Since I still hear about “microservices” more often than I’d like to. Here is another piece which points the trade-offs and that really there’s no one size fits all.
Good approach, you indeed need to immerse yourself in the context and see the recurring patterns. That’s the best way to figure out impactful solutions to recurring issues.
Interesting take, it’s more than just about hard drives being full. Indeed, there are many things on which we tend to procrastinate. But on the other hand… so many budgets to track otherwise!
Very interesting series I previously missed. It explores if software engineering is really engineering, and the differences with other engineering fields. Quite a few nice lessons to draw from it.
Performance on “secondary” GPUs has historically been suboptimal on Linux, especially with external GPUs. Let’s take a look at why it’s slow, and how we’re finally fixing it.
How Multi-GPU even works
To know how things work with multiple GPUs, we first need to look at how applications present images in single GPU systems.
With Wayland, the linux-dmabuf protocol is used for sharing images from applications to the compositor. The compositor advertises which GPU it’s using, and a list of drm formats1 and format modifiers2 it can make use of.
The application then allocates a dma-buf (direct memory access buffer) for each image with one of the supported formats + modifiers on the GPU of the compositor, and sends the compositor file descriptors for these buffers to share them.
This works quite well, but only for single GPU systems. In linux-dmabuf version 5, the compositor can only advertise one GPU as being usable! The app could still attempt to pass the buffer of any GPU to the compositor, but that can fail, requiring fallbacks, and it can cause really big performance issues.
Why it’s slow
When you import a dma-buf to a GPU, the kernel tries to be helpful and automatically ensures for you that the GPU you’re using can access the buffer. This sounds nice, and can be useful, but it’s also the source of all our performance problems.
Let’s say you have a laptop with an integrated GPU, and a dedicated GPU. When you start a game on the dedicated GPU, what would happen is
the game allocates a buffer on the dedicated GPU
it passes that buffer to the compositor
the compositor imports the buffer into the integrated GPU
To make step 3 work, the kernel would move3 the buffer to system memory. This can be really terrible for performance, since system memory is terribly slow in comparison to video memory, and the dedicated GPU has to go through the PCIe bus to access it, which has high latency.
To solve that problem, Vulkan and OpenGL drivers compare the GPU the application is using with the one the compositor advertises. If they’re not the same, then the driver will not share the buffers on the GPU with the compositor, but actually create a copy in system memory and share that copy with the compositor instead. So this would be
the game allocates a buffer on the dedicated GPU
the driver copies the buffer to another buffer in system memory
it passes that other buffer to the compositor
the compositor imports the buffer into the integrated GPU
Now imagine you’re not playing the game on your laptop display, but you have an external display connected to the dedicated GPU. The HDMI port on a lot of laptops is wired to the dedicated GPU, so this is a really common scenario. What happens in that case is
the game allocates a buffer on the dedicated GPU
the driver copies the buffer to another buffer in system memory
it passes that other buffer to the compositor
the compositor imports that other buffer into the integrated GPU
the compositor composites with the integrated GPU
the compositor copies the composited result to the dedicated GPU
So your game gets copied to system memory and back to video memory, for no real reason. The performance hit caused by that isn’t great in general, but it gets so much worse with external GPUs.
When all you have is one USB C cable to transfer data from and to the GPU, with high resolution monitors the bandwidth of that cable can be too small even for copying in one direction! Going in both directions completely kills performance. For example, I have a setup with
a Framework laptop 13
a rx 5700 XT in an external GPU enclosure
a 5120x1440 monitor at 120Hz connected to the external GPU
If I start vkcube (a super simple test app) on the external GPU, move it to the monitor and make it fullscreen, it only reaches 55fps!
How to fix it
In principle, we “just” need to stop the driver from copying the buffer around, and instead have it pass the original buffer from the dedicated GPU to the compositor. That’s exactly what I proposed version 6 of the linux-dmabuf protocol for: The compositor can support a list of GPUs instead of just one, and the application can tell the compositor which GPU it should import a given buffer into.
This all sounds nice and simple in theory, but as always, it wasn’t quite that simple in practice. There are basically three problems that needed to be solved for this change to be useful:
the driver needs to not do the copies unless actually necessary
the compositor needs to do the copies when required
to actually get any performance gains, the compositor needs to skip the copies whenever possible
Thankfully, I didn’t need to do this alone. Victoria Brekenfeld from System76 implemented support for the protocol in Mesa and Smithay, the library used by the cosmic compositor.
Unfortunately, it took a long time to support the protocol in a useful way in KWin. I needed to make it
aware of multiple GPUs beyond just displaying on them
properly handle GPU hotplug with that new infrastructure
deal with all the weird GPU setups in embedded systems
track which GPU each buffer is on
have generic multi GPU copy infrastructure
make those multi GPU copies actually as fast as the copies Mesa does internally
deal with GPU resets correctly while doing multi GPU copies
actually implement the Wayland protocol bits
Some of this was just moving code from out drm backend to more generic places, but a lot of it required doing things from scratch. Especially GPU reset handling was challenging because KWin’s effects APIs make a lot of assumptions about OpenGL contexts and didn’t support just stopping rendering a frame once it started.
While working on this, I also added Vulkan support to KWin specifically to make the multi GPU copies as fast as possible, which also took some time. As a positive side-effect though, we now have basic Vulkan infrastructure in KWin that can be used for a future Vulkan renderer.
Finally, just over 2 years after I first proposed the protocol, we had complete enough implementations to prove it works as expected and the protocol was merged. Since then, the KWin implementation was merged, the Mesa implementation should be merged soon and Nvidia has an (as of time of writing, unreleased) implementation of the protocol in their driver as well.
The actual performance gains
So, how much does this actually help? With vkcube, we go from 55 fps to the full 120 fps of the monitor, but that’s hardly a practically relevant test.
So I fired up Cyberpunk 2077 with the “low” graphics preset and it went from 27 fps with Mesa main to 50 fps with linux-dmabuf v6. That’s more than 80% better performance!
Mesa main
Mesa with dmabuf v6
From playing around a bit with the graphics settings, it seems likely that the game is bottlenecked by the USB C bandwidth, so the performance uplift could be even better with additional driver optimizations.
I haven’t been able to benchmark the impact of dmabuf v6 on “normal” laptop setups, because I don’t yet have access to a Nvidia Vulkan driver version implementing the protocol, and I don’t have a laptop with a dedicated AMD GPU. If I had to guess though, I’d expect improvements there to be much more moderate, probably something in the 5-10% range. I’ll update this post once I know more.
However, this performance uplift comes with two big caveats…
When does this work?
With the current KWin implementation, compositing always happens on the “primary” GPU. On laptops, by default that’s whatever GPU the internal display was connected to when KWin started. This means you can only get those large benefits if the game gets direct scanout.
My rx 5700 XT doesn’t have support for color pipelines, so I only get the performance benefit if HDR is disabled, night light is disabled and no color profile is used. I already have a merge request to lift that restriction though, stay tuned for part 2 to find out more about that!
There is another caveat though, and that’s a far bigger challenge: Most games still run on X11, and implementing the protocol (in a useful way) in Xwayland is incredibly challenging because of some assumptions X11 makes. I don’t know how that could be fixed, or even if it’s feasible at all, so for now, you can only get these benefits with games using Wayland directly.
For a lot of native games, using SDL_VIDEODRIVER=wayland does the trick, and for most Windows games, you can use Proton forks with the Wine Wayland driver. In my experience, Wine Wayland works really well and even HDR just works™ with it nowadays.
drm formats describe how the memory in the buffer is used, like “16 bits red”, “8 bits red, 8 bits green, 8 bits blue, 8 alpha bits” or similar ↩
format modifiers (more or less) describe how the pixels are laid out in memory ↩
meaning, it will create a copy, and then delete the original ↩
I attended the GSoC Alumni Camp in Delhi, it was an amazing opportunity to make people who are already familiar with open source aware of KDE Community
and I was actually able to do that. The event itself was from 8:00 A.M. to 10:30 P.M. at night.
The event started off with a formal opening session.
Soon after was the scavenger hunt (we were hunting people!)
it was a 5x5 grid where you needed to get signatures of people who fit the question in the box.
A few of them were :
I was born in July (I signed a lot of them )
I have held a koala
I know 3+ languages
I can juggle
I prefer go/Rust
and so on.
In all honesty it was an amazing ice breaker and something to get to know people.
After that there were a few unconferences followed by lunch and then more unconferences till 6:00 P.M. after that lightning talk sessions started and fortunately I got to represent KDE.
Unconference Session – From GSoC Student to Maintainer/Mentor
I also hosted one of the first unconference sessions titled:
From GSoC Student to Maintainer/Mentor
I wasn't expecting a large audience, but at one point the room was at its capacity around 40 attendees.
The audience included:
GSoC Organization Administrators
Mentors
Open source employees
Existing contributors
Google code-in students
The discussion focused on contributor retention after GSoC and how organizations can better support contributors beyond the program.
It led to some interesting discussions where we found out what some organisations are doing.
Discussion Points (Noted by an org admin)
Leverage Contributors as Subject Matter Experts
GSoC contributors naturally become subject matter experts for the components they build. When new issues, bug reports, or pull requests relate to their work, proactively invite them to participate in the discussion or review. This helps contributors feel that their expertise is valued and keeps them engaged with the project beyond GSoC.
Improve Communication Around Pull Request Reviews
Complex pull requests often require focused reviews and may take longer than isolated bug fixes. While taking the necessary time for a thorough review is important, mentors should acknowledge review requests and, where possible, communicate an expected timeline. Even a brief acknowledgment improves the contributor experience.
Using Organization Stipends for Contributor Retention
Explore using the GSoC organization stipend to support promising contributors after the program ends. This could fund continued maintenance or feature development through a transparent platform such as Open Collective, allowing contributors to invoice for their work and remain actively involved with the project.
Recognizing Mentor Contributions
Mentoring requires a significant investment of time in reviewing code, guiding contributors, and providing technical support. Consider allocating a portion of the organization stipend to compensate mentors, recognizing their efforts and encouraging long-term participation.
KDE's Community Support
Another topic that generated considerable interest was KDE's support for its contributors.
I talked about how KDE sponsors some students to attend Akademy, and how I was also sponsored to attend the event, which a lot of people agreed organizations should've done (some did, a lot didn't) but a lot of people agreed that when an organization consider these things contributors stay.
After all the unconferences ended a lot of people came up to me saying mine was the best and I managed and engaged everyone in the best way possible ( I was on cloud 9).
Lightning Talk
One of the highlights was being selected to give a lightning talk. Only 18 participants were chosen, and I was one of them.
My GSoC project Join KDE is an advertisement for KDE itself, so the project itself became an easy conversation starter. It was a great opportunity to introduce KDE and make more people aware of the organization.
Although I did fumble the last part of my presentation it was more of me being confused where my slide went (some slides were misplaced) aside from that it went as any other informative talk but hey I was funny on the stage (refer to the pics) so people remember me! You win some you lose some, oh well.
Conversations Throughout the Event
Outside the sessions, I had some interesting discussions with contributors from different organizations. Some contributors showed interest in working wwith us while some showed interest in our eduction/science side but honestly it's hard to bring contributors over as they're in different domains, org admins/mentors or they're busy with their careers.
SU2 Foundation
I had a conversation with a maintainer from the SU2 Foundation, whose work focuses on computational fluid dynamics.
We discussed KDE's scientific and educational initiatives, including KDE for Scientists and several education-related projects. I also shared the KDE For You website as an introduction to the wider KDE ecosystem.
After Party
The conversations continued during the after-party on the following day.
I met a group of students who had become interested in open source after interacting with contributors throughout the event. They had originally attended as friends of another contributor.
One of them is now actively looking for a project to begin contributing to within KDE. YAYYY!
New Contributors
Several first-time contributors approached me asking about projects suitable for beginners.
I introduced them to a number of KDE projects and explained possible contribution paths.
But the same issue they're busy with the beginning of their careers.
Overall Thoughts
The Alumni Camp was one of the best ways to make people aware of KDE and I tried my best while i had fun ( we played uno there! Will bill the cards to the organization jkjk), a lot of people seemed intrigued and asked me to share some stuff about KDE to them or they'll try it on their own. Met someone from GNOME (they didn't get any lightning talk HA! I know, I know no inter project fighting that's bad) had some good conversations, teased the person from GNOME, had breakfast with them, and tried to make the most of the unconference session and the lightning talks to at least get people to know the name KDE!
Today we're releasing Krita 5.3.3 and 6.0.3, containing many bugfixes and improvements across the board.
Android Donations and Supporter benefits
On Android, version 5.3.3 lets you support Krita's development directly through the application. This goes through the Play Store, so if you have a device without Google services or are using a debug or nightly version of Krita, it won't be available and instead will show you donation links.
In return for supporting Krita, you'll be able to download resource bundles directly in the application, including the paid Digital Atelier bundle. The splash screen and welcome page will also tell you about the supporter packs or donation links. You are not required to make any purchases, no features are locked behind a paywall and you won't get nagged about it while you're actually working. You can also of course still manually import resource bundles, the downloads are just a convenience feature.
Special thanks to Freya Lupen and Elena Sagalaeva, for multiple bug fixes, as well as the many newcomers who contributed to this release (named in the changelog).
Don't free hold frame undo command at scope exit (Bug 520732)
Added get started, community, website and source code icons (Thanks to Arkady Flury)
Try removing shell=True from the python calls on Windows
Get rid of division by zero in the KisMplTest.( Thanks to Elena Sagalaeva)
[android] All animation export and import functionality has been removed for now. Mostly because these functions didn't do anything. We're working on a properly working replacement for Android specifically.
Delete QuaZipFile before deleting the archive (Bug 520362)
Fix layers getting deselected when starting transform (Bug 517590)
Ensure a layer is activated for new views (Bug 518465 , Thanks to Gregg Jansen van Vuren)
Fix FFmpeg 7.0+ to render HDR animations (Bug 520669)
Fix Out-of-Bounds-Selection-Inversion (Bug 518235, Thanks to Ricky Ringler)
Fix Display Selection action check stats
Fix crash on null event pointer in Abstract Input Action (Bug 520581)
Force non-var vertical metrics to use default vertical caret. (Bug 517717)
Fix font-family sorting for fonts with wwsWithoutName set mixed.(Bug 518874)
Try to deduplicate font family names better, so we don't get sql errors. (Bug 517361)
Fix bug 522430. Switching between frames should always complete. (Bug 522430)
Fix deadlock when doing perspective transform (Bug 520602)
Expose tool and brush change signals to PyKrita (Thanks, Moritz Staudinger)
Various fixes to the selection action panel, including the ability to stick it to the sides of the canvas by pressing the "pin" icon.
Various fixes to the comic panel editing tool functionality.
Allow higher resolution of thumbnails on Welcome Screen.
⚠️ Warning
We consider Krita 5.3.3 suitable for productive work; 6.0.3 is, because of the many changes from Qt5 to Qt6 more experimental.
Download 5.3.3
Windows
If you're using the portable zip files, just open the zip file in Explorer and drag the folder somewhere convenient, then double-click on the Krita icon in the folder. This will not impact an installed version of Krita, though it will share your settings and custom resources with your regular installed version of Krita. For reporting crashes, also get the debug symbols folder.
The Linux AppImage and the source tarballs are signed. You can retrieve the public key here. The signatures are here (filenames ending in .sig).
Download 6.0.3
Windows
If you're using the portable zip files, just open the zip file in Explorer and drag the folder somewhere convenient, then double-click on the Krita icon in the folder. This will not impact an installed version of Krita, though it will share your settings and custom resources with your regular installed version of Krita. For reporting crashes, also get the debug symbols folder.
There is a bug in KDE's bug tracker that is almost as old as some of our
contributors: bug 342056,
"Ridiculously slow file copy (multiple small files)", reported by Alexander
Nestorov back in 2014. The report is blunt: copying a 15 GB folder of roughly 3
million small files took 5 to 10 hours in KDE, versus about 20 minutes with
rsync. One commenter summed up the mood:
I generally use cp and rsync instead of dolphin for anything more than a few files.
That bug has been stuck in my head for a while, so this post is about finally
fixing it. I have been working on KIO's copy path, and I want to walk
through where the time was going, some history that explains why, and show
numbers across KIO versions including a plain cp for reference.
At with any optimization, let's measure first:
Where a KIO 6.28 copy of 5000 small files spends its blocking time, and there
is no single hotspot. Each file blocks in a burst of short syscalls: reading the
mount table (/proc/self/mountinfo), opening the source and destination, and
round-tripping a command to the worker over an internal socket. The socket
round-trips, the send and receive plus the waits for each reply, are the frames
highlighted in red, about 15% of the blocking here, spread as thin slivers
across both threads because every one of the thousands of files pays them. The
frames in green are the real filesystem work the copy cannot avoid: the statx
and openat calls, copy_file_range moving the bytes, and the ext4 metadata
updates beneath them. That per-file storm, not any single call, is what the rest
of this post discusses.
(Off-CPU flame graph from sched:sched_switch, weighted by number of blocking
switches; click to open the full SVG.)
A little history
KIO is the layer behind almost every file operation in KDE software, from Dolphin's copy
and paste to the network transparency that lets you open sftp:// or smb://
URLs as if they were local. Its design goes back to the KDE 2 days, around the
year 2000. Back then the answer to "how do I do I/O without freezing the UI" was
not a thread but a separate process: a worker (we used to call them kioslaves)
is launched for a protocol and talks to the application over a socket. This
predates usable threading on Linux (Native POSIX Thread
Library only landed
in Linux 2.6 in 2003), so processes plus socket IPC was the portable, robust
choice, and it still is for network protocols today: if an smb:// worker
crashes, your file manager does not go down with it.
The flip side is cost. Every request and its reply are serialized and sent over
that socket, with an event-loop hop on each side. For file:// that overhead
buys you nothing, because there is no untrusted network on the other end, just
the local disk.
In 2022 David Faure changed that: Implement running KIO workers in-process
using a thread landed in Frameworks 5.95. Since then the file worker runs on a thread inside
the application instead of as a separate process (other protocols stay
out-of-process for robustness). That removed process launch and context-switch
cost.
No socket between the thread and the app (6.29)
There was still a catch that had been hiding in plain sight: even in-process,
the worker thread and the application talked to each other over a socketpair,
serializing every command as if they were separate processes. Once the thread
was used, that socket was pure overhead.
That's the red highlight in the first flamegraph.
So in-process workers now use a real in-memory transport instead of a loopback
socket. A new ThreadConnectionBackend handles commands, and for reads the actual
data buffer, straight to the peer thread, with no serialization and no syscall,
which makes in-process file reads zero-copy. This is the biggest single jump in
the numbers below, and it has landed on master for 6.29
(kio!2279).
Next: batching the copy and the stat (under review)
The changes above make each command cheap. The remaining cost is that there are
still so many of them: copying N files is N separate file_copy jobs, each one
going through the job scheduler and taking its own round-trip to the worker. For
many small files that fixed per-file cost, not the data, is what dominates.
CopyJob can dispatch a whole batch of plain local-to-local files to the worker
in one command, and the worker copies them (using copy_file_range,
and reflink on filesystems like Btrfs or XFS that support it). Each file still
gets its progress signal, its byte accounting, its undo entry, and anything
the worker cannot do blindly — an existing destination, an unreadable source, a
rename or skip decision — is handed back to the normal per-file path. The gate is
conservative on purpose so a hung mount can never freeze the batch
(kio!2282).
The same copy on 6.29 with batch-copy: the red socket round-trips are gone. Two
things removed them: the in-memory transport from the previous section (no
socketpair at all) and batch-copy folding a whole run of files into one command
instead of one round-trip each. What is left is real filesystem work: the green
frames are the copy itself, copy_file_range moving the bytes, the openat and
statx calls, and the ext4 metadata beneath them. The wide plateau that is not
green is the benchmark loop deleting each pass's files before the next run
(unlink), which is teardown rather than part of the copy.
Before copying, CopyJob stats every source, which is another per-file
round-trip. The follow-up batches the stat phase over the same in-process lane.
This is the batch-copy-stat column below.
Both of these are under review and targeted at a release after 6.29, so treat
their columns in the table as a preview of what is coming rather than something
you can install today.
Numbers
Methodology: a 13th-gen Intel Core i7-1365U laptop, everything built
RelWithDebInfo without any sanitizer, copying onto a real ext4 volume (so no
reflink shortcut, copy_file_range moves real bytes). Each KIO version supplies
both its own libKIOCore and its own kio_file worker. Times are the median of
5 runs (3 for the large set) after a discarded warm-up, so caches are warm and
equal for every version. cp -r --preserve=mode,timestamps is included as the
floor, since that is what the bug reporter reached for instead of Dolphin.
5000 files of 4 KB each (the many-small-files case from the bug):
Version
Median time
cp -r (coreutils)
81 ms
KIO 6.28.0
1616 ms
KIO 6.29-dev (master)
396 ms
KIO 6.29-dev + batch-copy
88 ms
KIO 6.29-dev + batch-copy-stat
93 ms
1000 files of 5 MB each (bulk data, where the copy is I/O-bound):
Version
Median time
cp -r (coreutils)
1312 ms
KIO 6.28.0
1997 ms
KIO 6.29-dev (master)
1687 ms
KIO 6.29-dev + batch-copy
1506 ms
KIO 6.29-dev + batch-copy-stat
1454 ms
The small-files case is the story. KIO 6.28 was about 20x slower than cp,
which is exactly the ratio the 2014 report complained about. Removing the socket
for in-process workers, together with no longer probing the destination
filesystem once per file, takes it from about 1.6 s to 0.4 s (roughly 4x).
Batching the copy takes it to 88 ms, essentially cp speed and about 18x faster
than 6.28. That "I use cp instead of dolphin" line finally has an answer.
The large-file case was already well covered before any of this work, and the
table shows it: the copy is bound by moving the bytes, so even KIO 6.28 stayed
close to cp there, and the per-file optimizations barely move the number
(batch-copy lands within about 15% of cp). That is the whole reason this
effort targets many small files instead, where the fixed per-file cost, not the
data, is what dominates.
batch-copy and batch-copy-stat come out tied here: the benchmark copies a single
directory, so its entries arrive in one listing and batch-stat has nothing to
batch. It helps the other case, copying a large selection of individual files
that would otherwise be one stat round-trip each. Since it removes the
round-trip and not the stat call, the win is largest when stats are cheap, not
when the disk is slow. Read the two as equal on this test.
And cp will almost certainly stay a little ahead, by design: KIO does more
than copy the bytes. It reports per-file progress, can be paused and cancelled,
keeps an undo record, preserves permissions and timestamps, resolves conflicts,
and tells any open file manager what changed so its view refreshes. That work is
not free, and the goal was never to beat cp, only to make the local case fast
enough that reaching for cp instead of Dolphin stops being the obvious choice.
Caveats and what is next
These are laptop numbers on one ext4 disk with warm caches, meant to compare
versions, not to be absolute. On Btrfs or XFS the batch path reflinks instead of
copying, which changes the large-file picture entirely. The batch fast path only
engages for plain local-to-local copies on a responsive filesystem; anything
else (network, FAT/NTFS, conflicts needing a dialog) falls back to the existing
per-file path, unchanged.
Thanks to David Faure for the in-process worker threads that made all of this
possible, to Frank Reininghaus for the earlier listing fixes, and to Alexander
Nestorov for a bug report that aged well enough to still be worth closing.
Months ago, some distributions accidentally removed cryfs and encfs
from Vaults dependencies (temporarily). And some distributions are
removing them intentionally from their package repositories (at least,
for the time being).
It might be a good idea to switch your vaults to gocrtyptfs.
There is no automatic process for this, just create a new vault (it
will be gocryptfs-based – other backends have been removed from the UI
in the latest versions) and copy the data from the old one there.
Some context
To avoid users thinking there is something serious going on, this is
some background on changes:
Plasma Vault creation wizard supports only gocryptfs since plasma-vault/-/merge_requests/62 as it is
the most maintained upstream project, and to avoid “confusing the
users”.
Some distro packagers thought this means Vault no longer supports
cryfs and encfs and they removed them from the package dependencies.
This change was quickly reverted as there are existing users that
already have cryfs and encfs-based Vaults.
Some distros are cleaning up FUSE2-based file systems and going
FUSE3-only. Some of them ship a FUSE3-based CryFS, and some seem to be
waiting for an official FUSE3-supporting release (see
github.com/cryfs/cryfs/issues/419).
So, nothing important has happened. Your vaults are as safe as they
always were.
It is just that it might bring you a headache if the backend you use
for your vaults disappeared after you update your system.
And headaches for me when I get bug reports that Vaults no longer
support encfs or cryfs :)
This is a weekly update from my Google Summer of Code 2026 project with KDE, improving effect widgets in Kdenlive, a free and open source video editor.
Gradient widget merged
MR !911 went through one more round of review, Jean-Baptiste caught a small inconsistency in how the 32-stop limit was applied and fixed it directly (c2d7a3d2, "Use limit everywhere"). After that, the Gradient widget merged. Two of the three widgets from the proposal are now in master.
One known issue flagged during review: MLT's gradientmap filter doesn't currently support alpha in its gradients, that's a bug on MLT's side, not the widget. Jean-Baptiste merged anyway to make the 26.08 window, with a plan to either disable alpha or fix it upstream before the final 26.08.0 release.
Speed Ramp: starting the third widget
With Gradient shipped, moved on to Speed Ramp. My original proposal described free-draggable bezier handles on the remap timeline's connector lines. Investigation before writing any code turned up a real problem with that: MLT has no bezier keyframe type to serialize a time map into, only preset easing types (linear, smooth, cubic, exponential, and more, as suffix characters on keyframe positions).
Checked with Jean-Baptiste. Turns out free bezier handles in MLT had already been investigated a few years ago and found difficult to integrate. His suggestion: use the same keyframe type system Kdenlive already uses for effects like volume and brightness, letting users pick a type from MLT's existing list rather than dragging custom handles.
Finding the right pattern to reuse
Kdenlive already has a widget that draws real keyframe curves in a plain QWidget, KeyframeCurveEditor, used in the effect stack for other parameters. It samples MLT's interpolated value per pixel and draws the result, no QML involved. That became the reference pattern for adding a curve to the remap dialog (RemapView).
Before building on it, tested MLT directly to confirm non-linear keyframe types are actually honored during playback, not just valid syntax. They are, sampled values matched the exact easing formulas. One gotcha found along the way: the keyframe type suffix has to go on the segment's starting keyframe, not the ending one, otherwise it's silently treated as linear.
Implementation
Four commits, each built clean before the next:
Added per-keyframe type storage (m_keyframeTypes) alongside the existing keyframe position map, synced across all 30+ places the position map gets mutated
Switched keyframe serialization and parsing to use MLT's own animation API instead of hand-formatted strings, so type suffixes are always placed correctly
Added a per-pixel sampled curve, drawn additively in the existing dual-ruler layout, nothing existing removed or changed
Added a keyframe type selector, starting with a curated list (Linear, Smooth, Cubic In, Cubic Out); the full MLT list includes bounce and elastic types, which cause the source time to briefly go backward on a time map, that's a product decision still pending, so it's left out for now
Manually verified: linear playback is unchanged, curve shape follows the selected type, old projects load correctly as all-linear, undo/redo works through type changes, and types survive keyframe drags and clip resizes.
What's next
Speed Ramp branch is local, not yet pushed. Bringing results to Jean-Baptiste before pushing and opening an MR.
With the main proposal items completed, I spent the last 2-3 weeks working on open issues in the KeepSecret tracker, fixing bugs and improving the overall experience.
Added a minimal user handbook for KeepSecret in DocBook format — the standard format used by KDE applications. The guide covers the main features: wallets, managing entries, generating passwords, importing and exporting, and locking. It's accessible via the Help menu. Writing it also involved adding the CC-BY-SA-4.0 license to the repo, which triggered a kirigami-app-components 1.0.1 release (more on that below).
Added the Find Action entry Ctrl+Alt+I to the global drawer. This opens the Actions Explorer dialog — a standard KDE dialog that lets users search through all available actions in the app by name. During testing, I discovered that kirigami-app-components 1.0.0 had a bug where FindAction wasn't working. Marco Martin fixed it and released 1.0.1.
Fixed a UX bug in 3-column mode: when switching wallets, the entry panel from the previous wallet stayed open. The fix was a single line — calling App.secretItem.close() when a wallet is clicked in the sidebar.
Added the Delete key as the default keyboard shortcut for deleting secrets. Since there's already a confirmation dialog, binding Delete directly is safe.
If anybody had told me I'd end up working on the very Linux desktop environment I've been daily-driving for the past three years, I probably wouldn't have believed them.
This summer, I got to learn and work on the main Network Manager component used by default in KDE, the desktop environment relied on by millions of users worldwide (source). Looking back, this has been the most challenging, and by far the most rewarding, journey of my open-source career so far.
How It All Started
I started looking for open-source projects to contribute to way before GSoC even announced its list of participating organizations. Back in October, I began contributing to major projects like OpenTelemetry, Lima-VM, and WasmEdge, repos I'm still active in today.
Somewhere along the way, I decided to explore the KDE ecosystem and give it a shot. That decision changed the trajectory of my year. From that point on, my co-mentors and I were actively working through a number of open-source issues together, including bugs in KClock, the native clock application used across KDE, long before the official GSoC coding period even began. That early groundwork ended up being exactly what prepared me to take on Network Manager as my actual project.
The Project
My GSoC project is titled "Make network related KCMs feature complete with desktop equivalent," under Plasma Mobile, mentored by Carl Schwan and the rest of the Plasma Mobile team.
Here's the core problem: Plasma Desktop and Plasma Mobile each ship their own, completely separate networking settings modules. The desktop KCM (kcm_connections) is mature and feature-complete, with full Wi-Fi security support (WPA/WPA2/WPA3, 802.1x Enterprise, various EAP methods), VPN configuration, and more. Plasma Mobile's networking KCMs, on the other hand, are QML/Kirigami-based and touch-friendly, but only support the basics (WPA/WPA2 Personal, essentially). Because the desktop editor lives in libs/editor/ as Qt Widgets, it can't be dropped into the mobile UI as-is, Widgets and Kirigami don't mix, so mobile has ended up stuck maintaining its own thinner reimplementation instead of reusing the desktop's mature backend logic.
My project is to close that gap: pull the shared backend logic in libs/ out from underneath the Qt Widgets layer, rebuild the connection editor in QML (in a new libs/editorqml/) using Kirigami, and get both the desktop and mobile KCMs consuming the same backend through QML property/signal bindings. This way, a feature added once (say, WPA3 Enterprise or a new VPN protocol) works identically on both form factors instead of being implemented twice. The work is broken into phases: project setup, the Wi-Fi/security connection-editor foundation, additional connection types (wired, hotspot, VPN, bonds/bridges/teams/VLANs), refactoring the KDED popup dialogs (like the Wi-Fi password and PIN prompts) from Widgets to QML, and finally integration testing plus writing proper architecture documentation for plasma-nm, which currently has very little of either.
The Challenge
The biggest hurdle, by far, was getting a real grip on QML and the architecture of the broader codebase.
Coming in, I understood C++ reasonably well, but plasma-nm's UI layer is written almost entirely in QML, a declarative language that describes what the interface should look like rather than how to build it step by step. On top of that, plasma-nm isn't a single self-contained app, it's a plugin-based system that sits on top of NetworkManager (the system daemon), talks to it over D-Bus, and exposes that state through a KDED module and a Plasmoid applet that the QML frontend binds to. Understanding which layer owned a given piece of logic, the daemon, the backend plugin, or the QML view, took real time.
Some of the specific things I had to work through:
Signal/property bindings in QML: tracing how a change in connection state on the backend actually propagates up through C++ signals into a QML property binding, and where things could silently break if a binding wasn't wired correctly.
The plugin architecture: plasma-nm supports different connection types (Wi-Fi, VPN, mobile broadband, etc.) via separate plugins, each with its own QML + C++ pairing, so I had to learn to navigate a fairly large, distributed codebase rather than one monolithic file.
Key Contributions
Here is the list of MRs that are merged in upstream as of this date:
Beyond the technical skills, QML, Network Manager core concepts, and plugin architectures, this project taught me a lot about contributing to a codebase that real people depend on every day. KDE isn't a toy project, it's a software that is used by millions of people to connect Wi-Fi and manage their networks without thinking twice about it. That changes how carefully you write and review code.
Thanks
A huge thank you to my mentors, Devin Lin and Carl Schwan, for the patience and guidance through the pre-GSoC Phase. And thanks to the broader KDE community and Plasma Mobile for being as welcoming as it is to newcomers navigating a genuinely large codebase for the first time.
This week saw a bunch of work on some core infrastructure components, like the RDP server, login greeter, and screen configuration tooling. That’s in addition to lots of other good work, too:
Notable new features
Plasma 6.8
You can now configure Plasma’s built-in remote desktop server to automatically lock the session after the last connected RDP client disconnects. (Nick Haghiri, krdp MR #211)
Discover’s transaction progress view now shows items in groups, with the ones currently in progress in a group at the top. (Aleix Pol Gonzalez, discover MR #1219)
On multi-screen setups, the existing Meta + 1-9 shortcuts for jumping straight to a task now target the panel on the screen you’re actively using, rather than always going to the primary monitor’s panel. This matches the behavior added to last week’s new task-switching shortcuts. (Salman Farooq, plasma-desktop MR #3884 and plasma-workspace MR #6848)
You can now find languages in Plasma Setup by searching for their locale code, localized name, or English name. (Tiziano Gaia, plasma-setup MR #109)
Frameworks 6.29
Further improved the visual fidelity of small icons drawn with the Kirigami.Icon component. (Méven Car, kirigami MR #2127)
Notable bug fixes
Plasma 6.6.7
Fixed an issue that made KWin crash on login for newer NVIDIA GPUs using the latest NVIDIA drivers in conjunction with various color management features. (Xaver Hugl, KDE Bugzilla #523287)
Restored the ability to play full-screen video in Chromium-based browsers on a virtual screen, such as those used when screen recording. (Xaver Hugl, Bugzilla #523353)
Fixed an issue that could make Plasma freeze and then take down apps as well after you switched activities with unusual values in your Plasma config file. (Shouvik Kar, KDE Bugzilla #522039)
Fixed two visual glitches affecting the Digital Clock widget: one that made the layout overflow on a panel right after turning on the “always show seconds” setting, and another one that made the time zone label fail to use the systemwide font family. (Hunter White and Nate Graham, KDE Bugzilla #523010 and KDE Bugzilla #523164)
When an image used in a slideshow in the Media Frame widget is changed on disk, the widget now notices it properly, rather than replacing it with a blank frame. (David Wild, KDE Bugzila #521538)
Plasma 6.7.4
Fixed a weird bug that could break Plasma Login Manager after you unplugged a screen and plugged it back in while on the login screen. (Oliver Beard, KDE Bugzilla #523313)
Fixed a bug that could make KWin crash on login on various laptops. (Xaver Hugl, kwin MR #9632)
Fixed a case where System Settings could crash when Fanatec racing pedals were plugged in. (Sebastian Sauer, KDE Bugzilla #522886)
Restored the ability to change the wallpaper and apply Plasma settings to Plasma Login Manager for users who upgraded from Plasma 6.6. (David Edmundson, KDE Bugzilla #517081)
Frameworks 6.29
Fixed a surprisingly common yet random-seeming way that Plasma could crash. (Shouvik Kar, KDE Bugzilla #519614)
QML-based KDE apps once again remember whether they were maximized or not across launches. (Nate Graham, KDE Bugzilla #522205)
Worked around a Qt bug that made some keyboard shortcuts in Kirigami-based apps not get assigned when using the app in certain languages other than English. (Manuel Alcaraz Zambrano, kirigami MR #2124)
Icons drawn by the Kirigami.Icon component now use the correct aspect ratio for portrait-orientation images when they’re being rounded to standard icon sizes. (Méven Car, kirigami MR #2126)
Notable in performance & technical
Plasma 6.8
We have created a new kscreenctl tool, intended as a future replacement for the old kscreen-doctor tool. It includes more features, more robustness, and a more conventional usage method. (Vlad Zahorodnii, kscreen MR #487)
Discover is now faster to load icons for Flatpak apps. (Aleix Pol Gonzalez, discover MR #1366)
How you can help
KDE has become important in the world, and your time and contributions have helped us get there. As we grow, we need your support to keep KDE sustainable.
Would you like to help put together this weekly report? Introduce yourself in the Matrix room and join the team!
Beyond that, you can help KDE by directly getting involved in any other projects. Donating time is actually more impactful than donating money. Each contributor makes a huge difference in KDE — you are not a number or a cog in a machine! You don’t have to be a programmer, either; many other opportunities exist.
You can also help out by making a donation! This helps cover operational costs, salaries, travel expenses for contributors, and in general just keeps KDE bringing Free Software to the world.
To get a new Plasma feature or a bug fix mentioned here
ervin
@zamundaaa:kde.org
GSoC
