Drugwash

It's possible that the resources occupied by the other video card could force the board to put the ATI in a nonstandard position that 9x cannot fully cope with. Or maybe you were just missing the monitor driver that I linked to in a PM. Install that driver, reboot and check again - you might get lucky. Otherwise, if there's other free PCIe x4 slot, try to move the ATI card there, maybe it will get other resources assigned.

There used to be an old package called 'Leaks' if I'm not mistaken, which can display GDI usage before and after running a certain application. It contains 'GDI Usage' and 'Menu Usage' - two separate executables. The first one says 'Should help you find GDI leaks / by Christophe' in the 'About' box (right-click the application titlebar -> About GDIUsage). It does not provide real-time statistics, however it may help finding out which application uses many GDI object, what type of objects and if there are any leaks in it, by clicking 'Compare!' after the applicaiton is closed. Hopefully someone here has a working download link at hand, otherwise I could try to find the archive on my drives/CDs and put it on my cloud.

Let's get one thing straight: by the new regulations - whoever may be enforcing that - the negative 5v (-5V) pin is no longer required in power source units. If you look at the main 20 (or 24) pin Molex on the motherboard, there may be a missing pin - that means the -5V output does not exist, in which case the software may yield weird results for that particular output. One possibility is that both the old and the new PSU are somehow defective as in not-so-accurate voltage output and they may both produce system lock-ups or other unusual behavior. Get a fairly precise multimeter and measure each voltage on the main Molex while the board is running (careful not to produce a short-circuit!) Does the video card have an additional power socket on it and if so, is it connected to the power unit? It may drain too much power, more than the PSU can offer. If that's the case, test with an older, more "gentle" card. Careful with drivers! Also check CPU temperature over time, the thermal grease between cooler and CPU may be missing or dry, the cooler may be improperly mounted, the fan may not provide enough air flow.

Yeah, I read about that earlier. But how about this one...? Not necessarily related to Win9x, but the mere idea of infecting the hardware...

Well, the Win95 tips and tricks have faded away in time. It's possible that the common controls pertaining to that OS family simply did not have the ability to use certain flags. What exactly is the version of comctl32.dll on your system? You may need to update that to 4.71 or whatever possible for a 95 system. The Plus! package probaly does that and some extra stuff that allows the stretch option for the wallpaper. As for the registry backup... I really don't know. I do keep the backup option active in my 98SE, as one never knows when it may be needed.

No idea if it works in 95 but you can try to change a registry value for the wallpaper: HKEY_CURRENT_USER\Control Panel\Desktop, WallpaperStyle --> 2. If the key doesn't exist, create it as a STRING value and set its value to 2. For the backup try Start > Run --> msconfig [ENTER] and see if there's something related to registry/backup in the Startup tab. Well, if msconfig exists and works in 95.

One thing I noticed: the return value for GetUserDefaultUILanguage() should be a hex value, so please try [Kernel32.dll]...GetUserDefaultUILanguage=r0x0407e0Please note that the above will only work in compatibility 'Windows ME' or higher. However, I have not used any compatibility mode for the installer other than KernelEx's default and the package still installed correctly. Other than that, it it is still possible that some module of the AV (or even ZoneAlarm which had its bad days - search these forums, you'll find details) could interfere with the installation process, even when disabled.

Minor graphical issues on a 98SE with Revolutions Pack and non-standard settings: screenshot Testing with Firefox 10.0.12 and KernelEx 4.5.2 official I get no patches needed in the results window, however the FileInfo plug-in (Total Commander) shows GetProcessIoCounters missing from kernel32.dll. Same missing API for Firefox 9.0.1 on the same system, however ImportPatcher40 does show it this time in the results window. Both firefox executables are in 'Default' compatibility mode. But... v10 is UPXed, while v9 is not, therefore I'd say unUPX would be necessary. If I may suggest a few improvements: - temporarily disable 'Analyze only' and 'Patch' while processing a file - show an 'Operation finished' notice somewhere for user's awareness - option where to save the log file - alphabetically sort APIs in log file, for each import - a readme text file în the archive, for those that may have gotten it from third-parties and have no Internet access to read details in this forum (or for those who forgot what's it for) Thank you for the new version!

For now I disabled multithreading for sheet and vxd, just as it was in 4.5.2. Honestly I wasn't sure what that dosstub was for. I just ignored the warnings. I'm also unsure whether size or speed should be the priority in KernelEx's modules. I've uploaded the full sources for R3 and R5 to my repository. Hopefully we get to common ground sometime soon. I'm using VC6-SP6. On another note, seeing the attempts at running XnConvert above, I've tried it myself. On the test machine, with both R3 and R5, it fails with various errors, including the ones mentioned above by loblo and others. However, on my main machine with KernelEx 4.5.2 (official), I got it to run correctly and add files without crashing. That's vague, I know, but it may help somehow. EDIT: readded screenshot here

I see an antivirus popping up right after the error. Try disabling the antivirus while installing WinPcap. I'm getting the same error on GetProcAddress - GetUserDefaultUILanguage, but it just goes on past that. It loads OLEAUT32.DLL, calls a few API in OLEAUT32 and OLE32 then it pops up a messagebox (in my case) saying it's already installed on this machine and if I want to force or abort the installation. Yours being a German system should be the last on the list of possible reasons. Rather the antivirus or an incorrect OLE component installation/registration may be the reason.

Make sure you don't have an unwanted compatibility setting for the installer. I have checked the 4.0 beta2 installer on my main system (official KernelEx 4.5.2) and it works without problems. I've also tried to compile WinPcap 4.0 beta2 and after some fiddling it appears to have succeded. However there is no NSIS installer project in the sources so the actual changes are unknown. It does appear to modify a few registry keys but they don't seem critical to the daily usage. Useful files should be packet.dll and wpcap.dll (and maybe npf.vxd which I haven't been able to compile due to out of environment space errors). pthreadVC.dll, Wanpacket.dll and npf.sys appear to be NT-related so they wouldn't have any business on a 9x system.

I'm trying to update the project to R5. However, I'm not sure I got all the changes since 4.5.2. A cumulative delta would be great. Right now I'm having problems with the sheet and vxd subprojects, won't compile if I leave the multithreaded switch (/MD) in. The switch is a new addition to R5. Also the new psapi and uxtheme subprojects are a bit difficult to integrate with the main project and especially with batch building. I've made a few minor changes to the main project to achieve that. There's also a warning in both aforementioned subprojects, related to multiple ".text" sections with different attributes (40000040) which may happen because of dosstub.exe. Finally I've managed to build the whole R5 but I won't upload it since I'm really not sure everything is in place and there's no need for such potentially unstable version at large. If anyone wants it (either Release, Debug or both), please send a PM.

You may get lucky using one of the Kaspersky tools on this page, specifically CleanAutoRun.

Please try on Tmee's server here.

Maybe the Prefetcher or something alike (even the sharing service itself) moved the files around on the HDD according to how often they've been accessed or similar basis. Or maybe just the last accessed timestamp has been touched in order to maintain an updated list of preferences. Just sayin'...

Ah, I see now! Well, I believe there may be a whole series of API calls when SiB applies its changes. it's only up to Tihiy to reveal these calls or not - after all, it's not an open-source application. So if he can't or won't do that, I'm afraid there's nobody else to help you on this matter.

Sources for latest changes...?

You could try to look into SiB's registry settings and change them accordingly, after applying various system-wide settings through the AutoIt script. If you don't know which registry settings to change and how, you may use a third-party utility that monitors the registry, apply a few sets of changes through SiB's own control panel and then check that utility's log to see what has been changed and where. But don't ask me about such utility, I have no idea if there is one to work under Win8.x+. There used to be InCtrl5, long time ago, for older OS versions.

I'll leave everything as is, for now. in CloudMe there's an option to download the entire contents of a folder by clicking on Download in the top-right. That should be an acceptable compromise to downloading files one at a time. Sorry for all the off-topic.

An application is required for synchronisation, if I'm not mistaken. CloudMe's works in 9x with UPX or latest KernelEx, but I don't know about DropBox'. Also, do they allow direct linking? I've heard about many others but none that keep files indefinitely. And free of charge, of course, because I can't pay for storage. My ISP also doesn't allow me to host files on my own HTTP server, otherwise it would've been a breeze using rejetto's HFS. My connection is at 14-15 kB/s most of the time so it wouldn't be of much use anyway.

Yeah, apparently they block direct linking to files. Everything must be accounted for: who downloaded what, when, how many times and so on... Sorry for the inconvenience, guys, I really dunno what else to do and where to go. Not even a picture can be linked to directly. This is madness! http://my.cloudme.com/#drugwash/myfiles/pics/KexDebugCloudMe.png

changed source files: link debug installer 4.5.2015.3: link debug installer checksum: link 27df222abed2315be159a9797c8da1db CloudMe (repository) seems to work fine on my 98SE under Firefox 9.0.1. Firefox would instantly crash on start when Kstub822 was enabled in core.ini - disabling it fixed the issue, Firefox runs correctly. This is with official KernelEx 4.5.2, original Kstub822.ini and no iphlpapi4, on my main 98SE machine.

So you plan on going with independent versions for those files? My thought was to keep all files together with a single version so crossversion tests would easily be tracked back to one build or another. But separate definitions for each file would do too. I have to build all files in a batch in order to build the installer package. You'll probably have to do it yourself at some point if you plan to release new versions officially. The small changes to the two project files would allow this but I'm not absolutely sure the order the core subproject is built does not affect the subsequent projects. Maybe there's no relation. Would it be a good idea to build and release debug versions of the whole package, that is a Debug installer? Since I'm building both Debug and Release versions of all files at once, I could create a new NSIS script for this. I'm thinking not to have Release versions of the files with Debug features around. Anyway, it's just an idea. EDIT: no need for a separate NSIS script, it works as is by choosing the appropriate item in script's context menu. /EDIT I'll try to pack the modified files again and upload them with direct links. That's quite difficult for me too since they do not provide an easy way to do it. Unfortunately I couldn't find a better reliable host and after my last one (FileDen) got wiped out and lost all my files, CloudMe was the best I could find. You could install their application in 98SE and connect to my repository, it may be easier to keep track of the updates. I always do the sync from my 98SE using the CloudMe application, although granted it has its quirks.

I've uploaded 4.5.2015.3 to my repository. jumper, why would you comment out the last two version variables in version.h? Since you started updating all subproject versions, I think it would be much easier to just use those globally. I've set my copy of the sources to use those variables. Also, is there a problem with the subproject order in main KernelEx.dsw (dunno if you checked my diffs)? I've slightly altered that and core.dsp to allow a full batch build in one go.

The original kexbasen.dll works fine with Kstub and I don't think installing your whole package would make a difference with this crash issue but I'll try eventually if you insist heavily. For now I use Jumper's new dlls except for kexbasen. GCC fix and Shell32 ordinal 727 are very cool to have directly in kernelex, the latter being now mandatory to run latest PotPlayer whose core dll can't be hexed anymore to another ordinal because of a new feature checking if it has been alterated. Well, installing the whole package is just one of the alternatives in hope that there could be an incompatibility between different builds. Slim chances at that but it would at least put the mind to rest. Since the only change in kexbasen is the addition of oleaut32 functions, it's safe to think the problem is right there. It's intriguing how it only crashes in ME while in 98SE it doesn't (according to mcv93us' reports). You didn't say if you tried disabling Kstub822 - it would help knowing whether it still crashes with it disabled or not. Thanks. I added Kstub822 to the test installation and so far two Firefox sessions run on subsequent clean boots have frozen. However no MPR-related crashes. I've had Firefox freezing before adding Kstub822 so this may not be related. More tests are in order. Waiting for suggestions on applications requiring KernelEx/Kstub.