Posts tagged ‘Chromium’


How to delete Windows 10 system fonts for real, not just remove registry references to them

13.08.2020

My last post implied that I ego-surfed and found a Wikipedia chat entry about me, but that’s not the case. I was searching for information on how to remove a system-protected font from Windows 10, and seeing as I often post solutions to obscure technical issues on here, I had hoped I recorded my how-to last time. The libel posted by some Australian Wikipedia editor came up during that search.
   Once upon a time, Microsoft didn’t care if you removed system fonts, but at some point, it began protecting Arial, whose design, for reasons I’ve gone into elsewhere, I’ve always considered compromised. There was one stage where you could replace Arial with something else called Arial, and as I had a licence for a very, very old Agfa version of Helvetica (do people remember CG Triumvirate?!), I decided to modify its file name to fool Windows into thinking all was well.
   The last time Windows did an update—version 1909—I had to resort to a safe-mode boot and taking control of the font files as admin, but I really could not remember the specifics. The problem is that when you install the ā€œnewā€ Arial, the existing roman one is used by quite a few applications, and you don’t really replace it—your only solution is to delete it.
   With version 2004, safe mode is quite different, and the command prompt and Powershell commands I knew just didn’t cut it. I realize the usual solution is to go into the registry keys—I’ve used this one for a long, long time—and to remove or modify the references to the offending fonts at HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Fonts. I’ve also used the HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\FontSubstitutes key to make sure that Helvetica does not map on to Arial (in fact, I make sure Arial maps on to Helvetica). Neither actually works in this case; they are ignored, even bypassed by certain programs. And, really, neither deletes the file; they just attempt to have Windows not load them, something which, as I discovered, doesn’t prevent Windows from loading them.
   By all means, use these methods, but be prepared for the exception where it doesn’t work. The claim that the methods ā€˜delete’ the fonts is actually untrue: they remain in C:\Windows\Fonts.
   The other methods that do not work are altering the equivalent keys under WOW6432node (which get intercepted and directed from the 32-bit keys anyway), using an elevated command prompt to delete the files (at least not initially), or doing the same from safe mode (which is very different now, as safe mode is in the same resolution and the Windows\Fonts folder displays as it does in the regular mode—so you cannot see the files you have to remove). You cannot take ownership of the font files through an elevated Powershell (errors result), nor can you do this from safe mode. Nothing happens if you delete FNTCACHE.DAT from the system32 directory, and nothing happens if you delete ~fontcache files from the Local directory.
   What was interesting was what kept calling arial.ttf in the fonts’ directory even after ā€œmyā€ Arial was loaded up. The imposter Arial loaded in most programs, but for the Chromium-based browsers (Vivaldi, Edge), somehow these knew to avoid the font registry and access the font directly. This was confirmed by analysing the processes under Process Monitor: sure enough, something had called up and used arial.ttf.
   This Wikihow article was a useful lead, getting us to delete the fonts under the Windows\WinSxS folder, and showing how to take ownership of them. I don’t know if altering these ultimately affected the ones inside Windows\Fonts, but I followed the instructions, to find that the original Arial was being accessed by three programs: Vivaldi, Keybase, and Qt Qtwebengineprocess. I shut each one of these down and removed the Arial family.
   Reboot: it was still there. Then it hit me, and I posted the solution in the Microsoft Answers forum (perhaps inadvertently prompting a Microsoft programmer to make things even harder in future!). Another user had told me it was impossible, but I knew that to be untrue, since it had been possible every other time.
   The solution is pretty simple: since you can’t see the full Windows\Fonts directory with Windows Explorer, then I needed another file manager.
   Luckily, I had 7zip, which I opened as an administrator. It allowed me to go into the folder and view all its contents, not just the fonts called up under HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Fonts, which we know is not an accurate representation of the fonts being used by the system. From there I could finally delete the offending four fonts without changing the ownership (which makes me wonder if the Wikihow advice of changing the owner under Windows\WinSxS wound up affecting the Windows\Fonts files). Once again, I had to close Keybase, Vivaldi and Qt Qtwebengineprocess.
   It took from c. 4 p.m., when my desktop PC updated to v. 2004 (my laptop had been on it for many weeks; soon after its release, in fact) to 2 a.m., with a break in between to cook and eat dinner. I’m hoping those hours of having typographic OCD helps others who want to have a font menu where they determine what they should have. Also, user beware: don’t delete stuff that the system really, really needs, including an icon font that Windows uses for rendering its GUI.


Using Google as a last resort—except this search, which I did again as an illustration, now displays in CG Triumvirate rather than Arial. Normally, Google is a big Arial user (Arial and sans-serif are in the CSS specs) and Chromium browsers are all too happy to circumvent the registry-registered fonts and go straight into your hard drive.


You may also like

Tags: , , , , , , , , , , , , , ,
Posted in design, technology, typography | 9 Comments »


Fixed Vivaldi’s poor type display, thanks to wmjordan

07.03.2020

It took two months but I finally got there.
   Vivaldi now displays type normally though the browser interface is slightly messed up. But I’ll take good type display, thanks.
   On the MacType forums, a user in China called wmjordan was in the same boat but had found a solution. In their words:

For the recent version of Vivaldi 2.10, 2.11, you need to create a shortcut, and modify the command line, append the "--disable-lcd-text" parameter behind the executable name, and MacType will work on the web page content window. The "--disable-features=RendererCodeIntegrity" parameter is recommended by snowie2000.

my command line:

vivaldi.exe --disable-lcd-text --disable-features=RendererCodeIntegrity

   I used the latter method, but the type was still quite poor for me. I had to do one more thing: start Vivaldi in Windows 8 compatibility mode.
   It’s messed up the top of the browser a little but it’s a small price to pay to have everything readable again.
   Snowie2000, the main dev for MacType, says a registry hack is their preferred workaround, at github.com/snowie2000/mactype/wiki/Google-Chrome#workaround-for-chrome-78.
   It turns out that Chrome 78 (and presumably Chromium 78, too) did indeed have a change: ā€˜Starting from Chrome 78, Chrome began to block third-party DLLs from injection. But they provided a way to disable the protection either from the command line or by policy.’
   I was right to have investigated which version of Vivaldi represented the change earlier (it was 2.9, which equated to Chromium 78). After testing wmjordan’s suggestions out on 2.9, I upgraded to 2.11, and it was still fine.
   Opera GX is still the more resolved browser (works as it should out of the box) but there are some aspects of Vivaldi that I’m familiar with after two-and-a-half years (to the day). Looks like I’ll be going back to it for my main browsing, but I know I’ve found another great browser along the way, and I’ve updated my Firefox, too.


You may also like

Tags: , , , , , , , ,
Posted in China, technology, typography | No Comments »


Directwrite isn’t the culprit

03.03.2020

That was confusing. Yesterday’s blog post was representative of my thinking: given that certain people were upset when Chromium took away the Directwrite toggle in 2016, and type rendering on Chromium-based Vivaldi deteriorated significantly for me with v. 2.10 (it turns out v. 2.9 was the turning-point), then did Chromium only switch fully to Directwrite for me earlier this year? Luckily I wrote a caveat: ā€˜There’s a possibility that what I saw from 2017 actually was Directwrite, and whatever they’re using now is yet another technology that no one has made any note of.’
   Snowie2000, one of the developers of MacType, suggested I try Cent Browser, arguably the only Chromium browser that still has a Directwrite toggle: you could still disable it in favour of GDI.
   Cent Browser by default is marginally better than what I was seeing on Edge, Vivaldi 2.10 and others, but once I turned Directwrite off, I saw a very different display, with far heavier type.

Cent Browser, Directwrite switched off

Cent Browser, default

Edge

   It wasn’t what I expected to see, and without taking issue with those who support GDI rendering in Chromium, it lacked fidelity (at least for me) with what the type looked like in print. I can see clearly why it has its adherents: it is superior to the default. But, in other words, what I experienced on Vivaldi between 2007 and January 2020 was using Directwrite, and whatever is going on now is using something else, or ignoring other settings on my PC.
   Yesterday I theorized that if the change happened between Chromium 77 and 78, then I should see that in the source browser. I installed a v. 77 from the repository. As you know, these are stand-alone and can run without a full installation. What I saw was the inferior rendering, so the ā€œswitchā€ didn’t happen then. It may have happened, as I was told on the Vivaldi forums, with Chromium 69, something I am yet to confirm.
   Therefore, whatever Chromium is doing isn’t something that’s been documented, to my knowledge, except for here. And Opera and Opera GX, if they are based on Chromium 79, seem not to be afflicted by this bug. Or they are interacting with other programs I have in order to keep the type rendering faithful, with decent hinting and contrast.
   The question is: what is causing the far inferior type display on Chromium today?

PS.: Trials on Chromium 68 and 69—they’re the same (i.e. poor type display). This may have gone on for quite some time.—JY



You may also like

Tags: , , , , , , ,
Posted in design, internet, technology, typography | No Comments »


Has Directwrite arrived on my Chromium-based browsers four years after everyone else?

03.03.2020

After considerable searching, the bug that I reported to Vivaldi, and which they cannot reproduce, appears to be one that the general public encountered back in 2016, when Chromium took away the option to disable its Directwrite rendering. I don’t know why I’ve only encountered it in 2020, and as far as I can tell, my experience is unique.
   It’s a good position to be in—not unlike being one of two people (that I know of) who could upload videos of over one minute to Instagram without using IGTV—though it’s a mystery why things have worked properly for me and no one else.
   When I switched to Vivaldi in 2017, I noticed how the type rendering was superior compared with Firefox, and it was only in January this year when it became far inferior for me. Looking at the threads opened on type rendering and Chromium, and the screenshots posted with them, most experienced something like this in 2016—a year before I had adopted Vivaldi. If my PC worked as theirs did, then I doubt I would have been talking about Vivaldi’s superior display.
   There’s a possibility that what I saw from 2017 actually was Directwrite, and whatever they’re using now is yet another technology that no one has made any note of.
   I’ve posted in the Vivaldi and MacType forums where this has been discussed, as my set-up could provide the clue on why things have worked for me and not others. Could it be my font substitutions, or the changes I’ve made to the default display types in Windows? Or the fact that I still have some Postscript fonts installed from the old days? Or something so simple as my plug-ins?
   Tonight I removed Vivaldi 2.11 and went to 2.6. I know 2.5 rendered type properly—Bembo on the Lucire website looks like Bembo in print—so I wondered if I could narrow down the precise version where Vivaldi began to fail on this front. (As explained earlier, after 2.5, no automatic updates came, and I jumped from 2.5 to 2.10.)
   It was 2.9 where the bug began, namely when Vivaldi moved from a Chromium 77 base to a 78 one. This is different to what Ayespy, a moderator on the Vivaldi forums, experienced: version 69 was when they noted a shift. Yet Opera GX, which works fine, has a browser ID that claims it’s Chrome/79.0.3945.130 (though I realize they can put whatever they like here). Brave, Chrome and Edge look awful.
   We can conclude that not all Chromium browsers are created equally (goes without saying) but I understand that the rendering isn’t something that each company (Vivaldi, Opera, etc.) has fiddled with. Therefore, something I’m doing is allowing me to have better results on Opera, Opera GX and Vivaldi versions up to 2.8 inclusive.

Vivaldi 2.8

Opera GX

Firefox Developer Edition 74.0b9


You may also like

Tags: , , , , , , , ,
Posted in design, internet, technology, typography | 1 Comment »


Returning to Firefox?

17.02.2020

I wonder if it’s time to return to Firefox after an absence of two years and five months. After getting the new monitor, the higher res makes Firefox’s and Opera GX’s text rendering fairly similar (though Chrome, Vivaldi and Edge remain oddly poor, and Vivaldi’s tech people haven’t been able to replicate my bug). There’s a part of me that gravitates toward Firefox more than anything with a Google connection, and I imagine many Kiwis like backing underdogs.
   Here are some examples, bearing in mind Windows scales up to 125 per cent on QHD.

Vivaldi (Chrome renders like this, too)

Opera GX (and how Vivaldi used to render)

Firefox

   Opera renders text slightly more widely than Firefox, but the subpixel rendering of both browsers is similar, though not identical. Type in Firefox arguably comes across with slightly less contrast than it should (especially for traditionally paper-based type, where I have a good idea of how it’s ā€œsupposedā€ to look) but I’m willing to experiment to see if I enjoy the switch back.
   In those 29 months, a lot has happened, with Navigational Sounds having vanished as an extension, and I had to get a new Speed Dial (FVD Speed Dial) to put on my favourite sites. FVD uninstalled itself earlier today without any intervention from me, so if that recurs, I’ll be switching to something else. I don’t like computer programs having a will of their own.
   A lot of my saved passwords no longer work, since I change them from time to time, and it was interesting to see what Firefox remembered from my last period of regular use. I’ll have to import some bookmarks, too—that file has been going between computers since Netscape.
   The big problem of 2017—Firefox eating through memory like crazy (6 Gbyte in a short time)—could be fixed now in 2020 by turning off hardware acceleration. It’s actually using less right now than Opera GX, and that’s another point in its favour.
   I also like the Facebook Container that keeps any trackers from Zuck and co. away.
   I did, however, have to get new extensions after having resided in the Vivaldi and Opera space for all that time, such as Privacy Badger.
   If I make Firefox the default I know I’ll have truly switched back. But that Opera GX sure is a good looking browser. I might have to look for some skins to make common-garden Firefox look smarter.


You may also like

Tags: , , , , , , , , , ,
Posted in design, internet, typography | No Comments »


Sticking with 24 inches, but going to QHD: a pleasant upgrade

31.01.2020


The Dell P2418D: just like the one I’m looking at as I type, but there are way more wires coming out of the thing in real life

Other than at the beginning of my personal computing experience (the early 1980s, and that’s not counting video game consoles), I’ve tended to have a screen that’s better than average. When 640 Ɨ 360 was the norm, I had 1,024 Ɨ 768. My first modern laptop in 2001 (a Dell Inspiron) had 1,600 pixels across, even back then. It was only in recent years that I thought my LG 23-inch LCD, which did full HD, was good enough, and I didn’t bother going to the extremes of 4K. However, with Lucire and the night-time hours I often work, and because of a scratch to the LG that a friend accidentally made when we moved, I thought it was time for an upgrade.
   Blue light is a problem, and I needed something that would be easier on the eyes. At the same time, an upgrade on res would be nice.
   But there was one catch: I wasn’t prepared to go to 27 inches. I didn’t see the point. I can only focus on so much at any given time, and I didn’t want a monitor so large that I’d have to move my neck heaps to see every corner. On our work Imacs I was pretty happy to work at 24 inches, so I decided I’d do the same for Windows, going up a single inch from where I was. IPS would be fine. I didn’t need a curved screen because my livelihood is in flat media. Finally, I don’t need multiple screens as I don’t need to keep an eye on, say, emails coming in on one screen, or do coding where I need one screen for the code and the other for the preview.
   Oddly, there aren’t many monitor manufacturers doing QHD at 24 inches. There was a very narrow range I could choose from in New Zealand, with neither BenQ nor Viewsonic doing that size and resolution here. Asus has a beautifully designed unit but I was put off by the backlight bleed stories of four years ago that were put down to poor quality control, and it seemed to be a case of hit and miss; while Dell’s P2418D seemed just right, its negative reviews on Amazon and the Dell website largely penned by one person writing multiple entries. I placed the order late one night, and Ascent dispatched it the following day. If not for the courier missing me by an hour, I’d be writing this review a day earlier.
   I realize we’re only hours in to my ownership so there are no strange pixels or noticeable backlight bleed, and assembly and installation were a breeze, other than Windows 10 blocking the installation of one driver (necessitating the use of an elevated command prompt to open the driver executable).
   With my new PC that was made roughly this time last year, I had a Radeon RX580 video card with two Displayport ports, so it was an easy farewell to DVI-D. The new cables came with the monitor. A lot of you will already be used to monitors acting as USB hubs with a downstream cable plugged in, though that is new to me. It does mean, finally, I have a more comfortable location for one of my external HDs, and I may yet relocate the cable to a third external round the back of my PC.
   Windows 10 automatically sized everything to 125 per cent magnification, with a few programs needing that to be overridden (right-click on the program icon, then head into ā€˜Compatibility’, then ā€˜Change high DPI settings’).
   Dell’s Display Manager lets you in to brightness, contrast and other settings without fiddling with the hardware buttons, which is very handy. I did have to dial down the brightness and contrast considerably: I’m currently at 45 and 64 per cent respectively.
   And I know it’s just me and not the devices but everything feels faster. Surely I can’t be noticing the 1 ms difference between Displayport and DVI-D?
   I can foresee this being far more productive than my old set-up, and Ascent’s price made it particularly tempting. I can already see more of the in- and outbox detail in Eudora. Plantin looks great here in WordPerfect (which I often prep my long-form writing in), and if type looks good, I’m more inclined to keep working with it. (It never looked quite right at a lower res, though it renders beautifully on my laptop.)
   I feel a little more ā€œlate 2010sā€ than I did before, with the monitor now up to the tech of the desktop PC. Sure, it’s not as razor-sharp as an Imac with a Retina 4K display, but I was happy enough in work situations with the QHD of a 15-inch Macbook Pro, and having that slightly larger feels right. Besides, a 4K monitor at this size—and Dell makes one—was outside what I had budgeted, and I’m not sure if I want to run some of my programs—the ones that don’t use Windows’ magnification—on a 4K screen. Some of their menus would become particularly tiny, and that won’t be great for productivity.
   Maybe when 4K becomes the norm I’ll reconsider, as the programs will have advanced by then, though at this rate I’ll still be using Eudora 7.1, as I do today.

Incidentally, type on Vivaldi (and presumably Chrome) still looks worse than Opera and Firefox. Those who have followed my blogging from the earlier days know this is important to me.

Vivaldi

Opera GX


You may also like

Tags: , , , , , , , , , , , , , , ,
Posted in design, internet, publishing, technology, typography | 3 Comments »


In my experience, the only browser that works with Jetstar’s website is Safari for Mac

23.07.2019

I’ve found some forum entries about this, but they date back to the beginning of the decade. I alerted Jetstar to this in March, and the problem has worsened since then.
   Basically, I can’t book online, and I don’t know why. Consequently, I booked one flight with Air New Zealand and only managed, after huge effort, to get the other (for a colleague) with Jetstar.
   Back in March, I couldn’t book with Vivaldi, but I was able to switch to Firefox. I let Jetstar know.
   Now, this strategy does not work.
   Before you suggest it, cookies and caches have been cleared.
   Here’s what happens after I’ve selected the cities and the dates, and I go to select times. Let’s begin with Vivaldi on Windows, which is based on Chromium (which, as we know, is what Chrome, the browser Jetstar suggests you use, is based on):

Switching to Firefox now results in this:

Switching to Edge on the same PC gives this:

   Fortunately, I also own Macs, so here’s what Firefox for Mac returns:

   The only browser that works with the Jetstar website: Safari on Mac. As I’ve sold my Ubuntu laptop, I was unable to test using that OS.
   Not many people would go to that effort, and while Jetstar’s Twitter staff (after some pushing from me in DMs) said they would refer it on, I don’t expect anything to happen.
   Maybe Chrome would work, but I’m not ever going to download it to find out. Why invite Google on to your computer? But if that is the case, it seems foolish to limit yourself to such an invasive browser. My experience is that whatever is blocking me from booking with Jetstar (some may argue that this is a good thing), it is expanding across browsers.


You may also like

Tags: , , , , , , , , , , , , , , , , , , , ,
Posted in business, design, internet, technology | No Comments »