Chrome main thread

[idontyboi]

Since this thread has been made the main Chrome thread (because I figured out the workarounds first?), the Chrome progress as of One-Core-API 3.0.5 is:

• Chrome 103-125 is now working, including WOW64. 126 doesn't work yet.
• Supermium 117 works. 118 requires Windows 7 compatibility to work without issues. Supermium 119 and 121 doesn't work at the moment without a workaround.
• All Chromium derivatives [Opera, Yandex, Vivaldi, Opera GX] work fine.

Known Issues with Chromium

SwiftShader WebGL crash in XP

Due to a issue with fibers, WebGL crashes immediately when it's used. The fix is to use the switch --disable-webgl. Various websites use this for... fingerprinting.

It's untested if NT 5.2 is affected: XP is definitely affected This is due to FIBER_FLAG_FLOAT_SWITCH being unsupported
This will be fixed in the next OCA version.

Hardware acceleration crashes browser a lot of times on OpenGL 2.1

Hardware acceleration on OpenGL 2.1 crashes the entire browser. To fix, use --disable-gpu switch. Chromium 122 is espically bad about this.

Chromium 111+ x64: massive pagefile requirements

Chromium 111 removes some Windows 7 fallback code, and under x64, each process requires 512 MB of additional pagefile to work. 32-bit Chromium and 64-bit Supermium is not affected.

You can mitigate this in many ways: switching to x86, using --single-process, reducing processes by one of many Chromium command line flags, increasing the pagefile by upwards of 16 GB.

Chromium 113+: flickering title bar

By Chromium 113, there is a flickering title bar on window moving or tab actions. There is no fix, the only fix is DWM.
Exception: Supermium.

Supermium 121: crash on startup

Supermium 121 crashes on startup, with debug.log showing that it failed to load chrome.dll. This is due to progwrp TLS code conflicting with already existing TLS implementations. This also happens under WinXP compatibility mode on Win7 and will happen to Windows 2000 extended kernel, ReactOS when they add this feature.

Workaround: Spoof your OS version using NNN4NT5 to Vista or higher. A potenital fix would be: inside progwrp.dll, if a NT6 API is found (or TLS is detected), then don't enable Thread-Local Storage. Or patch out progwrp:

Patching out progwrp

Download x64dbg and any hex editor (HxD) and open progwrp.dll. Copy progwrp.dll and rename it into something else. Load progwrp into there, and then go to:
_TLSInit_DllMain_ProcessAttach@4 (or _TLSInit_DllMain_ProcessAttach on x64) function. Right click the first instruction that pops up, and the bottom bar should say:

.text:72805B80 progwrp (2).dll:$5B80 #4F80 <_TLSInit_DllMain_ProcessAttach@4>

The positions will be different. Open HxD and open the copy. Go to what the hashtag says [#4F80], and type in C2 04 00 for x86 or C3 for x64. Repeat the same thing for ThreadAttach and save the file.

After that, Supermium should open without NNN4NT5.

Issue with Chromium-based Apps

VSCode 1.82.X+

Windows 10 versions of Electron apps have a low chance of working, although CEF apps work just fine.
You might be able to fix them with Supermium Electron.

[Skulltrail192]

Of course my friend. I want know reason of some bugs:

• Downloads not working
• Extensions cannot be installed
• File Dialog cannot be opened

[idontyboi]

Thanks, I only test in XP32. I'm posting this from Supermium 118 on Windows XP. I'm pretty sure that Chrome 103 - 109 will work with the same ncrypt.dll workaround.

Debug spam in dwmapi_main.c:196

Please mute this debug spam.

Extensions can not be installed

This happens:

DebugString: "Unexpected object name: L"\\Device\\HarddiskVolume1\\Documents and Settings\\Administrator\\Local Settings\\Temp\\a0229288-db85-4a03-8eac-4441e13dcdc5.tmp""
DebugString: "Unexpected object name: L"\\Device\\HarddiskVolume1\\Documents and Settings\\Administrator\\Local Settings\\Temp\\Supermium\\User Data\\Default\\Extensions\\Temp\\8acc5a46-6ab1-49d0-a7fc-8c21f83eb9d6.tmp""

There is a issue with GetFinalPathNameByHandleW. You should enhance this function to accept that, likely syncing this function with latest Wine.
The next thing that is required is accepting Windows 7 flags in FindFirstFileEx.

Downloads not working is likely the same reason.

File Dialog cannot be opened

Likely expects full Vista File Dialog implementation.

get.webgl.org crashes the tab or browser

It crashes on SwitchToFiber function. Not much else known about it.

[p1glynlol]

Did you use some dlls other than ncrypt.dll to get Supermium 118 to work? For me I can launch Chrome 103-109 and Supermium 117 (without crashing), but I keep getting Aw Snap error on Supermium 118.

Edit: Ok now it works fine by adding --single-process and --disable-webgl on the shortcut. Also I just noticed that the browser looks off (Did they readded GDI rendering?).

[idontyboi]

I used --single-process --disable-webgl shortcut to reduce memory and be able to debug the process [thus see what caused it to crash in the first place] and to make Chrome take less memory. I haven't tested without it: that is the only DLL I used.

More issues:
User data seems to be stored in the %TEMP%... for some reason. So if you clear temporary directories, browser data is gone.

Rendering a PDF crashes the whole browser. It's painful to investigate due to memory constraints and the debugger itself crashing, but it could be due to the extension issues.

[jonm58]

check it out
Chuyu-Team/YY-Thunks#29

[Skulltrail192]

Can you help-me about Firefox 55 issue too?

[idontyboi]

Oh, good timing. I was testing that and got Firefox 59.0.3 working.
Patch mozglue.dll

0000688E: 3D 90
0000688F: B4 90
00006890: 05 90
00006891: 00 90
00006892: 00 90
0000AFD7: 68 EB
0000AFD8: 20 0E

Also go on about:config and disable multiprocess otherwise every tab crashes
ntdll.LdrRegisterDllNotification is required to get Firefox 59+ working.
It's likely that you also need either a real implementation of LdrRegisterDllNotification or enhancements to condition variables.
If you want I can open another issue

[Skulltrail192]

Thanks for all! Do you have discord, skype or any way to contact you? Do you tested Opera too? I don't know why Opera traditional not work and Opera GX work. Edge only work if you install Visual Studio Debugger. All installers not working.

[idontyboi]

I do, what's your Discord username?
It's expected that all the other Chromium browsers call said ncrypt.dll API and crash/burn, but I have only tested Supermium. Some likely don't call the defect API.
Also consider running WINE tests and unzip the tests to see API conformance with Windows. Condition variables have very poor results, and this impacts Firefox 59 requiring a patch to skip over errors. kernel32_test.exe file also causes a BSOD.

kernel32_test.exe sync
sync.c:2797: Test failed: got 3735928559
sync.c:475: Tests skipped: RtlInterlockedPushListSListEx not available, skipping tests
sync.c:596: memory resource notifications not supported
sync.c:1159: Test failed: DeleteTimerQueueTimer
sync.c:1964: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on untriggered condvar, not 317
sync.c:1973: Test failed: SleepConditionVariableSRW should return ERROR_TIMEOUT on untriggered condvar, not 317
sync.c:1980: Test failed: SleepConditionVariableSRW should return ERROR_TIMEOUT on untriggered condvar, not 317
sync.c:1869: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on out of band wake, not 317
sync.c:1879: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on out of band wake, not 317
sync.c:1898: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on out of band wake, not 317
sync.c:1964: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on untriggered condvar, not 87
sync.c:1973: Test failed: SleepConditionVariableSRW should return ERROR_TIMEOUT on untriggered condvar, not 87
sync.c:1980: Test failed: SleepConditionVariableSRW should return ERROR_TIMEOUT on untriggered condvar, not 87
sync.c:1869: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on out of band wake, not 87
sync.c:1879: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on out of band wake, not 87
sync.c:1885: Test failed: SleepConditionVariableCS should return TRUE on good wake
sync.c:1891: Test failed: SleepConditionVariableCS should return TRUE on good wakeall
sync.c:1898: Test failed: SleepConditionVariableCS should return ERROR_TIMEOUT on out of band wake, not 87
sync.c:1904: Test failed: SleepConditionVariableCS should still return TRUE on crit unlock delay
sync.c:1918: Test failed: pSleepConditionVariableSRW should return TRUE on good wake
sync.c:1924: Test failed: pSleepConditionVariableSRW should return TRUE on good wake
sync.c:1930: Test failed: pSleepConditionVariableSRW should return TRUE on good wake
sync.c:1936: Test failed: pSleepConditionVariableSRW should return TRUE on good wake
sync.c:1757: Test failed: error occurred during SleepConditionVariableCS
sync.c:1761: produced 50, c1 5, c2 23, c3 22
sync.c:1765: producer sleep 21, consumer sleep 15
sync.c:2420: Test failed: thread commands were executed in the wrong order (occurred 11 times).
sync.c:2424: Test failed: AcquireSRWLockExclusive didn't block when called multiple times from the same thread (occurred
 2 times).
sync.c:2428: Test failed: AcquireSRWLockShared didn't block when the same thread holds an exclusive lock (occurred 2 tim
es).
sync.c:2432: Test failed: AcquireSRWLockExclusive didn't block when the same thread holds a shared lock (occurred 2 time
s).
sync.c:2436: Test failed: AcquireSRWLockExclusive didn't block when a second thread holds the exclusive lock (occurred 2
 times).
sync.c:2440: Test failed: thread waiting for exclusive access to the SHMLock was not preferred (occurred 3 times).
sync.c:2532: number of shared accesses per thread are c1 46, c2 52, c3 46
sync.c:2533: number of total exclusive accesses is 11
sync.c:2739: Test failed: Unexpected debug info pointer 00F83D70.
06e0:sync: 1338191 tests executed (0 marked as todo, 0 as flaky, 29 failures), 1 skipped.

[Skulltrail192]

My Discord is Samuel Marins#0436

[p1glynlol]

Did you use some dlls other than ncrypt.dll to get Supermium 118 to work? For me I can launch Chrome 103-109 and Supermium 117 (without crashing), but I keep getting Aw Snap error on Supermium 118.

Edit: Ok now it works fine by adding --single-process and --disable-webgl on the shortcut. Also I just noticed that the browser looks off (Did they readded GDI rendering?).

I also tried to run patched Google Chrome 117 (https://github.com/Blaukovitch/GOOGLE_CHROME_Windows_7_CRACK) and it works too! (I also don't need to add --single-process and --disable-webgl, I just put ncrypt.dll and start !NOSANDBOX_STABLE.bat).

Unlike Supermium, this one uses the Windows 10 titlebar instead the one from Windows Vista/7 (without Aero).

[Skulltrail192]

What is NOSANDBOX_STABLE.bat?

[p1glynlol]

It's the batch file containing the flags for the browser to work "chrome.exe --no-sandbox --use-webgpu-adapter=d3d11 https://cracklab.team/PAunlock/" you can also put it on a shortcut (copy anything except chrome.exe) and it will work fine.

[Skulltrail192]

However, i pass --no-sandbox by default

[win32ss]

Edit: Ok now it works fine by adding --single-process and --disable-webgl on the shortcut. Also I just noticed that the browser looks off (Did they readded GDI rendering?).

I did add GDI rendering back, but I didn't announce it because it doesn't work without the --single-process switch in all cases (and also because most of the browser UI text is offset high). When I figure out how to get it working without the switch, I'll make a chrome://flags option for it. I will also remove dependencies on GetLogicalProcessorInformationEx and other functions introduced in Windows 7 (specifically for extensions, I'm allowing Vista and below to use FindFirstFileExW without a flag introduced in 7 again), to meet the initial goal of the browser working on Vista without extended kernel.

And remember to check Supermium\User Data\Crashpad\reports for any crash dumps (which can be investigated using the Supermium PDBs which include line number references).

[idontyboi]

How can DirectWrite rendering be forced on OCA Supermium 118? Is there a switch for that?

[win32ss]

How can DirectWrite rendering be forced on OCA Supermium 118? Is there a switch for that?

I heard that there wasn't a good DirectWrite implementation for XP, so I didn't make a switch for forcing it. But it may work if you spoof OS version to Vista (the check will probably be replaced with a check for the presence of dwrite.dll itself).

[idontyboi]

Vista compatibility crashes but I bisected this to one last bit of XP code still remaining:
https://source.chromium.org/chromium/chromium/src/+/main:third_party/crashpad/crashpad/util/win/registration_protocol_win.cc;l=158;drc=8f5c47fd8d80208c191fe575f0817b26a9093837
Changing that parameter to return a null pointer whilst in a debugger works and it restores DirectWrite rendering as expected, but it has Windows 10 UI and flickering native XP windowbar

[win32ss]

Even with the force-xp-theme flag set in chrome://flags?

[idontyboi]

I tried force-xp-theme and it does revert it to the XP UI and fix the flickering issue. I don't know why the Windows 10 UI appears without it regardless of Vista/7 compatibility mode. DirectWrite rendering works with that flag, so that's good.
I'll simply disable compatibility mode for Supermium for the time being, since with it, I have to setup a debugger and edit a parameter every time.

[Stepman123]

I would like to try Thorium https://github.com/Alex313031/thorium-win7 as it supports SSE2 and also has a lower version than Supermium.
Where can I download ncrypt.dll(wine)?

[p1glynlol]

I would like to try Thorium https://github.com/Alex313031/thorium-win7 as it supports SSE2 and also has a lower version than Supermium. Where can I download ncrypt.dll(wine)?

https://cdn.discordapp.com/attachments/1030000423282675714/1153978374180585472/ncrypt_ocapi.zip

Put both ncrypt.dll and bcrypt.dll next to thorium.exe (DO NOT PUT IT INSIDE THE 109.0.5414.159 FOLDER).

Thorium 109 also works on XP/Server 2003 (64 bit) too, clearly something has changed between 109 and 110.

Edit: Forgot to say that the sandbox works on Thorium.

[win32ss]

Hello @win32ss , I have some questions about Chrome, since you developed Supermium and KernelEX for Windows Vista...

1 - How can we make compatibility mode with Windows 10 work and remove that warning that Chrome is no longer compatible with XP/Vista? Both in the titlebar and in the settings. At the time, Chrome 100 on Vista with kernelEX with osver.ini setting compatibility mode to NT 6.1 removed that warning that Chrome was no longer compatible with Vista/XP, But nowadays we would have to do it with NT 10.0 compatibility mode on XP with OCA

2 - Is it possible to make the Chrome 49-122 sandbox work?

When they added the banners for 7 and 8.x, the method for checking the version specifically for the banner was changed so that it checked the version of kernel32.dll instead of the actual versioning APIs. So you either need to change the resources for kernel32 so it reports a version number of 10.x, or kludge the function (GetFileVersionInfoW I think) used to get the file version resource string so it returns the "right result" for Chrome.

As for the sandboxes, they patch the syscalls in NTDLL on initialization. The problem is that the 32 bit syscall format changed in Windows 8 and Chromium 110 and up had the code that recognized the old syscall format removed. Furthermore, Server 2003 WOW64 suspend-on-create processes only have the 64 bit NTDLL loaded (Vista and later load both NTDLLs). This means that there is another variation in the patching process specific to Server 2003 (which was removed well before Chromium 110) to patch the section mapping functions in the 64 bit DLL until the 32 bit DLL is loaded, instead of patching a DLL that hasn't been loaded into the process' address space yet.

Aside from that, there are some problems with handle duplication (related to proc_thread_lists) and also object integrity levels, which were all resolved by Supermium and are relatively simple for One-Core-API to handle by implementing the necessary code (in the latter case, some TokenInfoClasses).

[idontyboi]

Is it possible to make the Chrome 49-122 sandbox work?

There are currently some conflicts between the workaround that makes Chrome 111-122 work, and the Chrome 49 sandbox.

The proper way on how to get the Chrome 49 sandbox work without having Chrome 111+ break requires kernelmode changes related to security, however my idea is a registry configuration to simply disable the workaround.

[DarioPlay]

chrome 123 working fine!

[DarioPlay]

Is it possible to make the Chrome 49-122 sandbox work?

There are currently some conflicts between the workaround that makes Chrome 111-122 work, and the Chrome 49 sandbox.

The proper way on how to get the Chrome 49 sandbox work without having Chrome 111+ break requires kernelmode changes related to security, however my idea is a registry configuration to simply disable the workaround.

oh yes, but out of curiosity I asked if it would ever be possible to make the sandbox of all Chromes work (including the most recent one)

[p1glynlol]

Update about Supermium: 121 requires os spoofer if one-core-api is installed, else it won't launch. Compatibility mode won't work. And yes I do know that this version does not require it.

Nothing seems to break based on my tests.

[TK50P]

One-Core-API is no longer needed for Supermium.
It works without One-Core-API.

[p1glynlol]

One-Core-API is no longer needed for Supermium. It works without One-Core-API.

Yea that's why I said "And yes I do know that this version does not require it."

I'm testing those versions with OCAPI installed just to see if anything breaks.

[Skulltrail192]

One-Core-API is no longer needed for Supermium. It works without One-Core-API.

Fine. Good luck. Try run any other new application without OCA.

[TK50P]

Even, sandbox is NOT disabled and no longer crashes (Aw, Snap!). but installer doesn't work. so you need to install manually.
That means, --no-sandbox and --single-process argument also no longer need.

However, if you run in vm, the number of cores must be set to at least 2. otherwise, system will freeze.
VMware:

VirtualBox:

[jonm58]

NSA QUANTUM Foxacid:browser sandbox?Tor?lmao

[sambow23]

Thorium 119 on OCAPI 3.0.5 seems to have some rendering issues unless you add --disable-gpu.

It looks correct with the workaround mentioned

I'm running this on real hardware (Core 2 Duo P8700 // NVIDIA Quadro NVS 160M)
Is there any way to get HW acceleration working without this rendering bug?

[jonm58]

vmware 17.5.0
windows xp 32bit(no 3d acceleration)
one core api 4.0.0 rc1
Thorium-Win M119.0.6045.214-1 SSE2

[jonm58]

https://learn.microsoft.com/en-us/windows/win32/api/winhttp/nf-winhttp-winhttpcreateproxyresolver

winhttp.dll
10.0.18362.778

lost DLL:
API-MS-WIN-NETWORKING-INTERFACECONTEXTS-L1-1-0.DLL
API-MS-WIN-SERVICE-CORE-L1-1-4.DLL
EXT-MS-WIN-FIREWALLAPI-WEBPROXY-L1-1-0.DLL
IESHIMS.DLL
WEBSOCKET.DLL

--------------------------------------
lost Function:

API-MS-WIN-CRT-PRIVATE-L1-1-0.DLL:
_o___std_type_info_destroy_list
_o__cexit
_o__configure_narrow_argv
_o__crt_atexit
_o__initialize_narrow_environment
_o__purecall
_o__seh_filter_dll

NTDLL.DLL:
RtlCanonicalizeDomainName
RtlPublishWnfStateData

API-MS-WIN-CORE-WOW64-L1-1-1.DLL:
Wow64SetThreadDefaultGuestMachine

API-MS-WIN-CORE-STRING-OBSOLETE-L1-1-0.DLL:
lstrcmpA

KERNELBASE.DLL:
AppContainerRegisterSid
AppContainerUnregisterSid
GetIsEdpEnabled
GetIsWdagEnabled
SubscribeEdpEnabledStateChange
SubscribeWdagEnabledStateChange
UnsubscribeEdpEnabledStateChange
UnsubscribeWdagEnabledStateChange

RPCRT4.DLL:
RpcServerInqBindingsEx
RpcServerRegisterIf3

DNSAPI.DLL:
DnsConnectionDeletePolicyEntries
DnsConnectionDeleteProxyInfo
DnsConnectionFreeNameList
DnsConnectionFreeProxyInfo
DnsConnectionFreeProxyInfoEx
DnsConnectionFreeProxyList
DnsConnectionGetNameList
DnsConnectionGetProxyInfo
DnsConnectionGetProxyInfoForHostUrl
DnsConnectionGetProxyList
DnsConnectionSetPolicyEntries
DnsConnectionSetProxyInfo
DnsConnectionUpdateIfIndexTable
DnsFreeProxyName
DnsGetProxyInformation

IPHLPAPI.DLL:
GetIfEntry2Ex

[ArtsemK]

I installed all the one core api components (6 pieces) in the order specified in the accompanying text file.

rebooted.

blue screen of death with some inscriptions. example:

[jonm58]

我按照随附文本文件中指定的顺序安装了所有一个核心 api 组件（6 个）。

重新启动。

死亡蓝屏，带有一些铭文。 例子：

what ver one core api?
In this case, we need your dump log file (.dmp file format) for further analysis.

[ArtsemK]

I reinstalled the system (now windows 7), so there will be no memory dump file.

I used this windows xp pro sp2 64 bit (wannacry protection)

https://windows-soft.info/743-windows-xp-professional-edition-sp2-vl-2017-by-lopatkin-x64-2017-eng-rus.html

[jonm58]

Supermium 122.0.6261.85 not working
4.0.0-20240210-experimental
compatibility mode need set windows 95 or windows 98/windows me

[ArtsemK]

why didn't you tell me that I don't need one core api anymore? I heard that supermium now works on windows xp, all that remains is to make a couple of corrections.

[jonm58]

why didn't you tell me that I don't need one core api anymore? I heard that supermium now works on windows xp, all that remains is to make a couple of corrections.

stop your off topic...

[p1glynlol]

Supermium 122.0.6261.85 not working 4.0.0-20240210-experimental compatibility mode need set windows 95 or windows 98/windows me

Does Supermium 122 launches with NNN4NT5? Compatibility mode doesn't seem to be effective for this.

I wonder if it's possible to implement a option to disable Local TLS for some application on OCAPI.

[Stepman123]

Does Supermium 122 launches with NNN4NT5?

The 32-bit version works the same as before 121.

[Stepman123]

Supermium 122.0.6261.85 Hotfix - Now supports single-core HAL on XP.
On OCA it also runs via NNNT4NT5

[jonm58]

One Core API 4.0.0-20240210 windows 2k3 standard
supermium not use NNN4NT5 log
Desktop.zip
By Dependency Walker

[p1glynlol]

Supermium 126 on XP x64 (with One-Core-API 4.0.0 RC4 installed): Does not work! (probably not compatible with OCAPI progwrp implementation)

To disable OCAPI progwrp you'll need to rename progwrp.dll (on System32/SysWOW64) to progwrp.dll.bak and restart the PC, then use NNN4NT5 to spoof the OS to Vista.

Hopefully this will get fixed before 4.0.0 RTM launches.

[p1glynlol]

Update: I tried the newest version "Supermium 126.0.6478.249 R3 (XP x64 Hotfix)" and guess what... It seems to work without NNN4NT5. Make sure to remove (or rename) OCAPI progwrp.dll in System32 and SysWOW64 and restart the computer (else the error will show up again).

Supermium generates a log called "condvar_use_log.txt" which might have to do with DirectWrite runtime from Supermium, it does work with OCAPI and the emojis also works.

Can anyone test it? If it works then there's no need to use OCAPI progwrp.

[Stepman123]

Can anyone test it?

Yes it works with Server 2003 x86.

If it works then there's no need to use OCAPI progwrp.

You are right if someone needs Supermium. I still can't configure Supermium to work as fast as Thorium 109.

[Stepman123]

I also tested it on Windows XP without OCA. It seems that win32ss optimized the code for processors like single-core Pentium-4. This is noticeable on modern websites like GitHub, Discord, when images and other page elements are loaded. The CPU load is lower and it happens faster than in Server 2003 OCA.