because GitLab was protesting. This a great addition made by Hao Chi Kiang.

Add Flatpak CI/CD Integration

Neelaksh Singh worked hard on adding a Flatpak CI integration. A nice thing for people using this form of a software distribution.

Extensions

I found a bit of time to fix additional extensions Cookies.txt and Readability. It was my mistake, I put a LICENSE file in the top level in the archive which resulted in install error from Falkon, I moved it into the “addon” directory and problem solved. And also as it was suggested to me, I added an entry point entry to the metadata.desktop file so that the older Falkon version can work with it.

RSSFinder

There is a slight change in the design to show a dedicated button for copying the URL and after clicking on the text part the feed will be opened in the new tab.

Conclusion

Thank to all the contributors Falkon is moving forward.

PS: As it goes, I can nicely describe what I work on and I am losing words when describing what others did since I see only the result of their efforts.

My little break

I took a break for a few weeks (turned out to be a bit longer) to get a different look at the features I am working at. (Good excuse, right?)

Development

Site Settings

I am struggling with the UI design and the UI programming itself. I am not that skilled with Qt and its way of doing these things so it will take a bit before I gain the required skills and experience to continue working on it in earnest.

On other hand, while working (for fun) on a other parts of the code I noticed that the PasswordManager is also using host name as an id but for some reason also introduces alternative methods of setting up an id when the general QUrl::host() fails. I should implement this here as well.

Dark design

I was a bit interested and worked a bit to put together a dark variant of Falkon internal pages such as:

• falkon:start
• falkon:restore
• falkon:config
• falkon:about
• tab crash page

The actual color design is just something I put together from colors on my screen and other stuff I thought would “look nice” on my monitor with inaccurate color representation and with RedShift program enabled during the night (Enjoy, lol).

I still do not want to play with chromium flags to add checkbox to prefer dark mode.

Waiting patches

There are few patches waiting for a check and inclusion for which I did not have time or will or mood to do so. Sorry for now, I will take a look soon

Extensions

KWallet

I was interested on how much work would it be to change the way how passwords are stored from the binary form to the map variant to allow users to browse or export Falkon passwords from KWallet without special tools. It turned out as a working solution (since it is based on the original one) and I learnt a thing or two while working on it.

I also almost forget to include the “updated” timestamp feature. I only found after looking at Clazy warnings on the GnomeKeyring implementation. This would be a really weird mistake.

Maintenance

Naturally, after the code was left as it is for a while it needs a top of care to be in top shape again. There are a lot of warnings and potential errors in the code which can be found with automated tools such as Clazy or CppCheck.

On top of that, there is a huge debt in Chromium flags and features support together with actual Falkon UI and usability improvements.

There is a work for a whole lifespan.

Conclusion

It goes well enough.

Development: Site Settings

At the start of the month I did something and I already forgot what it was (very sad) and than I took a “short” break for a week which turned out ot be for a whole month due to other reasons.

Extensions

During this break I worked on two QML extensions Cookies.txt and Readability. The Cookies.txt is in a way me trying to understand how to work with cookies from within QML extension in order to write a proper tutorial for it. The Readability was done on a request and because I thought it can be somehow hacked with QML. This took a bit longer than I expected, but it is nice to have.

Conclusion

I learnt quite a lot this month, just from this point it was a good month.

For ALSA I did not manage to find such a simple way to do it. I also did not find a way to disable it in KDE or Qt configuration files (or I did not search hard enough).

In the end, I resulted in a brute force approach and deleted files and made sure they will not be installed again. So how to do that? I use Artix linux and it is simple.

Open /etc/pacman.conf and add/edit it like this:

NoExtract   = usr/share/sounds/Oxygen*

The sounds are provided by package oxygen-sounds and it updates frequently and is a dependency with a VERSION for other KDE packages. This means I cannot ignore it or mark it as installed since other packages will request it at the specified version (greets KDE Gear).

By deleting the files provided by this package from /usr/share/sounds the sound will be gone and I will have my silence. Now I have to ensure that the package is being installed and updated properly by pacman and the sound files are not provided/extracted by it. This can be achieved by the NoExtract option in pacman.conf as said before.

PS: This is note for myself if I need it again (needed it twice already and always had to do a full research).

Click on button always opens popup

Desired or not the popup will always open. I used it to write there sone informative message, but it is unsued so far. In the future it an be used a bit more to dislay the debug message or a feedback from the script if the conversion fails.

The popup has to be there since without it the button would be without an icon, which is sad.

Communication between webpage JavaScript and the QML

This one is also a fannu one since there are multiple ways how I think this can be done. The communication is needed depending oo how and when the settings are modified and saved.

1. The Global settings can only be edited in preferences

This approach is interesting since it allows to set some defaults in plugin preferences which will be used all the time and any changes in configuration done on the webpage after activating Readability plugin is only for that single use and will be thrown away after the tab is closed or the Readability mode is exited.

2. Provide a “SAVE” button after clicking on the toolbal icon

After clicking on the toolbar button I could run a script to determine if Readability is running on the page and if it is I could try to fetch the current configuration and store it. If not detected the Readability will be executed as normal. Now at the time of writing I am not sure if the settings should be stored after clicking on the toolbar button or on the “Save configuration” button which would popup.

3. Change global configuration after each change on the page

This is the method I chose. It seems very intuitive and simple to use for the user (only from my limited understanding) since the change is interactive and user sees how Readability is configured at the moment. The downside is that QML API does not allow me to simply send these data the way how it was done before on C++ or Python version. In these older versions of this plugin the Readability script is executed in “Application world” which allows it access to the externalJjObject and this way can send data to the plugin code. The QML does not allow this at the moment, and the javascript executed is always in different js world (forgot the name, sorry). This is why I thought of using some obscure methods like I described before. The only way to make the communication running (allowing the JS on webpage to access the Falkon.external object created by QML) is to run the whole thing as UserScript which is executed probably on every single webpage. This is a huge waste, but it is a first working version in QML and sure, it can be improved.

As it can be seen, none of these options are great, but I decided to “ship” something instead of wasting years until it is “perfect”

Javascript code is weird

If yuu are JavaScript developer (a good one) you will notice that my skills in this language are very limited and I have no idea sometimes what I am doing since browsers these days are way too complex monsters.

Ensure that only current page is converted

Welcome to the train of action and reaction. Since I decided to use the userscript way to inject the script to the page I have to ensure that only the currently visible page will be converted and not all of them. The older versions simply ran the JS code on a current window and current tab which QML cannot do with desired JsWorld. So, I went with userscript and some Qt signals. This approach causes it to send a signal to every sinle instance of the userscrip in the browser and I am unable (or I do not know) how to limit it in the QML code. Thus in the JavaScript I added a simple condition which I found on the web (from web documentation, this w3c thing, should be legit) called document.hidden. This should be FALSE if the tab is active. I only tested it with one browser window and it did the job, I have no idea how it works with multiple windows.

Loading assets in QML

According to my knowledge QML does not have a file read interface or any file manipulation things, this should all be handled in some additional C++ code. QML should be just a nice face and all heavy lifting is done in C++. And here we go, the old versions loaded the required assets (scripts, style and images) on the fly and put the final JS together after by itself. This solution is a crude one and wastes a lot of CU power in a long run, but is nice for a development.

The sad part is that Falkon QML API only supports loading text files and not images. This is a bit sad but also not a thing to worry about since the images can be converted to base64 format and added to style while the plugin is being packages. That is exactly what I have done this time. I still left the final gue operation for JS in QML code (it joins the scripts and styles together into one). This should also be done in the packagin process together with minifying the scripts to save a bit extra on user side.

Conclusion

Now that I think about it, I used Readabilty QML rewrite as a good excuse to learn a bit more about the current state of Falkon QMl API and how to work around it in my own way. Playing with mysteiour QML thing together with monster browser and little Makefile for packaging was fun. Later I might write another chapters of my Falkon QML tutorials since I learnt a lot these past few months.

PS: Yeah, it all sounds chaotic.

Enjoy, or scold me.

The extension can be found at Falkon store and code at my Gitea.

This work is partially based on Firefox and Chrome extensions.

The GUI is shamelessly copied from the Firefox extension but I had a little trouble porting it to QML since I have still long way to go with this language. I like the GUI being a simple one and to the point. Either export “All” or the “Current Site”, sounds nice.

The harder part was to get the cookies for the current site since the domain set for a cookie can have a leading dot which probably (yeah I am guessing) mean it should apply for all subdomains it is even more trickier. The Firefox extension is cleverly using Firefox API to get the data and it works fine. The Chrome extension is getting all cookies all the time and filters the current domain on its own.

While Falkon Cookie API supports the filtering based on domain, the thing is, I was unable to get it working reliably with this leading dot and everything and I ended up using the approach from Chrome version. This can make sense since QtWebEngine is based on Chromium so there are some similarities. I did not want to import the whole domain thing as the original does and I only used the “fallback” for it and some hacking to get it somehow working.

The hacking consist of checking if the domain ends with the desired one and other way around. I consider it as a brutal hack since it should not work like this. But what am I to say, it is just a feature meant to easy life of people just a little bit. It is always required to check the results.

On top of everything, this short exercise is stretching my QML knowledge a bit.I found that the QML JavaScript is limited and provides only a handful of features and some are available through Qt module (or object, namespace….). The list of available QML JavaScript thing can be found at Qt docs. Yeah we are still on Qt5.

I wanted to use the URL thing to get the hostname or domain name but that is not available in QML JavaScript and executing JavaScript on the webpage is impossible since that would fail (at least in current state) when JavaScript is disabled, and that is a bit NO to me. I may try to look at the way to enable this to allow extensions to execute JavaScript even when the JS for site is disabled, I think there is a way and the Falkon QML API just hides it from me

For icon I wanted something at least a bit nice, so I went to SVG Repo and took one.

That was me rumbling about my latest achievement and partial fail.

Enjoy, or scold me.

The extension can be found at Falkon store and code at my Gitea.

Development: Site Settings

I continued working on some system for WebAttributes and it looks fine at the moment (well, it just works). The thought which I have is to put all the WebAttributes into one field in the database and parse it afterwords instead of having a separate field for each attribute. I did not try this yet and I will keep this idea for later date when everything will be ready.

Nest I moved to working on Cookie support and because of that started rewriting the SiteSettingsManager class to have sane interface.

Well, I can only blame my lack of preliminary research of the problems and possible solutions for this big rewrite. Instead of writing a good code from get go I change it according to the current needs. So far it works.

Bugs

Right after KDE Gear team released “Falkon 22.08.1” I took another look at the Bug 439268 (Typing text in the search bar sends data although suggestions are disabled). During my first look at this I thought I would need to connect/disconnect signals to this module or to conditionally create and destroy the OpenSearch object, or that I need to update this module. Well, I was silly and I was looking at it in a complicated way. On the second look I noticed that all I need to do is to set the suggestionUrl (or domain, or something) to an empty string when the suggestions are disabled and that fixes the problem. There is some left over code there which was supposed to hide entries (more like not show them to user) from search engine.

Misc

On the other hand I want to put documentation for Falkon to KDE API site and plugin examples and my short tutorials to KDE Develop site. This would make it easier for other to develop extensions and help Falkon to improve.

Conclusion

The development of SiteSettings is going well and there is even a bug fixed. I call this a success.

Any help is welcome and can speed up or improve the quality of this patch.

You can follow the progress of the development on KDE Invent where I created an issue and draft of a merge request.

In the Falkon example QML extension in order to set some basic (top level) element the original author decided to use Rectangle component (or class? or object? or whatever, I will call it component). The Rectangle component is not supposed to be used this way since it is by default also setting the background to white and thus it is not styled by the KDE theme.

This led me to believe that the QML plugins in Falkon cannot be styled by the system theme (colors in KDE systemsettings) since the background was always white even when my theme was set to dark.

Than while stalking the #kde-devel IRC channel I saw that for QML (QtQuick) when used in my window manager setup I need to set an environment variable QT_QUICK_CONTROLS_STYLE=org.kde.desktop to follow KDE settings (maybe there is a way to set this automatically outside of KDE Plasma but I was not successful at that).

Back to original thought.

The Rectangle component is probably not supposed to be used to manage layouts and there has to be better suited component to do that. So far I tried Pane which works exactly as I envisioned it to (it follows system colors set in kde system settings), maybe there is even better suited component than this which would help me even more but so far it is the best result I got from my QML journey.

So my conclusion is to use elements which are styled by Qt by default, but while I was lurking at #kde-devel IRC channel for even longer and asked more and more question I got an answer I half expected but did not like.

WARNING Next paragraph can be wrong and is recollection of how I remember it.

The KDE team recommends using Kirigami or other framework to get the colors from system theme and set them manually. This allow them a great deal of flexibility and it is also used to utilize the additional colors configurable in KDE Plasma System Settings which are not part of general Qt. But for my purpose, I do not want to burden my little extensions with a KDE library just to get some system colors.

PS: This article was postponed for many months for no good reason.

Why?

Sadly, even after project moved under the wings of KDE it did not see many contributors (am I as major stalker at fault?), maybe new contributors did not want to go through hassle of creating KDE account (joining KDE). This is hard to say for me.

Anyway, I wanted to do something what is already outlined and what someone already tried in the past and thus there is a bit of information on how it should be done. That is how I came across “Site Settings” feature. This idea comes from a merge request on Phabricator “Added feature to remember the zoom set on a host” by Anmol Gautam but was not accepted since it was a lot of code which only handles zoom while all data were still stored in the ini file which would make it slower and slower as time goes on.

In the comment David Rosca (The original developer of Falkon) proposed to move the backend to database and add to zoom other settings like javascript, images, html5 permissions etc. and we never saw any changes done after that. Yeah, that was like 4 years ago. I think it rested long enough so I started poking into it.

At first thought

When it comes to designing and programming in Qt and C++ I can be considered a better novice and I make a lot of errors and rewrites, This is exactly one such case. While I could spend days and weeks trying to think of all possible cases and desing an interface and everything in most efficient and nice manner, I lack the experience to do so and instead of that I decided to jump into the water and start programming hoping the code and my thoughts will adjust itself over the time.

And that is how it is still ongoing. I create something thinking it is good enough only for it to be replaced by a new code a little after. Or I did not think of some possibilities or edge cases and designed something in a way that adding this functionality will be a total mess. But that is the only way how I think I can make this work since if I tried to spend weeks “designing” it, it would never be done (like never ever).

The main goal

The aim is to add per site setting for some features to Falkon. There already is per site configuration of HTML5 permissions but this needs to be rewritten and integrated with the new SiteSetting manager. The rest of the features will be new to Falkon. The short list of features I would like to have in SiteSettings is:

• HTML5 permissions
• Zoom level
• Cookies
• Javascript
• Autoload Images
• Other javascript options (Available in Privacy tab in Preferences)
  • Pop up windows
  • Clipboard access
  • Paste from clipboard
  • Activate windows
• And some more located in QtWebEnginePage::SiteOptions enum

The list is pretty broad but all SiteOptions can be grouped together to one logic and I believe this can be done and this should simplify the development a little bit (maybe?).

The other part is to create Graphical User Interface for manipulating these settings. I want to have a default browser-wide settings and persite configuration which would override some parts of it when set to do so.

The current progress

There are already some working parts of this big patch.

HTML5 permissions

This became my first playgruond after getting some backend running, since I believe it is best to get some results quickly. I hooked the current code to SiteSettings and updated current HTML5 permissions dialog to be able to change default configuration. This was a first mock-up interface I did and it works, but it is not good enough so I will replace it with the one I used for SiteInfo window.

Next after I worked on backend a bit more I added a new “Permissions” tab to SiteInfo window with these permissions and created another mock-up interface which looks similar to the one from Firefox and I think I will stick with this one.

QtWebEngine SiteOptions

There is a lot of options in the enum and quite few which I want to add to the SiteSettings system. I have a small example containing Javascript and AutoloadImages options. I also updated StatusBarIcons to work with SiteSettings (but I believe I did terrible job). Hey, the prototype works so I only need to build on top of it.

Cookies

I looked at the code of CookiesManager and it looks like the whitelist and blacklist is stored in a ini file and it should not be that hard to modify it to work with SiteSettings. (Or so I hope.)

Zoom level

This one I think is a bit tricky since it requires special handling and thinking on when to store the new zoom data and when not to and how to work with it. This is my next target fr addition.

##x User interface

As I mentioned before it is a huge work in progress and I hope it will turn out well, otherwise others will be stuck with my little monster which I am totally fine with.

Backend

Because I did not spend a lot of time designing the whole thing I am rewriting and changing SiteSettingsManager class often, to the point that I worry a little bit since each addition of new feature means I need to rewrite or create something some methods and structures. What I need to do though is to have some method to get all permissions which needs to be set while age is loading so that all of them can be fetched at once and other to fetch all of them for SiteInfo GUI. Or maybe just one with everything? I want to avoid using raw SQL commands just to get some data when there should not be need for it, I have a small desire to hide the SQL backend when possible.

Conclusion

I know you wanted to read something nice, but this is all you will get from me.

Any help is welcome and can speed up or improve the quality of this patch.

You can follow the progress of the development on KDE Invent where I created an issue and draft of a merge request.

These feeds are ussualy in HEAD html element but sometimes can appear anywhere on the website. I decided to only include the ones provided in HEAD since this is a standard way of informing browser that there are RSS feeds.

Since I am aware of some characteristics of Falkon and the way how different plugin types work (written in C++, Python and QML) I decided to play with QML even when I do not like this language at all.

My decision was based on these facts:

• There is no need for real time changes, only update when clicked on button.
• To make it portable and working even when Python support is playing tricks on us.
• To learn a little bit about QML development (for Falkon at least)

There is one huge limitation which makes me not like QML support in general, and that is it ignores my Qt widget theme and in Falkon it is all shiny white. The KDE project manages to provide styled QML programs, so I guess there is something either my plugin or Falkon itself is missing. But since I wanted to make it portable and I will use it only once in a while it is good as it is.

The extension can be found at Falkon store and code at my Gitea.

On the end note, this project can be used as an example how to run Javascript code on the webpage and retrieve some information from it while using QML. Well, it is not perfect.

That changes now. This blog is around for over two years and it is time to put some nice numbers on it. At first I struggled about what to actualy count and how to do it within Hugo or if I should prepare the stats outside of it (bit of scripting can never hurt). In the end I managed to do it within Hugo and created somewhat usable template to display the data.

The link to the stats page is located at the main menu and it shows some basic sumary at the top and few bar charts under it. Specificaly the charts shows details of posts per each year, month, week day and usage of tags.

I know it is not that much and it is mostly there to masage my ego, but I made so that I have something to do since the times are as crazy as they are.

That is all for now, I just wanted to brag about the numbers from my little blog.

Intro

### Should be after module-*-restore but before module-*-detect
#load-module module-switch-on-port-available

/usr/share/pipewire/media-session.d/media-session.conf

#default-routes         # restore default route

# Media session config file for PipeWire version "0.3.33" #
#
# Copy and edit this file in /etc/pipewire/media-session.d/
# for systemwide changes or in
# ~/.config/pipewire/media-session.d/ for local changes.

context.properties = {
    # Properties to configure the session and some
    # modules.
    #mem.mlock-all = false
    #support.dbus  = true
    #log.level     = 2
    #alsa.seq.name  = Midi-Bridge
}

context.spa-libs = {
    # Mapping from factory name to library.
    api.bluez5.*    = bluez5/libspa-bluez5
    api.alsa.*      = alsa/libspa-alsa
    api.v4l2.*      = v4l2/libspa-v4l2
    api.libcamera.* = libcamera/libspa-libcamera
}

context.modules = [
    #{   name = <module-name>
    #    [ args = { <key> = <value> ... } ]
    #    [ flags = [ [ ifexists ] [ nofail ] ]
    #}
    #
    # Loads a module with the given parameters.
    # If ifexists is given, the module is ignored when it is not found.
    # If nofail is given, module initialization failures are ignored.
    #
    # Uses RTKit to boost the data thread priority.
    {   name = libpipewire-module-rtkit
        args = {
            #nice.level   = -11
            #rt.prio      = 88
            #rt.time.soft = 2000000
            #rt.time.hard = 2000000
        }
        flags = [ ifexists nofail ]
    }

    # The native communication protocol.
    {   name = libpipewire-module-protocol-native }

    # Allows creating nodes that run in the context of the
    # client. Is used by all clients that want to provide
    # data to PipeWire.
    {   name = libpipewire-module-client-node }

    # Allows creating devices that run in the context of the
    # client. Is used by the session manager.
    {   name = libpipewire-module-client-device }

    # Makes a factory for wrapping nodes in an adapter with a
    # converter and resampler.
    {   name = libpipewire-module-adapter }

    # Allows applications to create metadata objects. It creates
    # a factory for Metadata objects.
    {   name = libpipewire-module-metadata }

    # Provides factories to make session manager objects.
    {   name = libpipewire-module-session-manager }
]

session.modules = {
    # These are the modules that are enabled when a file with
    # the key name is found in the media-session.d config directory.
    # the default bundle is always enabled.

    default = [
        flatpak                 # manages flatpak access
        portal                  # manage portal permissions
        v4l2                    # video for linux udev detection
        #libcamera              # libcamera udev detection
        suspend-node            # suspend inactive nodes
        policy-node             # configure and link nodes
        #metadata               # export metadata API
        #default-nodes          # restore default nodes
        #default-profile        # restore default profiles
        #default-routes         # restore default route
        #streams-follow-default # move streams when default changes
        #alsa-seq               # alsa seq midi support
        #alsa-monitor           # alsa udev detection
        #bluez5                 # bluetooth support
        #bluez5-autoswitch      # automatic bluetooth HSP/HFP profile switch
        #restore-stream         # restore stream settings
        #logind                 # systemd-logind seat support
    ]
    with-audio = [
        metadata
        default-nodes
        default-profile
#        default-routes
        alsa-seq
        alsa-monitor
    ]
    with-alsa = [
        with-audio
    ]
    with-jack = [
        with-audio
    ]
    with-pulseaudio = [
        with-audio
        bluez5
        bluez5-autoswitch
        logind
        restore-stream
        streams-follow-default
    ]
}

Most of the things I search on the web are related to the similar things over and over againg. So the results I get are almost always to the same handful of websites over and over again.

Since my interest often lies in computers, programming and linux I truly get just handful of sites for my searches. A short example can be StachExchange, StackOverflow, language manuals or documentation, Archwiki pages and from time to time some other site (but this is very rare).

So I should start thinking a bit more about the topic before I put my question to the general search engine and rather use the correct site or information source (like local documentation) instead of these general search engines.

By looking at my blog from the perspective of the modern web it looked minimal and nice, but on the inside the theme I decided to base my blog on was using pretty modern styles a features which I had to get rid of.

The first thing I noticed was that my color scheme was not respected and I thought (mistakenly) that Netsurf is not being able to even color the background. But in the closer look my styles were using at some parts modern CSS3 (I guess) variables to set theme colors.
Netsurf does not know such a thing for style. My solution was simple, to move the color scheme style generation to the server side and thus I rewritten it to use SASL which provides same features regarding my use of variables as the previous modern solution.

AT this point I called it done and stopped working on it, since it looked kind of usable. But who would have thought it would stay in my head and thus I had gone on another journey of backward web development.

The other big point was to get rid of modern CSS options and replace them with older and supported ones and also while at it make it look much nicer in w3m since it is nice terminal “web browser”. There was only one feature used from modern CSS and that was display: flex; and its other options which are unsupported but makes it really easy to make webpage look nice with whatever elements. I somehow managed to replace most of the flex elements part with either float: left/right; and by NOT using div for everything but instead instead using span or removing it and reworking the part to look the same. I still kept the flex display option for the mobile view since it makes it just so easier for me to make (otherwise I would not even bother with mobile view, who is even using such evil devices to vie my blog anyway?).

These changes make it look almost same in modern web browsers and Netsurf also caught with them (round corners are not supported at the moment). The bonus of reworking it and getting rid of many divs is that menu and links, tags are now on the same line in w3m, while before each link, tag was on its separate line. This makes it look better in my eyes.

I also played with my video template and the alternative text is now nicely formatted and it even displays the file size of the video.

All in all it is a change which only few will ever notice (since I believe there is nobody apart from me who reads this blog.

Later Archlinux pushed PipeWire as hard dependency of few packages I had installed (namely kwin and PulseEffects). I can remove kwin safely since I do not use KDE/Plasma and had it for testing and as a backup, but it is really a pity that PulseEffects would be gone since it is great program to play with audio effects on the fly. So switching back to pulse would get rid me of these nice programs (unless I hack the packages which is another work, so why to bother?). The other huge issue with PipeWire is that it has some problems with playing audio.
When I start playing (or seek) some video it plays noise for a while until it catches up with something which is pain since PipeWire was advertised to me as a drop-in replacement for PA but it sucks even more.

So I decide to get rid of all that mess and use only ALSA. The basic switch was easy and almost none configuration was required. But I encountered 2 problems about which I wish to write about.

PulseAudio allows in its mixer to disable system sound which works nicely but ALSA does not have such feature (as long as I am aware of) and I have to disable sound effects for everything manually. In graphical session I use mostly Qt based programs and there was an annoying sound effect played whenever some pop-up window showed up. I searched hard and I was unable to find a solution on the internet and I found that Qt itself does not contain these audio effects and that they are part of some theme (or style), in my case it was part of Oxygen (maybe my configuration is somehow set up to use oxygen and I am not aware of that). After removing Oxygen from my system the sound effects are gone and silence returned.

All this I found after going almost mad and catching straws on the internet. My last straw which allowed me to fix this problem was that it should load “*.ogg” file so I started searching for such files and after few tries it bore fruits.

The search command I used is:

find /usr -iname "*\.ogg"

Second thing was default audio support in Wine which can be easily solved by running winetricks for given wine prefix.

winetricks settings sound=alsa

SystemD in my eyes is a tool developed by corporation (RedHat) for other corporations (their customers). RedHat guys (and other kinds) have many things (tools, configs, manuals, trainings, webinars…) prepared for it and thus it would be pain for them to stop using it.

Other corporations like it because it has support from other corporation which they can call when there is a problem and it is mostly one package which contains almost all basic stuff they will need. I see this as major selling point for it since huge corporations do not wish to play with small things, for them it is better to spend more money and buy something. In other words, it can give a lot of people work and thus a good livelihood.

Now when I look at it from the point of normal user like myself I do not see those benefits which might come into account if I was a huge corporation. All I see is a needlessly overcomplicated huge mess given (forced on) to me for testing purpose, so that the major bugs can be solved before it is used by huge corporations.

Next I see it as perfect example of EEE strategy (Embrace - Extend - Extinguish).
https://systemd-free.artixlinux.org/img/systemd-devours-all.gif

I dare to say we are at Extinguish phase since other projects which were providing similar functionality (see image above) to some parts of systemd are simple gone (or part of systemd now).

I also see it as a “killer” of a bit chaotic environment in Linux/open source world. I like the chaotic (modern word would be diverse) environment of Linux open-source world while systemd tries to bring some uniformity into this “chaos”.

PS: If you think systemd is just an init you are not thinking hard enough. Systemd is (as mentioned in some presentation about systemd I watched somewhere a long time ago) a system between kernel and user programs. (lets think of it in similar way Apple is doing things, you do not directly contact kernel or HW, you use their libraries to do the job you want).

PS2: Maybe the bar is too high up and thus people expect for everything to work in same way without any effort.

Enough of my rumbling.

Have a nice day.

This post is a bit refined version of my rumbling on Artix forum: https://forum.artixlinux.org/index.php/topic,1939.msg16354.html#msg16354

PS3: While polishing this post I wanted to write a bit more but it looks like the stars are not in my favour to do that.

This is all nice and great if you are using some more complicated music player with library function which can create index of the songs contained in such files and play them accordingly from its UI. But as a believer that GUI is bloated I am using Music on Console player which does not have this function and is strictly reading files (so I think). So I needed to split this huge file and extract each song with correct tags.

So I searched internet and even written a script to automate this (a monster to say). I was too lazy and did not check manual at all just copy pasted commands from different places and glued it together. To a script which first created WAV files and than converted it to FLAC with tags. This conversion took a long time while my machine was under a huge load, truly a waste. But I was happy that it works.

Now a smarter me read a manual and found out that the utility which is used to split the wave (flac) file shntool has an option to configure the output format which works flawlessly and is very fast. It does not include tags though, and as I am writing this channel I found a very good article on Arch Wiki which describes exactly what I wish to do.

So instead of boasting how good I am at scripting and messing around as I originally intended I ended up learning to revise my old scripts after reading the manual and reading wiki. What a night.

EDIT: After being happy for a while I ran into a wall which consist of shntool being unable to process big files. So I did another bit of research and found utility unflac which is able to do the job a bit better. The code can be found here.

The official docker image works fine, the only thing I needed to do was to add ruby gem local cache so it would not need to download all that mess over and over again (what a hell that is). This setup worked fine until for some reason docker started crashing from time to time.
It somehow was always down after some time when I wanted to create some post. And since this is the only thing using docker on my VPS I decided to throw it away and once again jump into the rabbit hole of internet to find a good solution to replace it (I am too lazy to craft one from the ground up).

I did a bit of searching here and there and I also had some memories from the time I was first setting my blog and I decided to go with Hugo. Hugo is written in go and which I am not a big fan (dependencies listed as urls scares me) but for me as end user it is great since the thing I need to use/get is only one binary. No need for docker and for gem (libraries) cache.

During my research of Hugo I discovered that themes and templates are done in a bit different way than in Jekyll so if I wanted to keep my theme from Jekyll I would need to rewrite it on my own (which is pain) so I decided to dive into another rabbit hole of design and search for the “good” theme. Lucky me I found one good looking (to me that is) called Hugo Dusk which I took and modified to suit my weird taste and needs. I think it looks even better than the previous one.

With Jekyll I was using few include templates to make writing posts easier like inputing images and videos. Luckily Hugo provides similar functionality with its “shortcodes” which already contains one for images (figure) but lacked one for video (I do not count youtube and vimeo) so I crafted one by myself and everything works just fine.

Here and there I still play with design a bit, adjusting details might take a while.

To sum it up, I got rid of docker and ruby. I added Hugo and changed the theme of my blog (it was a good opportunity to do so). Overall it was a good move, I still have a lot to learn with Hugo to make a better use of it and for now I am fully satisfied with it.

That is all, you can find the source codes at the website repository

But before I have to mention that Pacman prints the “Installed size” value when we try to install packages. So one way is to try to install all packages but this can also include dependencies and may require some pacman magic which I am not aware of. I forgot to mention, this method may require root access or sudo privileges which is not good enough for me.

There was also one more brute-force idea which consisted of downloading of all packages and checking their uncompressed sizes and counting this value somehow.This is way too brutal.

The basic idea behind this script is based on the fact that pacman prints the installed size value when we run pacman -Si "package"and as such it can be automated with bit of skill and luck.

$ pacman -Si pacman
Repository      : system
Name            : pacman
Version         : 5.2.2-1
Description     : A library-based package manager with dependency support
Architecture    : x86_64
URL             : https://www.archlinux.org/pacman/
Licenses        : GPL
Groups          : base-devel
Provides        : libalpm.so=12-64
Depends On      : bash  glibc  libarchive  curl  gpgme  artix-mirrorlist  artix-keyring  archlinux-mirrorlist  archlinux-keyring
Optional Deps   : perl-locale-gettext: translation support in makepkg-template
Conflicts With  : None
Replaces        : None
Download Size   : 838,79 KiB
Installed Size  : 4538,70 KiB
Packager        : Artix Build Bot <jenkins@artixlinux.org>
Build Date      : Št  2. júl 2020, 00:28:07
Validated By    : MD5 Sum  SHA-256 Sum  Signature

So all we need to do is get this information out of this output. This can be done with few commands or some smart awk or whatever you are good at using. The tricky part is to get the actual size because it has different prefixes we need a bit smarter tool to count formatted sizes like numfmt (this tool is truly great and I am happy it was recommended to me after I use awk script which I stole from something else) or some smart awk.

After adding a bit of counting magic and formating the result into human readable format I thought “I am done” (how foolish of me).
During the creation of this script I did not realize i was working with my local locale and thus the number formatting was totally messed up and I reflected this mess with few sed replacements in the script which made a mess when run in universal locale LC_ALL=C.

Well, I do not wish to write about every little change which I made and everything which was done, if you wish for a bit more detail you can only regret since you probably do not have logs from that channel.

So here is the final script, which is a bit of shell magic and I know that it can be improved, but it is good enough to run it from time to time and just for fun.

$ export LC_ALL=C && pacman -Sl system | cut -d" " -f2 | xargs pacman -Si | grep -i "installed size" | cut -d: -f2 | sed "s/\s//;s/ //;s/B//;s/i//" | numfmt --from=iec | paste -sd+ - | bc | numfmt --to=iec-i -- && export LC_ALL=

Example output:

2.9Gi

This script as it is written here will show the installed size of all packages in Artix system repository. It can be adapted to show some other package group or repository.

The result is nice and it works just fine, for me that is. As a bonus I learnt some nice utilities a way to use them, for me this was a win.

That is enough of my rumbling.

• New domain name.
• Self-hosted on VPS which is powered by Artix.
• Own git server powered by Gitea.
• Openring integration which provides links to articles from blogs I follow.

For the future I plan some things, but the current setup is working fine so I have no idea when it will be done. The things I plan to do in the future are:

• Move to different static site generator
• Change the html template to something of my own creation
• Prepare some nice landing page at my top level domain, which is for now empty
• And probably many more but I will be glad to make even one of these work.

For years I thought of configuring my computer to suspend at around night time so I will be somehow forced to go to sleep. Well, the main problem with this was that I only thought about it and never actually did any research on this subject.

At normal times under my normal user I suspend my machine with elogind loginctl command and this works fine but there might be a problem with sessions and stuff around that which I refuse to work with. So I searched a bit and found out that there is nice system API to manage power state.

The system/kernel API consist of a file located at /sys/power/state.
When I try to print the content of this file it shows the currently available states (or values it accepts) which is my case are:

$ cat /sys/power/state
freeze mem disk

And when I looked at the printed values I saw mem option which is probably what I want (I was lazy and did not look at any docs) to suspend my machine into RAM. The power state can be changed by writing desired state into /sys/power/state file.

$ echo "mem" > /sys/power/state

Sure, you need to be root to do so, but so what? This is my machine and I can. Since I figured this out I put it into a file, set it executable and configured cron to start it at 23:00 from Sunday to Thursday like this.

0 23 * * 0-4 root /usr/scripts/suspend

And done, now I hope it can help me get my desired rest.

The script as I made it only works with files exported by the GUI, when you export whole scheme and save it somewhere. I don’t think it will work with only data for one file type (it can be easily modified to do so). So, if you try to use your full kateschema/highlighting file which stores multiple themes at once the script can “export” a random mix (I wonder how usable that would be, maybe we can get some funny results).

The interface/command line options part of the script was done at the last moment when it was actually working (for me that is) and I was deadly tired. Before you start open the script in some text editor and look at the top for “settings” dictionary and its values, which I tried to briefly explain there. There are two configuration options available:

1. prefferStandAloneData
2. IncludeCustoms

prefferStandAloneData forces script to use settings stored for that data type. The reasoning for this option is simple, in old file format you could specify color/style settings for one data type differently based on current file type. In other words your CSS file content could be styled in different way when opened standalone in editor, different when viewed in HTML (inside <style></style>), different in PHP and so on. This is one of the reason for moving to theme files, but it is pain to convert. SO select what you wish or design a better solution. A sad thing is that you will not see the generated values as defaults, but I can live with that.

IncludeCustoms allows script to attempt (I am only aware of one tester which is me) to convert the style for all data types/syntax highlighting settings available. When set to False it will generate simple theme file without any custom things, it is nice base to start adapting your current theme which you spent years using.

And start the program:

script.py ThemeName mytheme.kateschema

The script requires 2 arguments.

1. ThemeName - specify how you wish to name the theme, it is so far only used for metadata
2. kateschema file - path to your exported kateschema file.

The results will be printed to stdout. I was too tired and this was the simplest way I could think of. You can always redirect the output to the file with your shell:

script.py ThemeName mytheme.kateschema > new.theme

And in the end, import the theme to Kate/KDevelop and enjoy.

For brave one here are my suggestions on improvements.

• The theme name could be used to select right highlighting setting from global kateschema file, probably by just adding check if string ends with the name. The INI file section for highlighting are names as “Highlighting language - Schema MySchemaName” and such I think this.
• Some user interface could be made to select theme to export from global file (in my thoughts this is done when you export it from Kate/KDevelop directly). It can be useful in the future
• There re probably many checks which I did not do in the code. It is probably not possible to convert files for single data type/language.

In short:

1. Export your theme
2. Open script in text editor and adjust settings to your needs
3. Run the script and save the output. (the result is printed to stdout)
4. Import theme to Kate/KDevelop
5. Use the script as an example and create even better one.

And I almost forgot, the script can be either found at my repository or in KDE Syntax Highlighting repository.

With the latest changes in ksyntaxhighlighting library the old themes (they are called “kateschema”) will become obsolete and one will have to either use some new theme and hope it will fit or try to adapt the old one and spent hours doing so.

The old format is a bit complicated, the settings for one language can vary depending on the current file type. In other words you could have different settings if you opened CSS file directly or when styles were used in HTML. This looks like a nice feature to have which allowed users to tweak everything but also added a huge overhead when doing changes to the theme. So one has to decide how to export these changes, I decided to take 2 approaches:

• First come first serve - whatever setting is read first for given data type, that is used
• main type rules - In other words it would prefer settings stored for that standalone data type and will overwrite any previously read ones

Another huge drawback while writing this script was lack of documentation which forced me to go through the source code to find answers to my questions, which took a lot of time and energy. A huge part of the script is comprised of dictionaries which are used to convert old settings to new one.

I also got a bit of help from the developer Christoph Cullmann on reddit.
There you can see how big spammer I actually can be, at least I managed to somehow finish it.

In the end I managed to convert my theme with some drawback compared to old solution. In new version you cant define default style for something by using the theme. These defaults are set by syntax highlighting definitions. So the “default” values are almost always not met and will restore setting to some theme keyword defined in syntax file (well, nothing to do here from my side).

The script was than pushed to the frameworks/syntax-highlighting repository.

I was too tired to do some proper user interface and such it is very crude. In general you need to supply 2 arguments:

1. Theme name
2. kateschema file

And the results will be printed to the stdout, it is up to user to save it into the file or whatever.

In the end, the script did the job for me and I can have same colors (for time being until I fix or change the theme) as I have on old one.
I use it to convert exported theme with all the mess around (16k lines) and I got a nice theme as a result together with messy custom styles. But there is nothing to do with that, I did not wish to lay with theme for hours to make it looks nice again.

So these are my messy thoughts about this script, it works but it is not perfect. It can be used as an example for your own (hope better) creation.

After a month at new work I realized I need a good way to manage my tasks and things todo lists. Managing things manually is not exactly what I wanted to do and even through the main operating system at work is M$ Windows I decided to try my luck and compile taskwarrior within Cygwin which is luckily present.

After going through a bit of problems due to my lack of knowledge of how to do things under M$ Windows and to top it of I also added timewarrior. My primary reason for adding timewarrior is because I like graphs. Seeing colorred graphs representing my work is the best way how to end the month.

My workflow

I tried to use taskwarrior for personal tasks multiple times in the past but I did not managed to stick with it. Now as I started to use it at work I see the true potential and flexibility of this program.

At first I simply put in the main (read huge) task which should be completed and as I progressed I put annotations to it and created some small tasks around to help me switch focus a bit. The downfall of this strategy is that I ended up having this one task in my task list for a long time and as I progressed the number of annotations also increased and it was a mess.

It was a bit later as I again went through taskwarrior website and read the “Best practice” recommendations and realized how stupid I was. Now I create small tasks which can be cmpleted in a short period of time. I assign them to project according to my main task and add some tags to represent nature of what it is.

I am also slowly learning how to use other commands to manage “time” with taskwarrior like wait, until, some reccuring things and so on (go read manual).

Final thought

I am truly amazed that such a flexible program was laying in my computer for years unused.

Bonus

I truly love how timewarrior can be set to automatically exclude certain time ranges so I do not need to even move a finger to get a nice looking graphs and summary from it.

I bought an SSD to replace classic slow HDD on home M$ Windows 10 machine. This SSD is a smaller than the original HDD (250 GB vs 500 GB) so I could not clone whole disk which I hope would be easier (anyway I will probably never know).

Initial research

Before I started I decided to check internet (DuckDuckGo) and asked how to migrate M$ Windows 10 from HDD to SSD. Sadly all I got and wasted time on were articles advertising commercial product (or features of commercial version) as free product and nothing constructive which would help me. (Such a costly system and it does not even has a migration tool.)

My second thought was to use CloneZilla as it is great tool for cloning driver in general.

Cloning

Before starting I prepared partitions on SSD in the same way as they were on the original drive apart from the one for data.

I cloned all partitions one by one, started M$ Windows rescue medium (or something of that sort) to fix the bootloader (which worked after few tries) and was happy to get to M$ Windows loading screen but than I got only blackscreen with working mouse cursor (my thoughts at the time were as follow: wtf? You can find the cursor so show me the rest!). So I searched how to fix this (maybe error during cloning, or some M$ Windows thing) but nothing helped solve my problem.

Solution

I ran out of patiance for this crazy product and downloaded installation media from the website, “burn” it on USB and installed new “clean” system on SSD. After doing so I had to code the data and install programs and set few things (it is not very customized machine). This also took a while.

Conclusion

It is probably me who was unable to find the “right” manual for migration and also me who does not understand the essence of M$ Windows.

Final thought

Commercial products suck.

I continued to slowly work o Falkon and also resumed work on Falkon Plugins and once again started to fill Falkon store.

I am also slowly investigating / researching Falkon Qml API which should work out of box on all platforms (while some may have problems with Python), the API is insufficient, in other words “there are too many restrictions”. I document my progress at file which can be found in _drafts folder of this repository.

Falkon contribution

• Speeddial can now properly work with Unicode characters 406095
• SearchBar - Show number of matches found on page search 395429

On a side note, I am preparing a VPS with nice domain to move this blog and many other things on. My domain and VPS are up just for few weeks and I already have too many bots to kick out.

Hello, while using Vertical Tabs extensions in Tree Style mode I developed a habit of keeping tabs and only unloading them. This leads Falkon to have hundreds and hundreds of tabs always prepared for me to load.

In face of this I became curious about how many tabs there actually are and how many are un/loaded so I developed Tab Counter extension.

Let me show you it’s features!

Toolbar Button, actually Label

The tab count is shown in a toolbar button, which can be moved where user desires is to be.

Settings window

There is also a settings window in which user can change which count is actually displayed.

There are following options:

• All
• Loaded
• Unloaded
• Loaded/All

How it works

The counts are updated whenever the tabs are changed. This includes: added, removed, moved, loaded, unloaded.

Conclusion

This is a great extension for people who like me keep a huge number of tabs (and use tabs instead of bookmarks) to see how many tabs there are.

This was too short but also a very long month. Many things I wanted to do I was unable to and had to push it to the near future.

Artix Linux

• To make information more accessible we prepared RSS feed for our news. You can subscribe at https://artixlinux.org/feed.php

Falkon

• After a long while the bug 409575 (Python plugins don’t load with PySide 5.13) has finally a working with Qt (PySide2) 5.14.1.
• GreaseMonkey can now grab userscript URLs (*.user.js) with newer Qt versions. 6974f8c2
• Falkon can now handle web notifications (since QtWebEngine 5.13) with it’s notification system 409045

Other

• I helped to hunt very annoying bug in ktexteditor 413848 (Drag&copy function (by using Ctrl Key) does not work anymore)

This month was full of work and I did not have much time to do things which I would love to do.

KDevelop

Some patches I made (!66, !68, !75) were accepted and will be part of the KDevelop 5.5.

Falkon

We are still walking on edge when it comes to Python plugin support. The problem is there is probably “only” me who is testing the patch for real to ensure the patch is correct, but since there was no time…

In this way I would like to ask anyone to help with solving this issue.

On the other end, at the end of this month (year) I finally found a bit of time and tried to port Falkon code away from deprecated things. If there is any problem please create report at bug tracker

Conclusion

That is all what I deem worthy of praise.

Happy new year and happy coding.

Few days ago I wanted to play with Falkon and create some basic Qml plugin. (What else to do to play, Python support is broken and C++ is for built-in plugins).

I started KDevelop (the big IDE I use) and went to create new Empty project because as I am not big Qml lover I did not create template for it yet. But I was met with errors which made me change my focus and fix the tool I use (KDevelop) before playing with Qml and Falkon.

Man must shape his tools lest they shape him. - Arthur Miller

Problem

During the project creation AppWizard is trying add files and than commit changes into new DVCS (Decentralized Version Control System) repository. When there are no files to add there will naturally be nothing to commit which results in error message from DVCS and the project creation is stopped and project removed.

Solution

Lets dive into KDevelop code and check what can be done about it. My humble solution relays on checking the state of the DVCS repository before commit. To achieve this I am trying to get actual status of the repository from VCS (Version Control System) and then checking if the result contains some items.

The KDevPlatform provides nice API to work with VCSs and thus I only need to call few method to get all the data.

Actual code is truly simple and works for me. (only my changes)

qCDebug(PLUGIN_APPWIZARD) << "Checking for valid files in the DVCS repository:" << dest;
job = dvcs->status({dest}, KDevelop::IBasicVersionControl::Recursive);
if (!job || !job->exec() || job->status() != VcsJob::JobSucceeded)
{
    vcsError(i18n("Could not check status of the DVCS repository"), scratchArea, dest);
    return false;
}

if (job->fetchResults().toList().isEmpty())
{
    qCDebug(PLUGIN_APPWIZARD) << "No files to add, skipping commit in the DVCS repository:" << dest;
    return true;
}

Brief description

• Print notice into debug output
• Prepare job with status command to run
• Execute the job and check for failure
  • If failed put notice in debug output
• Check if the response contains some items
  • If not put notice in debug output
  • Skip add files & commit by returning true
• Test many times and done.

Results

KDevelop is able to create Empty project from template with initialized DVCS repository which is really helpful when one wish to start a project from scratch.

Git

Works just fine.

Bazaar

The world is never perfect and there is always a room for improvements.

Links

The merge request is waiting for review at invent.kde.org !73.

So I searched a way how to disable it, but as almost always there was none. In the end I used “Feedback” function and described how unusable it is on desktop and now (at the time of writing this post) I have older homepage design.

Meanwhile

While YouTube was playing tricks on me I decided to play tricks on YouTube. My first idea was to change the user interface with some UserScript but that failed. The second idea was to simply discard web browser and use YouTube Viewer, to be exact its GTK version to browse and watch videos.

The other big hurdle of YouTube is that videos often do missing or are simply blocked, audio muted, cut and so on. Due to this I decided to create a little video cache. This cache also helps me to get better video quality out of youtube-dl and YouTube-viewer.

Another things is that my machine does not like modern software which supports playing YouTube videos such as MPV (restarts? GPU) or VLC (does not work at all whatever I set in settings) I went for simpler Mplayer which works fine to my surprise.

The goal was to get videos with 1080p or lower with fps 30 or lower. I was unable to set this special condition directly in YouTube Viewer. I decided to create little “proxy” script which will be called from YouTube Viewer with video URL and video title (I am truly lazy here).

Little script

Since youtube-dl is written in Python I also decided to use Python and use youtube-dl python API to call it instead of cli. In this way the script has full control over youtube-dl configuration and I could play.

The biggest hurdle is that youtube-dl does not provide final filename for downloaded video. The only way to get the final filename which is used the script has to parse (with regular expression to have some fun) the debug output of youtube-dl to snatch the real filename in which the file exists or will exist in file system.

In the end I just called mplayer with video filename and title which was provided in argument.

The script is still not up to the standard and only works thanks to luck, use with a caution. You can download it here

Conclusion

I have to thank YouTube for providing me with this oportunity, without I am afraid I would be still mostly locked in YouTube interface, not wanting to go out.

Really big Thanks goes to youtube-dl and youtube-viewer projects. Thank you.

At the time I decided to use Bumblebee + bbswitch (to disable the card), it worked fine. There was a bit of drawback on CPU but it worked kind of nicely, I could run things on NVIDIA with simple command without closing current X.

Years went by on and everything was mostly fine, small things over the time as always. As plus there was support in mainline of NVIDIA drivers. But after a bit more this support was dropped and after an update I was welcomed with not working card, welcome legacy branch (390 for me).

Meanwhile libglvnd landed in Archlinux, what a pain. At first I did not manage to configure Bumblebee properly with it. But after some time I managed and it worked fine, using the card went fine.

There appeared a problem when NVIDIA card just randomly turned on and stayed on. I only heard noise, felt heat or experienced empty battery. This truly turned into HELL, waking up this hungry friend is costly. But I endured for a long time.

After a long time with help of some internet people and entities I managed to resolve one case but it still did not help and other things like Steam, Lutris and even SimpleScreenRecorder still somehow managed to activate the NVIDIA gpu.

It is sad to say but I desire a bit more stable environment. So I had disabled NVIDIA gpu in BIOS.

It was a good ride.

Good night NVIDIA, have a sweet dream.

At first I went on journey to find a nice dark theme which I would use as a base. After a bit I found tocttou/hacker-blog which is really nice. See demo.

Customization

It sounds all nice and fine but it is a bit plain. I decided that categories or tags would be nice. I chose tags because it is shorter and looks better. After a bit of search I stumbled upon a nice example. In the end it required a bit more work to add all functions but the basic idea was there. I added list of tags for each post on index:

memory page:

and tags page:

Even if it looks easy it took me a while. I also added some hidden features which noone will probably see unless you look at the code.

RSS / Atom feed

The theme / blog came with some atom.xml file which generated Atom compatible feed. I thought it is pain to manage and went for Jekyll plugin jekyll-feed. This plugin can create nice feeds, it should also be able to create feeds for each category. After many trials and errors there was a catch. the plugin could create category feeds only if category variable for post (memory) was present. Because I decided to use tags I changed it to categories, and so it did not work.

So I went and downloaded template from jekyll-feed repository and adapted it to work with my project. Surprisingly it was easy to do. After it was done I saved it as feed.xml and also saved a copy at feed/KDE.xml

Making the generated code nicer

After a while of working on the site I was more and more dissatisfied with the html code it generated. HTML was fine but there was a LOT of blank lines with white spaces left from templates.

The way how to get rid of this is to ass - at the beginning of template tags. That will make the Liquid template engine to remove the line on which this element is.

{%- template logic ... %}

After using this syntax on almost whole template the generated code looks much nicer.

Add videos

In my second post I wanted to add few videos. There are already prepared solutions to load youtube, video and some other external video services. But I wanted to host these small mini videos in git repository, so it was no go. It is sad to say but there is not {% video url %} tag. After a bit of searching there is truly a nice solution.

The files in _include directory can take a parameter when included from other files. See documentation.

While I was at it, I made it so that it is nice and easy to use. The code can be found in the repository.

In the “template” file I also included ffmpeg commands to convert video to mp4 and webm which are supported by browsers these days.

Example of usage:

{%- include video_embed.html video="video name" extensions="mp4,webm %}

It is easy and nice.

Revising feeds

After I learnt of this nice feature I adapted RSS feeds and made them use one template called atom.xml.

The whole xml file can now be reduced to few lines:

---
layout: none
tag: KDE
---
{% include atom.xml %}

Conclusion

Third-party plugins sucks.

Make your own and be happy.

The code with all changes and “tweak” can be found at Gitea.

Apart from other widgets in KDevelop window the main culprits in “Template Preview” toolview are kMessageWidged, line policy combo box and its label.

The combo box is wide, the proposed solution would be to set some fixed width or to move the label above or under it.

The other solution is to enable word-wrap for KMessageWidget and the line policy label. While I was at it I also removed one spacer which was only causing trouble. The effect of this change is very nice, the toolview requires a bit less space than before and the KDevelop window can now fit my screen.

But as always, there is a drawback. There seems to be a bug on KMessageWidget which does not look nice.

After a bit of playing with layout I sent a bug report and prepard merge request at invent.kde.org

The final results can be seen here. And it seems like the bug report should be edited….

The KMessageWidgets size is not changing when the width is decreased and words needs to be wrapped, it looks bad.

Based on the comment in source code, it should be possible to add tooltips for custom options in file templates.
How foolish I was to trust the comments. The option loading logic never counted with such possibility, it takes any data it can find in ’label’, ’tooltip’ and ‘whatsthis’ field and sets it as option`s label. The displayed text is the last read value of these three fields.

So I went and changed it to properly load the tooltip and adjusted the custom option page to display them. The merge request is waiting at invent.kde.org

And the result is: