Chrome on Windows 8.1

[superleiw]

On 2/10/2025 at 12:16 PM, mjd79 said:

Opera will probably work up to around version 110 (Chromium 125?) with DWrite.dll from Windows 10, replacing DiscardVirtualMemory with VirtualAlloc (or what I also recently discovered, DiscardVirtualMemory is in 8.1 in kernelbase.dll, so you can add by ExportTableTester redirection to this function in kernel32.dll exports and using it for opera_browser.dll e.g. as kernel33.dll) and with --no-sandbox. This is completely pointless due to depriving the browser of security, but for viewing DRM content you can do so. In my case, interestingly enough, DRM works on 8.1 after copying the widvine from Firefox to any compatible browser on Chromium.

I had done some tests with Opera and it worked until version 112 with the modifications. Version 113 didn't work anymore because it started using Chromium 126.

https://msfn.org/board/topic/184454-last-version-of-software-for-windows-81/page/10/#findComment-1274614

 

[D.Draker]

22 hours ago, superleiw said:

I had done some tests with Opera and it worked until version 112 with the modifications. Version 113 didn't work anymore because it started using Chromium 126.

https://msfn.org/board/topic/184454-last-version-of-software-for-windows-81/page/10/#findComment-1274614

 

It's because in 126 they introduced the so much hated by @NotHereToPlayGames UI, and I agree, it's ugly. No way you're getting it to work on 8/1 due to the missing elements.

It's actually the same with Vista. First, you'd need to tool back the old one,

[D.Draker]

On 2/10/2025 at 7:16 AM, mjd79 said:

replacing DiscardVirtualMemory with VirtualAlloc

This leads to a huge memory leak overtime. Which Braunovitch wasn't able to fix.

[Karla Sleutel]

21 hours ago, D.Draker said:

This leads to a huge memory leak overtime. Which Braunovitch wasn't able to fix.

Thanks, now this mystery is solved, at least we know the reason. All roads for Win 8 lead ro Supermium, I guess. I still not want to use it. And seeing its streaming problem, it adds more to the point.

[mjd79]

I finally got to what's messing up the Chromium 110-126 sandbox on Windows 8.1, and it's the UpdateProcThreadAttribute feature. When I redirected it in the original kernel32.dll to the pwrp_k32.dll included in Supermium 132 R2, the sandbox started working.

Just find a 100% sure solution to the RAM and virtual memory usage problems, and you will be able to use any Chromium-based browser (except Edge) up to version 126. I still don't know what is causing the problem with Chromium 127 and newer preventing the browser from running.

And fixing the sandbox in version 126 is just a milestone. It is 17 versions higher than 109, but it is also 9 versions older than the latest official Google Chrome.... If anyone knows how to debug, I would greatly appreciate your help. All I know is that the problem is definitely in chrome.dll.

126 also worked on 8.0, but the sandbox after the repair still does not work, the browser starts up, but with the error SBOX_FATAL_CLOSEHANDLES, and does not display any page, of course.

Edited April 3 by mjd79

[D.Draker]

On 4/3/2025 at 8:19 AM, mjd79 said:

to debug, I would greatly appreciate your help.

Goof ol' days I used https://en.wikipedia.org/wiki/SoftICE

.... but it's only up to XP.

Some say https://www.ollydbg.de/ is somewhat fine, but the developer dumped the project, therefore leaving the 64 bit version incomplete.

 

[mjd79]

I found a “solution” based on Blaukovitch's crack. It turns out that all it takes is to change a few bytes in chrome.dll (unfortunately, completely different in each Chromium-based browser and in other versions) and the original Chrome 130 works, including the sandbox fixed by my method described above. @D.Draker, could you check this with v127 on Vista? For Chrome (not Chromium, etc) 127.0.6533.73 x64, just replace offset 621D030 in chrome.dll with: B8 10 00 00 00 C3 90 48 8B 05 02 E6 8B 06 48 31. Also need to binary change the import in chrome.dll to API-MS-WIN-CORE-REALTIME-L1-1-0.dll and remove the "precise" addition from the QueryUnbiasedInterruptTimePrecise function. It's still not a complete success as I don't fully understand what I'm changing, hence I'm not able to port it to anything other than an identical version of Chrome.

 

Edited April 4 by mjd79

[D.Draker]

18 hours ago, mjd79 said:

@D.Draker, could you check this with v127 on Vista? For Chrome (not Chromium, etc) 127.0.6533.73 x64

Wouldn't it need Ex. Kernel? I don't want to install it, as it adds brightness to the system.

[mjd79]

7 hours ago, D.Draker said:

Wouldn't it need Ex. Kernel? I don't want to install it, as it adds brightness to the system.

I have no idea, Chrome 125 failed to run even with the 9th March 2023 extended kernel, so does your version run with or without it? Unfortunately, it is not possible to redirect ntdll locally, so if the extended kernel is required, it is impossible to avoid installing it.

Edited April 4 by mjd79

[D.Draker]

On 4/3/2025 at 8:19 AM, mjd79 said:

sure solution to the RAM and virtual memory usage problems

Haven't I told you in private with what you need to replace the buggy, old and obsolete Win2000 era VirtualAlloc?

[D.Draker]

19 hours ago, mjd79 said:

I have no idea, Chrome 125 failed to run even with the 9th March 2023 extended kernel, so does your version run with or without it? Unfortunately, it is not possible to redirect ntdll locally, so if the extended kernel is required, it is impossible to avoid installing it.

Chrome 126 will surely fail on Vista without the ported Dwrite.dll from Win10 and "platform update" from 7 (forgot the KB number), be it with or without the Ex. Kernel by win32,  not to mention the Ex. Kernel is literally obsolete and even got unsafe to use since the development had been dropped several years ago. I trusted win32ss for many years, after the case with Supermium, not anymore, I'm not getting his severely outdated kernel installed just to try, sorry.

[D.Draker]

19 hours ago, mjd79 said:

does your version run with or without it?

My version is based on what was partially used in Kernel32 before, I don't need someone else's victories, but user32 and shell32, all by myself.

Redirecting to VirtualAlloc is not clever, I was always against it, I warned win32ss. Here we see the result. They all are in love with VirtualAlloc.

https://github.com/win32ss/supermium/issues/1304

Most likely, Chinese student (as he claims) E3kskoy7wqk used another, more modern approach, and his ported browsers, on either 8 or Vista, do NOT have such terrible leaks like Supermium or Blyaunovitch, I already posted screenshots.

[mjd79]

1 hour ago, D.Draker said:

Haven't I told you in private with what you need to replace the buggy, old and obsolete Win2000 era VirtualAlloc?

I tried, in place of VirtualAlloc in the kernel32 import table I can't fit this function (too long name), and redirecting it in the original kernel I use for Chromium on 8.1 causes an error in ntdll, which is very strange...

BTW Memory usage alone does not seem to be an issue in 8.1, with the same number of tabs with the same pages the v130 uses a similar amount of memory to Chrome 109 on Win7.

The bigger problem is with virtual memory, after some time, such as an hour of using the browser suddenly starts to occupy all the virtual memory.

Edited April 4 by mjd79

[mjd79]

52 minutes ago, D.Draker said:

Chrome 126 will surely fail on Vista without the ported Dwrite.dll from Win10 and "platform update" from 7 (forgot the KB number), be it with or without the Ex. Kernel by win32,  not to mention the Ex. Kernel is literally obsolete and even got unsafe to use since the development had been dropped several years ago. I trusted win32ss for many years, after the case with Supermium, not anymore, I'm not getting his severely outdated kernel installed just to try, sorry.

Despite the backported DWrite and the platform update (which is required by ext kernel), I have not been able to run anything beyond Chromium 111.0.5550 probably.

Edited April 4 by mjd79

[D.Draker]

16 hours ago, mjd79 said:

Despite the backported DWrite and the platform update (which is required by ext kernel), I have not been able to run anything beyond Chromium 111.0.5550 probably.

And this 111 version is most likely all those spring-offs like Supermium, Thorium are based on, whereas the version made by Prybludovich is real, but it has a heavy memory leak of which Prybludovich honestly writes at guthub, and Prybludovich agrees with me about the reason -  VirtualAllo "c.x64/x86. x64 still affected by VirtualAlloc (HUGE MEMORY) serious bug."

And people also write about the same in Supermium, plus CPU overload.

https://github.com/win32ss/supermium/issues/1307

Therefore, for Vista-8.1 I advise to use the Chinese student honest, true port.