UI regressions in newer release versions, user agent is still wrong

[InterLinked1]

I just tried out the latest Supermium 132, and it seems there are some things that aren't working right anymore.

I remember at some point, something was added (an installation option?) to emulate the old trapezodial tabs in Chromium up through version 70, when it was removed in 71. In 132, this no longer seems to work.

In the release notes, I see "This is the first Supermium release after the removal of legacy Chrome theming code. Some user interface changes have occurred as a result; most notably, the #compact-ui switch mostly covers the side/app menu at this time." It's not fully clear to me if that means this capability isn't available anymore. If so, could we get some clarification on why that was removed? I thought that was kind of one of the main points of Supermium so I'm confused now. I tried the tips in #377 but they don't seem to work now. The compact UI isn't it either. I went back and couldn't find anything about this being removed now, apologies if I missed something.

Also, circling back to issue #57, this issue was "closed" - I can't reopen it but this is still an issue. The user agent still reports "NT 10.0" for some reason, even though I'm using Windows 7. This issue also existed in 126. This makes the browser unusable for me. I am not going to pretend I'm using Windows 10 when I'm not. So for now, I'm still using regular Chrome 109, and incidentally, I've never had a compatibility issue with a site blocking me for my OS - only for my browser version, if that. If this is Supermium-specific logic, I think it should be removed as it's not needed in most cases, and the user agent should be correct by default. I do not want to be part of the OS detection problem by false advertising that I'm using Windows 10 when I'm not. I'd think many users would find this unacceptable as well.

For now, I've reverted to 122.0.6261.152 (R6) due to the UI regressions, as in testing, that seems to be the last one that worked with the old UI. I'm not using it regularly at the moment due to the user agent problem - so far (knock on wood), my health insurance is the only site that doesn't work in Chrome 109 due to some new JavaScript thing there.

If we could get some clarification on the direction of the browser and whether these things can be addressed or not, that would be great! I assumed that one of the goals of Supermium was to patch back in stuff that removed upstream, but now I'm not sure if that's the case given what I've been reading about rebasing on what they are doing upstream. Is the new goal now only to be able to run on older Windows, without retaining any of that older functionality?

[win32ss]

The tabs are in a bit of transition. The tab settings will ultimately be controlled by the "scs" file mentioned in another issue. I should be able to produce "scs" files for certain types of tab metrics soon.

The user agent was "reduced" to NT 10.0 to match all other major web browsers that stopped reporting other Windows versions and avoid fingerprinting of Supermium, especially to avoid the wrath of websites like this. The default user agent can always be overridden using the switch --user-agent.

Supermium continues to feature other legacy features such as FTP support, the download shelf, native Windows tooltips and the legacy titlebar (#force-xp-theme).

[InterLinked1]

The tabs are in a bit of transition. The tab settings will ultimately be controlled by the "scs" file mentioned in another issue. I should be able to produce "scs" files for certain types of tab metrics soon.

Awesome, looking forward to it, I'll be patient and await updates!

The user agent was "reduced" to NT 10.0 to match all other major web browsers that stopped reporting other Windows versions and avoid fingerprinting of Supermium, especially to avoid the wrath of websites like this. The default user agent can always be overridden using the switch --user-agent.

Was this actually a change that occurred in Chromium at some point? In Chrome 109, it reports the correct OS (NT 6.1)

I've never encountered a website that blocks Windows 7 and doubt most users do either. I still think the handful that might run into the scenario should be the ones overriding, rather than overriding it by default and breaking it for everyone. I think this actively does the user base a disservice as many of the remaining Windows < 10 users are now falsely reporting they are using Windows 10, which further reduces the OS usage stats for legacy Windows versions. Then people will look at it and say "oh, look, nobody's using Windows 7 anymore, guess we should not care about that." So Supermium is actively feeding that at the moment, which really bothers me.

Put it another way: every request that's made is like a vote on the web. By defaulting to an override, Supermium users are all voting for Windows 10, not for Windows 7 or whatever OS they are really using. This should really bother any freedom-of-OS-choice advocates.

Supermium continues to feature other legacy features such as FTP support, the download shelf, native Windows tooltips and the legacy titlebar (#force-xp-theme).

Awesome, thanks! Glad these were just exceptions. Thanks again for all your work on this!

[win32ss]

Was this actually a change that occurred in Chromium at some point?

Yes, both Chromium and Firefox did it not long after Chrome 109 was released.

I'm not aware of anything blocking Windows 7 in that way, but it has happened to Vista recently. Complicating things further is that the user agents in Chromium (but not Firefox) are to be deprecated and replaced with "user agent client hints" (UACH) which doesn't even include proper Windows versions in its specification, in order to properly recognize separate Windows 10 releases and Windows 11. So NT 6.x becomes 0.x; 10/11 becomes 1.x. And NT 5.x doesn't exist.

Here is an example of a scs file that implements trapezoidal tabs. However they lose their shape on the right after a few tabs are opened, because the end of the tab is marked by the right edge of the tab close button. I will have to shift the button to the left by subtracting the x-value of the button by a new value corresponding to the "tab close button offset", which would allow the tabs to exist fully within their prescribed space. Unfortunately I'm not able to work on Supermium itself for several days and had to cram in some work originally intended for a later period during the end of February, which means that this will have to do for now (place in the root user profile folder): scs.zip

[InterLinked1]

Was this actually a change that occurred in Chromium at some point?

Yes, both Chromium and Firefox did it not long after Chrome 109 was released.

Okay, that's good to know, albeit disappointing.

Still, I don't think that's a reason to follow suit. Those browsers explicitly no longer support NT < 10, while Supermium's explicit goal is to support NT < 10. So for that reason, I think it should continue to report the traditional user agent.

I want websites to know that I'm using Windows 7, and so are all the other Supermium users, whatever version of Windows they're using.

Would it be terribly difficult to reintroduce the ability to report it as has been done up through Chromium 109?

I'm not aware of anything blocking Windows 7 in that way, but it has happened to Vista recently. Complicating things further is that the user agents in Chromium (but not Firefox) are to be deprecated and replaced with "user agent client hints" (UACH) which doesn't even include proper Windows versions in its specification, in order to properly recognize separate Windows 10 releases and Windows 11. So NT 6.x becomes 0.x; 10/11 becomes 1.x. And NT 5.x doesn't exist.

But it remains that traditional user agents will still be around for quite some time, and probably never really disappear, given all the users using older browsers on older versions of Windows. i.e. all the non-Supermium users will still be sending traditional user agents and proudly reporting what version of Windows is in use.

While I get it's what upstream is going, I think this is very much something we should not be going along with. Client hints also have their own downsides and I suspect many people may continue to want the traditional user agent supported indefinitely, no matter what happens in Chromium.

Here is an example of a scs file that implements trapezoidal tabs. However they lose their shape on the right after a few tabs are opened, because the end of the tab is marked by the right edge of the tab close button. I will have to shift the button to the left by subtracting the x-value of the button by a new value corresponding to the "tab close button offset", which would allow the tabs to exist fully within their prescribed space. Unfortunately I'm not able to work on Supermium itself for several days and had to cram in some work originally intended for a later period during the end of February, which means that this will have to do for now (place in the root user profile folder): scs.zip

Thanks, and understood! I'll keep an eye out for updates on this and continue testing as they come out. Thanks again!

[Snowball-V]

---

First of all, a huge thank you for what you’re doing.

A lot of people fill out issues, and some keep bombarding you with requests, but most forget how lucky we are to have someone dedicating their time to maintaining a Chromium fork for older versions of Windows. And even more so, someone who allows us to customize it!

You could be making a fortune by industrializing forks, but instead, you choose to spend countless hours making the world a better place for anonymous users.

Just as a reminder: there are many of us who deeply appreciate your work on Supermium !

---

Regarding Trapezoidal Tabs

Here is an example of a scs file that implements trapezoidal tabs. However they lose their shape on the right after a few tabs are opened, because the end of the tab is marked by the right edge of the tab close button [...] (place in the root user profile folder): « scs. Zip »

I tested the example file you provided in #1204 (comment), and I can confirm the difficulties in customizing trapezoidal tabs in Supermium, particularly when the browser automatically reduces tab width based on the number of open tabs or the window size.

This automatic shrinking distorts the Bézier curve defining the right side of the tab, which affects the intended appearance.

Based on your example, I deduced the following for creating a trapezoidal tab:

• H = Tab height
• W_Top = Right offset (in pixels) of the tab’s top compared to its base (= slope of the trapezoidal diagonal)
• W_Base = Total width of the tab (= base width of the trapezoidal tab)

Thus, the formula would be:

tab{1=0,H,W_Top/2,H/2,W_Top,0}{2=W_Top,0,W_Base/2,0,W_Base-W_Top,0}{2=W_Base-W_Top,0,W_Base-(W_Top/2),H/2,W_Base,H}

For my specific case (see attached file):

• H = 28
• W_Top = 8
• W_Base = 180

This results in:

tab{1=0,28,4,14,8,0}{2=8,0,90,0,172,0}{2=172,0,176,14,180,28}endtab

Additionally, some constants in the SCS file need adjustment, such as TAB_WIDTH, TAB_BOTTOM_CORNER_RADIUS, TAB_HEIGHT, TAB_STRIP_HEIGHT, etc. (Once again, thanks for making all of this possible ❤️).

---

Question: Can We Use a Variable for Tab Width in the "SCS" File?

For example:

int_layout_constant_α = 28
int_layout_constant_β = 8
int_layout_constant_γ = 180

Then, in the source code, it could be used like this:

tab{1=0,[α],[β/2],[α/2],[β],0}{2=[β],0,[γ/2],0,[γ-β],0}{2=[γ-β],0,[γ-(β/2)],[α/2],[γ],[α]}

However, I think this might not be a great idea since we’d also need to consider TAB_CLOSE_BUTTON_X_OFFSET, which would need to be shifted based on int_layout_constant_γ.

But more importantly, this formula only applies to trapezoidal tabs with sharp/acute angles… There should probably be a variable like "Old_Trapezoidal_Tabs_Appearance" = true / false instead.

---

Final Thoughts

The more I think about it, the more I realize trapezoidal tabs are going to be a nightmare to maintain.

No, I did try asking Copilot, ChatGPT, Claude, DeepSeek, Grok, and 132 others. They are completely off the mark. This is something only a human can figure out for now.

So, I’ll start learning to code and dig through the source to find a possible solution.

That said, your example works beautifully with fewer than 10 tabs in my case, producing a clean and satisfying result.

Thanks again! I’ll be back in 2–3 weeks once I’ve figured out what’s responsible for shrinking the tabs and how to handle it.

---

Few tabs :

Many tabs :

[win32ss]

I was able to find the solution to prevent the right edge of the trapezoidal tab from being cut off on account of reduced tab width; in a yet-to-be-released modification to the tab shape code, I gave the integer k at the beginning of each tab coordinate statement (in the format {k=x1,y1,x2,y2,x3,y3}) scaling properties. This property has the following definitions:

k = 0 : no scaling
k = 1 : scale all x values
k = 2 : scale only x3

(I may change this to a bit mask for scaling individual x values. There is no benefit in scaling y values in this way, which will be done through the other parameters, and will also be corrected to better accommodate the custom tab shapes)

This means that the tab coordinates, in the example of the trapezoidal tab, would have k = 0 on the left-hand-side of the tab, k = 2 for the middle of the tab and k = 1 for the right-hand-side of the tab.

The scaling itself is performed by reducing the x values by a value f obtained from subtracting the actual tab width (represented by tab_->width()) by layout constant TAB_WIDTH. That f value has a minimum of 0 (since tab_->width() can be larger than TAB_WIDTH when only a few tabs are open) and a maximum of TAB_WIDTH - 16 to prevent tabs from becoming too small or disappearing.

This is the result:

However, I have noticed one caveat where the favicon was shifted upward on a few random tabs. This may be rectified by fixing the y value of the favicon to a certain value.

[Snowball-V]

Damn, impressive responsiveness, and the scaling factor idea is actually really smart!

Selfishly, on my end, TAB_FAVICON_Y_OFFSET & TAB_FAVICON_X_OFFSET are already set to fixed values since I use 28-pixel-high tabs (the same size as the circle for adding a new tab).

By the way, TAB_FAVICON_Y_OFFSET & TAB_FAVICON_X_OFFSET didn’t work in Supermium 132.0.6834.191 R0  (chrome.dll / hash 0x0D6EB8F3), but they were fixed in Supermium 132.0.6834.191 R1  (chrome.dll / hash 0x0D6EC16F).

So, in my case, I’ll have to check if "the favicon was shifted upward on a few random tabs"
can be fixed using my constants, which lock the favicon’s position in both X and Y.

And actually, that might also resolve some of the issues I’ve seen here about misaligned favicons.

---

Lastly, I just realized that Raymond Hill / gorhill (the creator of uBlock Origin) is also Canadian. Is it a habit for you 🍁 guys to work this hard and efficiently for the greater good? You two 🫎 should totally team up — that would be a powerhouse collaboration! 🚀

[Wen19752024]

I have no idea when it comes to coding. I do appreciate how much time and effort must go into this. Seems we're still waiting for the trapezoid tab fix to be updated/released in a newer version? That's perfectly fine, I will continue to wait. Just clarifying so I know to keep using version 126. Thanks again for all the time and hard work, win32ss.

[Snowball-V]

I'm on version 132.0.6834.213 R2 (last at this date) and I'm using a custom configuration like this:

tab{0=0,28,4,14,8,0}{2=8,0,90,0,172,0}{1=172,0,176,14,180,28}endtab

This setup is supposed to account for the «  k  » scaling factor and its properties — but the patch hasn’t been applied yet.

So I’d say for now we just have to wait a bit, since this is a minor aesthetic detail and there are still crashes to handle first.

But as always with Supermium, it's only a matter of time — everything ends up being addressed eventually, kind of like in an 🇪🇺 emergency room: you never know who will get handled first, but in the end everyone gets their issue fixed!

[win32ss]

I decided to implement the scaling factor in the form of a bitmask. This means that bit 0 scales x1, bit 1 scales x2 and bit 2 scales x3. So if you want to scale only x2 and x3, k will be 4, and if you want to scale all x values, k will be 7.

A more recent commit made a few days ago will add support for custom new tab buttons.

[InterLinked1]

@win32ss Just out of curiosity, do you know where Chromium removed support for user agents < NT 10.0?

The discussion came up elsewhere and I spent some time looking through the Chromium codebase but couldn't find where that happened, if it did happen. I imagine one of these three files would be the place: https://github.com/chromium/chromium/blob/master/base/win/windows_version.cc

But it seems like all the NT versions are still there from what I can see.

If it really was removed in Chromium, maybe a compromise we could do is add an option to the install. Something like "Use real OS in user agent", which would be enabled by default. If somebody disabled the option, then it would behave as currently and send 10.0.

[InterLinked1]

@win32ss Just out of curiosity, do you know where Chromium removed support for user agents < NT 10.0?

The discussion came up elsewhere and I spent some time looking through the Chromium codebase but couldn't find where that happened, if it did happen. I imagine one of these three files would be the place: https://github.com/chromium/chromium/blob/master/base/win/windows_version.cc

But it seems like all the NT versions are still there from what I can see.

If it really was removed in Chromium, maybe a compromise we could do is add an option to the install. Something like "Use real OS in user agent", which would be enabled by default. If somebody disabled the option, then it would behave as currently and send 10.0.

Okay, I found it: https://privacysandbox.google.com/blog/user-agent-reduction-android-model-and-version

From Chrome 107 we replaced the desktop operating system version and CPU information with a fixed value for the platform.

Mac	➡	Macintosh; Intel Mac OS X 10_15_7
Windows	➡	Windows NT 10.0; Win64; x64
ChromeOS➡	X11; CrOS x86_64 14541.0.0
Linux	➡	X11; Linux x86_64

https://groups.google.com/a/chromium.org/g/blink-dev/c/kC-AeZ1fSdY/m/a_ICuXZRBQAJ

Looks like the flag is still present at least in 109: #reduce-user-agent-platform-oscpu

I did some more looking, looks like it was removed in 113: https://www.chromium.org/updates/ua-reduction/

I think this is the actual commit: chromium@9a0a414

The flags seem to still be there in current master Chromium, so I think we just need to revert that commit and also make the flag work again and then we're groovy! (And bonus points if we can have an installer option that would disable the flag by default) What do you think?