mjd79

https://github.com/win32ss/supermium/issues/1367 I don't have a github account, so if someone could pass on. Win32ss needs a bcryptprimitives function, specifically ProcessPrng. Supermium doesn't have VMP anyway, so the author can change the import to his own implementation of bcryptprimitives in the patcher anyway. GetProcessMitigationPolicy and SetThreadInformation are also needed, but from what I've seen the author of Supermium certainly knows. Oh, right, I forgot. On Supermium 132, including R3 widevine does not work at all, even on Windows 10, and even more so with the repaired widevine under 7, or with original widevine with VXkex. The world does not end with Netflix, there are many sites that use DRM, for which unnecessary for the author to buy VMP license for Supermium. Works in 126 R7:

Unfortunately, versions 2830.0 and 2380.100 have recently stopped working. For now, my temporary workaround for 4.10.2891.0 on Win 7 is to use pwrp_k32 from Supermium instead of kernel32, an p_cryptp instead of bcryptprimitives. There are actually a few functions missing, so perhaps I will write my wrapper and make it available to the public. That's basically what Mozilla does. https://hg-edge.mozilla.org/releases/mozilla-esr115/log?rev=widevine You can find bcryptprimitives in the widevinecdm.dll file by searching 62 00 63 00 72 00 79 00 70 00 74 00 70 00 72 00 69 00 6D 00 69 00 74 00 69 00 76 00 65 00 73 in hex editor (two occurrences), replace, for example, 62 at the beginning with 78, and name the p_cryptp.dll file xcryptprimitives.dll.

Indeed, the sandbox is fixed. #supermium-tab-options works fairly decently, although v109 mode has minimal bugs (at least on 7 SP1 with Aero).

No, any pure Chromium lacks API keys and OAuth2, I remind you that I only applied e3kskoy7wqk patches to the completely original Chromium. I have not tried this flag, I will check it under Win7, on which, of course, my rebuilts work.

Unfortunately, Chrome Sync still does not work after adding Google API keys, you still need OAuth2 keys. As a not being a Chromium developer, you can only use the default keys, which can be added with the --oauth2-client-id and --oauth2-client-secret flags. There is a simple solution for this, so that you won't have to add flag it to shortcuts or in the registry (in case you set it as default browser) void ChromeMainDelegate::PreSandboxStartup() { base::CommandLine& command_line = *base::CommandLine::ForCurrentProcess(); std::string process_type = command_line.GetSwitchValueASCII(switches::kProcessType); if (!command_line->HasSwitch("oauth2-client-id")) { command_line->AppendSwitchASCII("oauth2-client-id", "77185425430.apps.googleusercontent.com"); } if (!command_line->HasSwitch("oauth2-client-secret")) { command_line->AppendSwitchASCII("oauth2-client-secret", "OTJgUOQcT7lO7GsGZq2G4IlT"); } With this code, replace the PreSandboxStartup() function in the chrome_main_delegate.cc file. Edit: in testing/variations/fieldtrial_testing_config.json there are flags about TabStripComboButton and disabling MV2 support. They override the default values declared in the corresponding files, hence you need to remove them from this json.

The patcher included by e3kskoy7wqk, along with the diff files for each compilation (a total of 5 apply_patches.bat files), require a bit more correction than the author claims. At “REPO_PATH=E:\win7dep\chromium\src” change the path to yours, of course, but besides that, you also need to replace “!REPO_PATH!” with “%REPO_PATH%” in two places. In addition, it is required to add clang_use_chrome_plugins = false to args.gn.

For now i5 6500, I just had a Ryzen 1700 lying around for a long time, I bought a MSI B450 motherboard for it yesterday. On the i5 compilation takes 15-20 hours, depending on the number of objects. On your processor it will be at least about 36 hours. 16GB of RAM is also the absolute minimum. And no, unfortunately it can't be done on 8.1. Edit: 8 hours on Ryzen 1700, the rest of the hardware configuration except the motherboard the same. On the i5 6500 as I wrote earlier 15 hours. However, 32GB RAM is not necessary for me for now

The issue described in this comment has been resolved, see https://msfn.org/board/topic/186741-firefox-and-chromium-running-on-windows-7-by-e3kskoy7wqk/page/9/#findComment-1279595

This is my personal lowest priority, but from what I've gathered, it won't be difficult.

I don't know, maybe I'm just stupid, but the compiler kept telling me what to do with them. There were two errors, regarding elevation service and updater legacy. In both cases, the compiler components created corrected files in temporary files, which had to be copied to src\third_party\win_build_output.... I was doing this correctly all the time with elevation service, but for some reason not with updater legacy, which I copied to out\Default\gen\... ninja has the ability to compile individual objects, it seems that I will have to compile those problematic elements first and paste the fix generated by it even before I start compiling the whole. I'm currently finishing compiling already I can't even remember which version (it's been compiling since 3 pm on Sunday, and it's 8 am on Monday) I have 2000 objects left, which will take maybe half an hour. I purchased slightly better hardware (Ryzen 7 1700, 32GB of DDR4 RAM) to make the compilation time significantly shorter, should be at my place tomorrow. Edit: YES!!!!!! I GOT IT!

This man compiles almost pure Chromium. He only adds basic codecs and ffmpeg. You can always compile yourself, but as you can see, despite the seemingly trivial instructions, it's not so easy.

This version of the guide no longer applies to compiling with e3kskoy7wqk patches. I will at least write a compilation tutorial in Windows and for Windows What you need: - Windows 10/11 (before 10 20H1 additional steps with SDK required) - Visual Studio 2022 (can be free Community edition), git and Python (I recommend 3.14) Run the Visual Studio installer from CMD with the following arguments. PATH_TO_INSTALLER\VisualStudioSetup.exe --add Microsoft.VisualStudio.Workload.NativeDesktop --add Microsoft.VisualStudio.Component.VC.ATLMFC --includeRecommended. Then install the Windows 11 SDK in the latest available version. Install git and python. In Git Bash, configure: git config --global user.name "My Name" (optional) git config --global user.email "my-name@chromium.org" (optional) git config --global core.autocrlf false git config --global core.filemode false git config --global core.preloadindex true git config --global core.fscache true git config --global branch.autosetuprebase always git config --global core.longpaths true Now we will take care of downloading depot_tools. In my example, you first need to create a src folder in C:/. Then in the CMD: cd C:\src git clone https://chromium.googlesource.com/chromium/tools/depot_tools.git Next add the environment variables: DEPOT_TOOLS_WIN_TOOLCHAIN=0 vs2022_install=C:\Program Files\Microsoft Visual Studio\Community (or other edition) NINJA_SUMMARIZE_BUILD=1 PYTHONIOENCODING=utf-8 PYTHONUTF8=1 Edit the "Path" variable with path to depot_tools (eg. C:\src\depot_tools) and move it up before any Python installation. Next simply run the gclient command in CMD, it will download the basic components needed. In the folder where depot_tools is located, create a directory, for example, named chromium. Go to it in the console and run the command "gclient sync" if you want to download the latest version, or gclient sync --revision src@xxx.x.xxxx.xx (the version number you need). Use --no-history parameter at the end of command, downloading entire history, including branches and tags is about 50GB, I think it's not worth it. Next, find /testing/scripts/common.py, and change: try: existing_acls = subprocess.check_output(['icacls', acl_dir], stderr=subprocess.STDOUT, universal_newlines=True) to: try: existing_acls = subprocess.check_output( ['icacls', acl_dir], stderr=subprocess.STDOUT, encoding='mbcs', errors='replace' ) Also in build/toolchain/win/midl.py find: returncode, midl_output_dir = run_midl(args, env_dict) if returncode != 0: return returncode and change to: returncode, midl_output_dir = run_midl(args, env_dict) if returncode != 0: return returncode # WORKAROUND: Copy MIDL outputs from temp dir to outdir for f in os.listdir(midl_output_dir): shutil.copy(os.path.join(midl_output_dir, f), outdir) Now we are ready to compile. Go to the src folder inside the folder we named chromium earlier in the CMD, and run gn args out/default (you can name it whatever you want). This will open your default notepad, where you can set all the arguments. If compiling under Windows, I recommend using at least the following: is_official_build = true is_component_build = false is_debug = false target_os = "win" target_cpu = "" (x64, x86) proprietary_codecs = true ffmpeg_branding = "Chrome" chrome_pgo_phase = 0 You can also add arguments about Google API keys or widevine google_api_key = " " google_default_client_id = " " google_default_client_secret = " " enable_widevine = true ignore_missing_widevine_signing_cert = true enable_media_drm_storage = true Save the args.gn file and close notepad, gn will generate the necessary files. Now just use the ninja -C out/Default (you can also add the parameter "-j X", where X is the number of clang compiler processes)

After 10 hours of compilation, I received errors. Then I would face another 10 hours, and in the meantime more errors would appear. Not for my patience, I pity the money for more powerful hardware, as long as I can't compile without errors, and I'm not able to do that at the moment at this compilation rate. I apologize to the users I disappointed, apparently I'm not smart enough after all and despite following the instructions I keep doing something wrong all the time

It's now a question for users whether ungoogled or Vista support is more of a priority. My computer can take more than 24 hours for one compilation. Unfortunately, in the previous system installation and previous software setup, I made such a mess of things with the trial-and-error method (and doing some things at my own discretion, not how the tutorials dictate) that something went wrong during compilation. This time I found much better tutorials and have experience from the previous tries. Compiling Chromium in theory is very simple, but very many things can go wrong. Now I downloading the entire repository with every possible version from the main branch. You can also fetch with the command -gclient sync --revision src@137.0.7116.0.

I have just managed to start compiling for now a clean Chromium 137.0.7116.0 without any patches, just to be sure that no problems will occur. If successful, my intention is to apply all the patches from https://github.com/e3kskoy7wqk/Chromium-for-windows-7/tree/main/137.0.7116.0 and restore Vista and 7 SP0 compatibility, and 32 and 64 bit versions using https://github.com/Chuyu-Team/YY-Thunks/ - what the author of Chromium for Windows 7 does not intend to continue for some reason, despite the fact that it is the simplest and very well working solution.

It seems that yes, chrome.dll is already 270mb....

Ok, so now it's time for specifics and evidence. I'll start with v135 ungoogled and v137 on Windows Vista SP2 x64 (update level 2017.04 - final). Let's start with what extended kernel files can be redirected locally (v2023-03-09 rev2). Most except ntdll.dll, bcrypt.dll, and kernel32.dll. Included in this version is kernel33.dll, and it is this that should be used by renaming it to kernel32.dll. Then paste that below into notepad, save it as a file with a .reg extension and add it to the registry. This turns on the ability to redirect DLLs locally. Windows Registry Editor Version 5.00 [HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options] "DevOverrideEnable"=dword:00000001 Just create a folder chrome.exe.local in the folder with it and copy these files. Now, using CFF Explorer, change the NtOpenKeyEx import in chrome_elf.dll to NtOpenKey (thanks to D.Draker for the hint) On some graphics cards with very old drivers supporting, for example, only DX9, the GPU sandbox may crash. In this case, create a shortcut to chrome.exe and add --disable-gpu-sandbox. In chrome://flags set "Choose ANGLE graphics" backend to D3D9, from now on the browser should work correctly without this flag. Now I'll move on to 7 SP0, in which it turns out. you need advapi32.dll from Vista's extended kernel or from Supermium 132 (it's called p_advp32.dll there) In the first case (x64 only), use CFF Explorer to replace the import from RPCRT4.dll I_RpcBindingIsServerLocal with I_RpcBindingIsClientLocal. Then open chrome.dll in HxD, for example, and search for advapi32.dll. You will find it in two places, change it to e.g. advapi33.dll and save. In the same way, of course, name the file from Vista ext kernel/Supermium. That's it, the browser should work fine. I also recommend adding this to the registry so that non-ungoogled versions don't disable MV2 extensions on their own, and so that they can be added from the Chrome Web Store. Windows Registry Editor Version 5.00 [HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Policies\Chromium] "ExtensionManifestV2Availability"=dword:00000002 Screen confirming disabling client hint working in one of the previous posts, this one would be too long.

NtOpenKey without the "Ex" appended was enough. shell32 you mentioned and most other libraries I redirected locally from the extended kernel of March 9, 2023 (rev2). It is not possible according to what I have determined for now to redirect bcrypt.dll and ntdll.dll, kernel32.dll, which is a wrapper for Chromium 110 and 111 support does not work, probably due to lack of platform updates, however, there is also kernel33.dll in the files, which is the original kernel last modified by win32ss in December, 2022, and it works with ungoogled 135 and Chromium 137 without problems. My Vista update level is end of April 2017, no post-eol updates.

I am using this library for now for personal use and temporarily. I will try to use advapi32 from SP1 Absolutely zero success, even though advapi with SP1 does not lack dependencies in SP0. BTW I forgot to mention, but in v137 the flag to disable clients hints actually worked, including on Vista. Besides, every time I use these libraries, I mention that they come from Supermium. And the fact that I use them to run other, better browsers at the moment, is just my business. As long as I don't create ready-made mods and make its libraries available with them (no, I won't do that, I will prepare simple tutorials at most)

The good news is that I've already run ungoogled 135 and the latest 137 on Vista SP2 x64 (without the 2019 platform update BTW), with some March 9, 2023 extended kernel files redirected locally. On 7 SP0 the problem at least for ungoogled 135 appeared to be with advapi32, I will also check if it can be transferred from SP1 without too much trouble, or if you will have to use the wrapper included in Supermium. I'll show the screenshots later For now I'm also talking only about x64 versions, in Vista Extended Kernel there are unfortunately only wow64 files and not for x86, in addition, mostly flawed. On 7 SP0 it should be possible to run any combination.

@D.Draker, I already know what break ungoogled Chromium 135 and Chromium 137 by Chinese stundent on 7 SP0. It's ADVAPI32.dll, I temporarily used the whole library from Supermium, but in a day I'll check which specific function is responsible for it. On Vista x64 SP2 it should be possible to run these browsers with an extended kernel redirected locally, as long as the “ntext.dll” file created a few years ago by win32ss is enough to replace ntdll.dll.

Unfortunately, as you probably already saw in the relevant topic, it will not support Vista. I checked the version for 7 SP1 on 7 SP0 out of curiosity, the browser takes a suspiciously long time to start and I get a STATUS_BREAKPOINT error on every tab. I, unfortunately, do not have the skills or a sufficiently powerful computer to be able to compile his browser with the design that helped him maintain Vista compatibility, but perhaps there is a suitable person on the forum. On 7 SP0 it is probably possible to solve this problem, I checked much older compilations of the Chinese student, and its run without this problem.

I am not against this project, I would like to see it develop, and get to as good a level as the “Chinese student” fork, but since the man just mentioned solved the problems with memory usage and more, the more the author of Supermium can do it. In my opinion, Supermium should be created completely from scratch, and in two versions. The first for Vista (without the extended kernel) all the way up to 11, and the second for XP, containing all the necessary features in system library wrappers, OCA-style, which runs successfully the same version of Chromium that Supermium supposedly is, only without the sandbox for now.

After thinking about what you wrote, I have a conclusion - how is it that super-duper prominent authors of forks like Supermium or Thorium are not able to determine what causes such high memory usage/errors associated with it on some sites, and it was reached by an MSFN user, who, in addition, does not work with the source code, but already compiled browser?

It is up to D.Draker to whom he makes this method available. I received it in a private message for a reason, probably one that he doesn't want it available to the public.