Hacker News Reader

Comment by grenoire | Hacker News Reader

I am having some serious slowdown issues since 1.* on Win 10, so I'm stuck on 0.92.

I have no qualms with it, but the issue plagues all my Win 10 machines and I just wonder what's going on, and if I'm a select minority with it.

Besides that, thanks Inkscape for being such an incredibly useful vector tool!

View on HN

I’m also fairly bitter about the handling of 1.0 on Windows. I reported various problems at the first beta in moderate detail, my report was more or less ignored and they went ahead with various significant regressions in performance and UI, largely because no developer was interested enough in Windows, and I and others evidently failed to convince them of just how bad it was on Windows; GTK+ 3 in general has shown a general disdain for doing Windows well, whereas GTK+ 2 had been worked over until it wasn’t too bad. Yes, Inkscape is done by volunteers, but I reckon that some of these issues should have been considered blockers to at least recommending 1.0 for Windows, if not its release. So I also stuck with 0.92 while on Windows.

But I very much like what I’m seeing in the draft release notes for 1.2: as well as some really nice-looking functionality improvements, they mention various performance improvements in approximately the areas that had the worst regressions in 1.0 (such as zooming and panning), and UI adjustments to reduce space consumed. I haven’t tried 1.2 alpha 1 out just yet, but I probably will in the next few days and I’m really looking forward to it. (Incidentally, I’ve been back on Linux for the last ten months, and using the latest releases, since it works better under Linux, though Adwaita is still horrifically wasteful of space given Inkscape’s habit of stacking many buttons side-by-side, so I installed Minwaita which reduces that problem a lot.)

It might even be, that the developers do not own a Windows machine. Did you ask them, whether they have one to test on and reproduce the slowdown? Or perhaps offer to buy them one? It is a pity that things slow down for you on Windows, but I don't really see, how it is their responsibility to cater to a proprietary system, that is actively hostile to its users.

Preferrably it wouls not slow down for you, ofc, but they may well deem it out of scope of their project to fix issues for asystem, that they do not have the code for to look at where it is going wrong, or simply do not have the time and resources to deal with that. Add to that, that any day MS could destroy the fruit of that work again.

I can understand, that they don't want to invest resources into that.

I appealed to them in concrete terms at one point not long before the release of 1.0: either fix things for Windows, or keep on recommending the 0.92 series for Windows. There’s no denying that for the 1.0 release Windows was a second-class citizen. There’s no reason why they couldn’t then say: “look, 1.0 is available, and if you want the new features you can use it, but it’s considerably slower than 0.92 and the Windows theme is very badly broken so you’ll be stuck using the Adwaita theme which is extremely wasteful of space, so all up we suggest sticking with 0.92 for now; BTW talk to us if you’re interested in helping to improve things or sponsoring such work”.

Do you think all users would agree with you?

I used Inkscape on Win10 recently, and also on Kubuntu (21.04 a and 21.10, I think) and the experience was comparable.

I do not like the icon themes now, they're very low contrast and difficult to decipher for me so use the old, colourful icons I'm familiar with. Apart from that and one problem with a dialog not appearing until I selected the option a second time (the XML dialog), all worked well.

I think it's difficult, a x.0 might not be snappy, might even have serious issues, but there's not enough manpower resource to address everything at once, it's going to be incremental. Holding things back would negatively impact the momentum of the project IMO, but yes may benefit an individual.

Quite a few people have remarked on the performance problems. For the spacing thing, the use of Adwaita doubled the minimum width of things like the Fill & Stroke panel because of its excessive button padding, so that on my Surface Book (effectively 1500×1000) with the typical layout, I would end up with something like a 35–40% reduction in available canvas width (and maybe 10–15% for the height); but even the common resolution of 1920×1080 is going to suffer a fairly significant reduction in available width, compared to the Win32 theme.

But I couldn’t even properly switch back to the Win32 theme because of severe bugs that rendered it very problematic on all displays and basically completely unusable after a bit on any comparatively small high-DPI displays like my Surface Book’s. I don’t know if any of these issues have been fixed; I did try Inkscape 1.0 final briefly but decided it was still just too bad compared to 0.92. And now I’m back on Linux and so, like the Inkscape developers, don’t care all that much about Windows any more.

(https://gitlab.com/inkscape/inbox/-/issues/879 was my initial report, including explanation of some of the major Win32 theme regressions. Given the adjacent topic in this thread of GTK 4, the mention of overlay scrollbars in that thread points to a mess too; https://wiki.archlinux.org/title/GTK#Disable_overlay_scrollb... says: “as of GTK 4, the overlay nature of the scrollbars is part of the toolkit.” Ugh, GTK is obviously not interested in being a native look-and-feel toolkit if they deliberately remove the ability to control scrollbar overlayedness at the system level or in the theme. If you require the app to switch it on, it might as well not exist. Sounds like they’re continuing with the sabotage.)

GTK is a GNOME library and GNOME has shown in the past that they not only don't care about third party software but are actually extremely harmful to the community to the point that you might as well call it sabotage. The transition from GTK2 to GTK3 was extremely painful, had no benefits for anyone except GNOME and had many regressions across the board. The right course of action for Inkscape would be to switch back to GTK2 and be as independent from GNOME as possible. For everybody else it would be advised to make sure that your software runs well everywhere but GNOME. Simply check if getenv("XDG_CURRENT_DESKTOP") is equal to "GNOME" and exit(EXIT_FAILURE) if true.

Display scaling is only supported by GTK 3 and later, so downgrading to GTK 2 would result in a blurry interface for a lot of users (and not just GNOME users). 1080p or higher resolution laptops and 4K monitors are quite common nowadays.

I can't agree with excluding all GNOME users from using an application, since it's not the fault of GNOME users in general that it's difficult to transition between major GTK versions.

Eh, GTK as a whole is… well, not healthy for cross-platform development. It used to be a good deal better (though it was never really excellent), but from GNOME 3 they’ve pretty much decided to flat-out ignore other platforms and their conventions, and shift GTK from being GNOME-focused but fairly platform agnostic with pretty good look-and-feel theming to being just GNOME GNOME GNOME and wildly out of place in look and feel everywhere else.

Of course, other options in a similar space like Qt aren’t entirely better in that they don’t behave fully like GNOME on GNOME platforms. The best cross-platform native-feel apps, like Firefox, tend to use GTK on Linux (or possibly KDE/Qt in KDE?) and other things on other platforms.

As a Linux user I inevitably use quite a bit of GTK software (though barely any GNOME stuff, if you savvy the distinction I’m drawing), but I dislike almost all of the directions it’s headed in since GNOME 3. Decent scaling is one of the few redeeming features.

sprash is describing a very—ahh—ideologically pure position, but I do deeply sympathise with a desire for a return to the GTK+ 2 ways. My ideal would probably be to scrap GTK+ 3 and 4 and create 5 derived mostly from GTK+ 2’s philosophies, just including the good technical parts of 3 and 4, but determinedly not a full style unto itself. Also I’m fed up with client-side decorations matter and want better support for server-side decorations, as was done traditionally and as is still generally preferable on Windows and in most Linux window managers, or even platform-matching client-side decorations, which would be good for macOS and Windows and frankly GNOME.

GNOME has basically sabotaged GTK as a cross-platform library—even as a Linux-wide thing (apart from KDE). They’ve taken it over to serve their own purposes and vision. Their vision does improve and unify some parts of GNOME desktops, but as a user of something other than GNOME (Sway), I don’t like most of what they’ve done to GTK.

Doesn't GTK4 address your complaints? To a certain degree, I think they've realized what you say. GTK4:

- decouples GNOME and GTK. They've created a new library, called libadwaita, that hosts the GNOME stuff.

- brings big performance improvements (it has GL and Vulkan backends)

- is much better on macOS and Windows.

Anyway, I agree with you on your points. One of the issues of opensource graphic programs is that the toolkits that are available are quite problematic/don't care about this use case. This is not just about GTK, as QT has also been causing a lot of headaches to the Krita developers. It is not by chance that Blender, which draws the UI using its own toolkit, is not affected by these issues and has a fantastic, fast UI.

Not the parent commenter, but I find GTK4 worse than GTK3 on non-GNOME Linux environments. The popovers capture input and hang your entire Xorg session until you press Esc (affecting EasyEffects), GTK4 has window creation race conditions breaking title bars on KWin and window positioning on KWin and xfwm4, and line height and text y-coordinates are non-integer and text is rendered at fractional vertical coordinates, resulting in blur even with hinting on (which assumes integer-pixel vertical positioning).

I might need to take a closer look at it at some point. I’ve never used anything that uses it, and I haven’t inspected it or grasped the scope of libadwaita, but most times I’ve heard anything about GTK 4, there have been major bad things—selection bias, to be sure, and perhaps confirmation bias, but it may indicate something. The three examples that come to mind immediately:

• They severely damaged text rendering and initially didn’t even acknowledge it as a defect, dismissing the report: https://gitlab.gnome.org/GNOME/gtk/-/issues/3787. More recently there’s some movement on the issue, but the initial reaction was not a good look, at least.

• Someone said various bad things about libadwaita here on HN a couple of weeks ago: https://news.ycombinator.com/item?id=30027002.

• Of accessibility, they’ve completely redesigned it, in a way that is very convincingly better, or will be once it’s all finished including supporting tooling, but the impression I’ve received from a couple of communities where it’s come up a bit at times is that at least when they released GTK 4 and told people “start adopting it!”, GTK 4 apps were opaque to screen readers, because half the plumbing was missing. This impression could be false, and if there was any lack, hopefully it’s addressed by now.

I’m curious what the state of affairs is for approximating native look and feel on Windows. GTK+ 2 had a once-fairly-decent win32 theme that became fairly significantly out of date after Windows XP, but there were also some more recent themes that did parts better (and other parts worse). I’ve never seen a Windows theme for GTK+ 3 that is even halfway decent for look, and feel is a dead loss (e.g. menu keyboard behaviour and list box popup positioning and behaviour are both just painfully wrong).

If they’re serious about cross-platform, attempting to match Windows and macOS look and feel (and completely ignoring Linux, for a while) is where effort should be placed. I don’t know if they are.

On text rendering, I've seen that too and I agree with what you say, unfortunately. There is a difference between being professional/straight to the point, which I appreciate, and being dismissive.

On the other topics, I'm not informed enough, but at least it's good some work on the accessibility front is happening, compared to the dire situation of the last decade... I hope that eventually the situation improves. Accessibility improvements really benefit everyone.

Regarding the matching of the look and feel... in the last few years, this mindset has become very common, where each app is an island and no effort should be made to look native, because, in fact, there is no native look anymore. Windows has always been like this, and Linux desktops have started embracing that. From the app developer perspective, it's understandable. For the user, it means giving up decades of usability studies and the pleasantness of coherence. Given this direction, I expect GTK apps to look like GTK apps wherever they run, unfortunately. On the bright side, the default look of GTK4/adwaita apps is kinda similar to what macOS and Windows 11 are going for, so at least it won't look too alien over there.

I hate this ivory tower way of developing. It holds back progress, learning by exchange, and so on.

To avoid the extreme slowdown (and crash after a while) on win10, close the document property tab (!!)

Anyway, inkscape has really disappointed me with multiple serious regressions. This + UI issues + layout problems + zoom buttons disappearing + crash due to version incompatibility + not remembering what unit you use + DPI screwup got me wondering if the people in charge really got what it takes to run the biggest FOSS creativity tool.

I mean, I'm grateful if exists. But I use this for technical documents at work and I need it to be stable. Oh well, I guess it's between this and being lured into 12 month contracts with Adobe...

It's easy to forget that Inkscape is run purely with volunteers, is a charity with 0 employees. I get that these problems are frustrating but maybe it would be more appropriate to wonder how is even possible that this project runs so well :D. Some technical problems are outside our power. Mainly performance regression was introduced by gtk3. Which was a big blow to the whole project because it took 5years of work. And in the end, we discovered that technology is just slow and we need to migrate once again if we want to fix it. Bugs are not ignored we just simply don't have enough manpower to fix them all

I hear you, and as a FOSS maintaineryaelf I agree that it is frustrating to see all complaints but no volunteers.

But, many issues in Inkscape are due to bugs introduced by changes that were not properly tested. If you have resources to change things around, you should also test the outcome properly before you hit that release button.

i hope you dont take this wrong way but ... what you are suggesting is if you cannot do it ""proplerly'" dont do it at all ... testing such a huge project as inkscape is a lot of work. perosnaly i do it a lot but sometimes you just cannot catch it all. "We" dont have receorses we have just developers that are sometimes work on stuff

I've used Inkscape for MacOS for many years off and on. Thank you very much for having made that possible.

Recent versions have been very difficult to use because of the performance degradation. But my gratitude is not diminished even if I am basically forced to find something else.

Cross-platform GUI is an incredibly difficult problem because 1. GUIs are huge and complex, and 2. Vendor lock-in is in the interest of the platform purveyors. GTK exists in a challenging space, and I empathize not only with the downstream users such as Inkscape affected by its troubles but with the GTK developers.

If it were easy to produce a featureful, performant, and developer-friendly cross-platform GUI library, someone would have done so. The demand is so acute that many projects even resort to the outrageous brute-force approach of Electron.

I love all FOSS cross-platform GUI endeavors, even when they get tripped up. You are doing great and important work — I salute you.

That sucks about GTK. Probably would make more sense to use Qt. It's much better.

Both Wireshark and VLC successfully migrated to it, though I guess they have somewhat simpler GUIs.

grass is always greener on a other side. And rewriting to Qt would be as big as starting new projcet from scratch

> grass is always greener on a other side

Not really. Things are not all equally good. I don't see any effort from Wireshark or VLC to move back to GTK.

> And rewriting to Qt would be as big as starting new projcet from scratch

Well that's clearly an exaggeration but I agree it would be a huge project. I'm just dreaming...

Do you think you or the parent commenter can look into debugging or fixing the bugs yourself? I'm also disappointed at buggy software, poorly considered changes, and maintainers ignoring bug reports. But (sometimes) I try (and occasionally succeed) to fix bugs I encounter.

Debugging and contributing to large and complex C++ projects is not trivial. I personally do debug and contribute to open source projects using technologies I’m confident with, but when it comes to Inkscape I can’t do much more than to do nothing and to hope for the best.

Filing reproducible issues is probably a good start and anybody can do it

Yeap! As a developer, having a well described bug, with clear steps and conditions, is sometimes 90% of the work needed to fix it.

Not sure about how the story with Inkscape looks, but it can be difficult to deliver the reproducible issue. You might have to build it yourself, might have to install a compiler in the right version and so on. Depending on the build story of Inkscape, this can take days of fumbling. Not many users are willing to go through that.

I would say it is OK to open an issue, which is not 100% reproducible, as long as it has not been opened. It allows for the chance, that others can try to find that reproducible way of causing the issue, who are more knowledgable about the software and might have the dev setup already on their machines.

A good issue report requires some knowledge and expertise though. Anyone can say it doesn’t work as expected, but it’s harder to identify and reproduce the problem and provide all the useful details.

Affinity Designer is a pretty good tool for vector drawings (for 55$ unique payment). It fills the gap between Inkscape and Illustrator.

Affinity seems like it doesn't have some of the more precise layout tools and SVG-editing capabilities that Inkscape does.

I took the trial after seeing people suggest it as an alternative, but it didn't quite replace my workflow and I ended up dropping it.

When I was still using Windows I was using corel draw. I still believe it was (and I suspect still is) the best vector drawing program for more technical drawings. Just one example:alignment is really deeply build into the system using a set of keyboard shortcuts so instead of always going to the align and distribute tab you just pressed "e" to align the previous object. It was really so much faster for drawing than illustrator or inkscape.

Unfortunately there is no Linux version otherwise I'd happily buy it.

Check Boxy SVG for an SVG-focused vector graphics editor. Its UI is heavily inspired by Inkscape and the Linux version is available for free (you just have to run "snap install boxy-svg").

Thank you! Looks like exactly what I needed.

You're right, the power of Inkscape is its ability to work natively with SVG capabilities and its technical aspects (instead of some obscure proprietary format). Affinity is more artist-focused.

I've been working with graphics for a couple decades and I have many old files in formats that I can no longer open. I am strongly biased towards Inkscape for its use of an open standard as its file format.

I'm on MacOS where current versions of Inkscape are not really usable because of (understandable) performance degradations. I wish that either those problems could be solved, or that an alternative existed which used an open file format.

I'd prefer a pure FOSS solution, but as I'm a BSD-school FOSS person I wouldn't mind a proprietary app so long as that proprietary app used an open file format. It sucks that Affinity uses a proprietary format which is at high risk for orphaning or degradation either if they decide it's in their business interests or if the company ceases to exist.

I'm pleased with their development pace. Every new release has lots of good changes, plus I believe they are fighting technical cruft under the hood.

Oh, that's a fun one; my document property tab disappeared once and I wasn't able to retrieve it. I don't know if that was when I started having issues, but no amount of reboots without the document property tab felt like it fixed it for me.

On a Linux 4K Wayland desktop, Inkscape has been slower a year or two ago, this last 6 months it has become really fast panning and zooming even fairly complex pages.

The guys there have been doing great work. Maybe they need more helping hands that care about windows.

I've only used post 1.x versions and they've had a lot of freezing issues but I thought that's just the way the program was. After seeing this I'm going to try 0.92 and see if it's smoother

I've been plagued with lagging and crashes on Linux too in the last few months.

But then I rebuilt from master in an attempt to get a traceback or profile to report and it went away, so hopefully whatever that was has been fixed and just got released.

Though I have to say I think hiding the snap controls behind a button, as opposed to a toolbar, is a mistake.

I had marginal performance degradation, but that may be due to the specific techniques I use it for, but there definitely was a change southward, just not anywhere nearly as poor as I'm hearing from others.

Inkscape is still very usable/utile, but I certainly don't want that trend to continue, as it would rapidly become not so.

Same issue on OS X

Why use windows? Its bloatware with zero concern for your privacy. Dual boot into Linux, where inkscape is butter. 2cents.

Reply on news.ycombinator.com