KernelEx 2022 (Kex22) Test Versions (4.22.26.2)

[Goodmaneuver]

I have manually installed Java 1.8.0_51 in  WinDir subdirectory JAVA\JDK\jre after installing JDK 1.7.0.0. Since some executables and control panel now need 2K settings I needed to update KEX AppSettings. My original settings were *\Java*.exe = DCFG1. This original setting results in other executables having names not starting with Java in the JRE folder having BASE setting. Tested in 2 different builds and happens with JDK/JRE 1.7.0.0 as well. Is this intensional? I still use DCFG1 for *\JAVA*.EXE and now use *\*.CPL.EXE  = NOHEAP and over riding the BASE auto setting that I found for these two exes Jp2Launcher.exe and  SsvAgent.exe to NOHEAP.

Edited May 11, 2020 by Goodmaneuver

[jumper]

BASE and DCFG1 are the same, so original definition is no longer needed. Try *\jre\* = NOHEAP or WIN2Kstd instead.

 

Edited May 11, 2020 by jumper
Added WIN2Kstd

[Goodmaneuver]

DCFG1 is not the same as BASE as if JAVA*.EXE is set to BASE then the JAVA control panel does not work. It complains that Javacpl.exe has a device not functioning. This is because of the VerSetConditionMask and VerifyVersionInfoW functions. I run Explorer with DCFG1 to allow more programs to run without the need for extra individual settings. *\JRE\* is a idea that I had not known about and assigns all modules to the desired setting, but JVM.dll only works with BASE or DCFG1 and there are the Java, Javaw and Javaws executables in the WinDir as well. The phenomenon of why the other files had the BASE setting in the Java\Bin location is still a mystery when they should have just been in following mode.

The other problem which may not be a KernelEx one but is important is the assigning of folders like .folder in Java 1.7 and above. I think Java has a problem finding the correct folders. Some programs I have that use Java are VUZE 5.4; Art of Illusion vs 3.1; and Eclipse Luna 2014. It is important as saving of work in Eclipse can not be done properly and Vuze always wants to install the Swt download in an incorrect location or something like it. Edit: The assigning of folders like .folder in Java 1.7 and above happens with other NT operating systems; it puts a full stop in front of the folder and places it in a parent folder.

I am still looking into this issue but would appreciate help. Eclipse is storing in correct location but 0 kB, see picture.

I have decided to have the java 1.7.0.0 runtime as is and let the Java control panel update the 1.8.0_51 runtime location. This will negate any errors associated with manual registry updates for Java 1.8. In so doing this, the 1.8.0_51 runtime location bin has all the executables KEX set to DCFG1 with KEX AppSettings *\Java*.exe = DCFG1. The DLLs in the 1.8 folder are in follower mode.

Edited January 29 by Goodmaneuver
Important read Purple text, the full stop in front of folders happens in modern OS.

[Goodmaneuver]

Java uses local variables which after the function ends, the variable name does not exist. C/C++, Visual Basic, and VB.NET use static variables ( local ) that stay in RAM which allows a value to be retained for different function calls. I am wondering if there is a race situation that does not favor the saving in Eclipse as JVM not working unless assigned DCFG1 or BASE - could be a timing speed problem. MSHTML vs 6.2800 needs KEX disabled due to speed before the graphics output outputs. Experiment using explorer in a window and then go full screen and back with different MSHTML KEX settings. Will have to restart Windows for accurate results between KEX setting changes on the MSHTML as it is loaded at all times. MSHTML also has many registry settings and I found that it will not register any more, time wise, I would imagine since the original install of IE6 or IE5.5 : - ( WinME original install ).

Edited May 16, 2020 by Goodmaneuver

[jumper]

OT. Please reread Warning in TP.

 

[DosFreak]

If there was an issue with the files in the package then likely everyone is going to have the same issues, removed since you know the software the best.

Edited May 18, 2020 by DosFreak

[jumper]

Thank you for your enthusiasm, but you need to coordinate with me before redistributing collections of my files. This one is flawed. Please remove it immediately. 

Remember that these files are only for developers and still under development. Anyone who cannot read/scan through the whole topic and compile their own full update set should not be using them.

That said, I am working on a full release package suitable for end users. Please be patient.

 

Edited May 18, 2020 by jumper
Tone

[jumper]

KernelEx v4.5.2016.25 updates.7z

Readme's inside.

 

[schwups]

Thank you so much for all of your work. First start failed with error messages. "Rundll32 has caused an error in Kernel32.dll" ... "KernelEx.dll cannot start". All page faults are in Kernel32 at 0177:bff6bb07, e.g. 

RUNDLL32 caused an invalid page fault in
module KERNEL32.DLL at 0177:bff6bb07.
Registers:
EAX=00000000 CS=0177 EIP=bff6bb07 EFLGS=00000246
EBX=00000480 SS=017f ESP=0063f6a4 EBP=0063f6b4
ECX=ffffffff DS=017f ESI=7d01c510 FS=229f
EDX=819c9de8 ES=017f EDI=00000094 GS=0000
Bytes at CS:EIP:
ff 76 04 e8 8a 87 ff ff 5e c2 04 00 56 8b 74 24

Stack dump:
00000480 7d00b736 7d01c510 ffffffff 0063f6cc 7d00b67c 00000482 00000177 bff6a444 000819cb 0063f6ec 7d00aa2e 00000482 00000000 bff6a501 819ba000

I will figure out what causes the problem. (My MSVCRT.dll usually is version 7.00.9981.0.)

Does the KEXVista.dll file have to be registered? Should I replace my apphelp.dll 5.1 with version 5.2?

Edited November 16, 2021 by schwups

[schwups]

First I removed all not essential files of the KernelEx folder, also Kstub824. After I replaced Kexbases by version 24 I'm able to run ME. Then I reinserted the other files taken out before. KernelEx basically works, but isn't very functional with this configuration. I was able to start OpenOffice3.2 and Opera 12.02. All Firefox versions I tried didn't start, even 9.0.1. Kexvista is listed in Ktree without registration.

[tyukok]

Tried it out on 98SE, was able to boot without problems (my MSVCRT.dll is 6.10.9848.0). Unfortunately any Firefox versions (and other Gecko/Goanna-based browsers) that were working before didn't start. I tried adding Kexstubs in CORE.ini, but after that they all started crashing on start, with problem being, apparently, kexbasen.dll, address 0167:7d011b65. Replacing it with the one from version 24 didn't do anything, so I reverted back for now.

Edited November 16, 2021 by tyukok

[Dave-H]

@jumper

Thanks very much for the latest update, but could I just clarify something?

I have version 4.5.2016.20 of kexbases.dll installed, 4.5.2016.21 of kexbasen.dll, 4.5.2016.19 of kernelex.dll, 4.5.2016.18 of kexcom.dll, and Kstub626.dll.
This looks a mix and match mess, but I had problems with later versions, which appeared when I replaced the files.
What files exactly should I replace from this latest package to try and prevent problems again?
I notice there are now files in there which appear to be for Vista.
Should every file be copied to the Windows\KernelEx folder, overwriting the originals?
Thanks,
Dave.

[Goodmaneuver]

Kstub824, kexbases and kexbasen all do not work from the K25 upload. I tested them individually and the best result was that KernelEx.dll could not be loaded message occurred. EDIT Kstub824 and kexbasen are good after fixing core.ini. See https://msfn.org/board/topic/157173-kext-diy-kernelex-extensions/?do=findComment&comment=1208164

Edited November 22, 2021 by Goodmaneuver

[schwups]

I tried to update KernelEx on two more Win ME's to rule out that it is a special problem with my main machine. But that isn't the case.

Kstub824: I merged both files the new ini with my ini (of my Kstub823) to test the new definitions and possibly to benefit from these. I found there is a problem with BaseUpdateAppcompatCache=v3. I had to disable it to avoid a page fault "EXPLORER caused an invalid page fault in module MSVCRT.DLL at 0177:7c00145a." on OS start.

So far I have successfully updated KexVista and Kstub.

I'm not sure KexVista may not work as it should, for example Firefox 54.0a1 (XPSP2 mode) runs without Kexvista but not with (Vista mode).

Firefox event log: "AbortMessage=###!!! ABORT: unable to find a usable font (serif): file c:/builds/moz2_slave/m-cen-w32-ntly-000000000000000/build/src/gfx/thebes/gfxTextRun.cpp, line 1844.....GraphicsCriticalError=|[0][GFX1]: no fonts - init: 1 fonts: 84 loader: 1 backend: gdi system-uptime:   148.702 sec (t=13.6201)"

minidump-analyser: 2021-11-18 21:02:46: minidump.cc:4808: ERROR: ReadBytes: read 0/32
2021-11-18 21:02:47: minidump.cc:4453: ERROR: Minidump cannot read header

Edited November 18, 2021 by schwups

[jumper]

Core.ini ApiConfigurations need to be renumbered without gaps in the sequence. This should reenable the legacy modes.

I will refresh the package with that fix, a version-matching Kernelex.25, and also the Core.24 that got left out of the v24 package.