Drugwash

So, would the FLAC play in 2.1.1 with waveout set as output? Nevertheless, dsound appears to be broken and whatever player depends on it will not be able to play FLAC, at least not the file in question. Doesn't it strike you as odd that a much newer library (dsound.dll from DirectX 9.0c) is unable to play a file type that a much older (system original, assumingly) can? Unless there's something I'm missing here... I've just installed VLC 2.0.5 in Win2000 compatibility and it plays the file in both Dsound and waveout modes, but both choppy as reported somewhere above. This could be due to missing CPU instructions (PIII doesn't have SSE2) and/or low speed (mine is 667MHz/133MHz bus). Darn, we're just assuming things.

Thank you for the listings above. It does seem that the new VLC switched to DSOUND processing/output and the old 9x-compatible version can't handle certain formats. Most likely any player that relies on dsound will fail on those formats. On another note, today I've installed the MPC-HC version mentioned by Mikl above and tried to play more media files - none of them would. This means I have to fix some other issues on my system. Once I do that (hopefully), I'll have a look at that script, Mikl.

I've performed a few tests with GOM Player, Trout and my own BASS-based player on my old 98SE. First of all, the BASS library simply crashes with a MSVCR70.dll used as MSVCRT.DLL. So any media player that uses BASS will likely crash. Not to mention BASS v2.4.10 (latest at the time of posting) crashes the audio driver in the DSP test application that comes with the package. Then, none of the players would play the 2L50SACD_tr01_stereo_192kHz.flac on my 98SE, regardless of KernelEx compatibility settings for BASS.dll, bassflac.dll or player executables. I got to the conclusion that the culprit may be dsound.dll and/or other 9x-specific media library common for all players. The file plays correctly in Trout under XP-SP3 and after installing oggcodecs, GOM Player was able to play it too under XP. Installing a few older versions of oggcodecs in 98SE did not help; codecs appear to be loaded but all players would bail out. Could anyone check the difference between the loaded modules of older (working) and newer (not working) VLC on a ME machine? We might get an idea of which file(s) break the playback.

Apologies for my fuzzy memory, but I vaguely recall something about CPU microcode update - an inf file, a dll or something like that. Maybe the new .cpl requires a newer version of that microcode file or a registry key would have to be created beforehand. A comparison between a vanilla 98SE and an updated one (with the Hotfix or ME .cpl installed, maybe with uSP3 too) might reveal the differences between files and/or registry keys (and their locations). I really don't recall what I did, but my old system that suffered so many updates and whatnot and now has a modded version of the ME SYSDM.CPL, does display CPU information fine (GenuineIntel x86 Family 6 Model 8 Stepping 3). Not sure if activating OEM information (OEMINFO.INI and OEMLOGO.BMP) has anything to do with this, but I do have those files in the SYSTEM folder. Here's the registry keys on my 98SE system: Apparently unrelated but maybe not so: can anyone point me to detailed information on (re)creation and structure of SYSTEM.1ST, located in the root of the bootable drive? Mine got corrupted a year or so ago and since then boot time increased very much. I wonder if the lack of this file triggers some hardware/software redetection at boot time and if this could help in creating/updating the CPU info displayed by SYSDM.CPL...

Glad you found the culprit! Yes, disabling Firefox's crash reporter was a good move. Looking into my system I see my version of usp10.dll is 1.471.4030.0. It doesn't say where it comes from. So far I had no crash that could link to it as the culprit but from now on I'll try to keep it in mind. I do have a certain font issue in Firefox 9.0.1 that may or may not be related to this file. Thank you for sharing the finding and the fix!

Apparently your problem has nothing to do with KernelEx. The IE6-SP1 setup installs certain updated system files that remain on the system even after uninstalling IE6; that is why you keep experiencing the crashes. Based on your description, most likely one of the updated system files clashes with Firefox or any add-ons you may have installed. You may obtain a detailed crash log from Dr.Watson after enabling it but even then it may be difficult to pinpoint the issue and fix it.

SourceForge is full of browser projects. Unfortunately, many of them are simply lame or straightforward advertising certain site/search engine/etc, not to mention most are written in .NET language, which means waste of resources or incompatible with 9x. I've tested a lot of them some time ago on my 98SE+KernelEx machine and couldn't find one to call acceptable or usable. Here's some names: dplus, Internet Surfboard, Weltweitimnetz Browser, Sleipnir, Fast Browser. I believe at this point we don't need a new browser, but a whole new Internet, free of useless additions that favor advertising, DRM and other limitations.

Related to running MPC on 98SE, I played with KernelEx. Found that the missing ordinal 61 in uxtheme.dll is actually OpenThemeDataEx which is exported by name starting with Vista. I added it to the list of stubs, recompiled uxtheme.dll but there's a problem: exporting by name works fine, but that would only help in a Vista+ environment. Trying to export by ordinal succeeded at some point, but I wanted to respect the original positions and while ordinal 1 is not defined, all ordinals will be shifted back by 5. I wonder what ordinal 1 does and if it could safely be replaced by a stub, so that original ordinal positions would be respected. I found that it takes no parameters and on XP with themes enabled returns 0x3 while with Classic style returns 0x1. I'd say 0x1 should be the return value in KernelEx's stubbed uxtheme.dll. However, in the case where ordinal 61 was recognized, a kernel32.dll error popped up: missing HeapQueryInformation. And there are more missing APIs in kernel32.dll, user32.dll (related to raw input devices) and shell32.dll. Some may be covered by KernelEx, but not all. Another ugly thing I discovered while testing in Dependency Walker: the MSVCR70.DLL 7.0.9981.0 I used as replacement for MSVCRT.DLL (and at least another one, v7.0.9466.0) doesn't seem to export __unDName under the same ordinal number and as a result the function names cannot be displayed in Dependency Walker. Apparently, DW loads __unDName by ordinal, which is 138 in MSVCRT v6.x while in MSVCR7x is 168. This is a major bugger with replacing MSVCRT by MSVCR70 since any other application out there may be doing this. Reason why I'm trying to keep uxtheme.dll's exports in original positions. But who knows how many others may have to be fixed. EDIT: Just stumbled into this WINE library that I believe covers some ordinal exports in uxtheme.dll. Should help if we ever need to add something to KernelEx's uxtheme.dll or in Kext.

Sure, I sent you a PM.

That's quite complicated and if it were to require feedback in regard to failure or success of operations, it would become even more complicated. If all this is to run when Windows already finished booting, I can provide AHK scripts that offer flexibility and feedback. But not in DOS mode or before the GUI environment is loaded. Problemchyld, I've replied to you at FD forums as well.

Sorry, I've read this but never been a batch guy so can't help with this the way you want. However, if you need all these operations to be launched after Windows finished loading and is fully functional, I can write AHK scripts, provided you first deploy the AutoHotkey exe (old ANSI version, not AHK_L) somewhere within %PATH% (Windows folder might work). First batch above will not launch fullscreen and will also not wait for the .chm to close before launching the executable. I believe this can be done by launching an .inf (remember how TweakUI 1.33 launches its CHM - but not maximized, although it may be possible - and only after closing it the installation continues?)

After having compared a few msvcr???.dll versions, I've reached the conclusion v7.0.9981.0 is the optimum choice for my system. Replacement was much more easier: open RegEdit, navigate to HKLM\System\CurrentControlSet\Control\SessionManager\KnownDLLs and add/modify a String value called MSVCRT to read MSVCR70.DLL as Data (or whatever version/path you may have to it). Personally I renamed a copy of msvcrt70.dll to msvcrx.dll in the System folder and that's the Data I set in the above-mentioned registry key. Paths are relative to the System folder, so if you deploy your library in a different folder, you'll have to use standard relative path naming. I believe absolute path may work too. If you have KernelEx installed, you'll notice a few relative paths leading to its folder. Personally I use Trout as audio player and GOMPlayer for video (beware of 2.2.53.5169, it's buggy even on XP - last 98-compatible is 2.1.43.5119 I believe and haven't tried newer versions under KernelEx).

For me the video issues appear mostly when scrolling a web page up/down, but some pages get broken when loaded, appear unscrollable (no scroll bar) and may trigger corruption if scrolling is invoked through hotkeys. None of my machines has any kind of PCI-e slot and the one in question is a VIA chipset with AGP 2x max. while the video card can take AGP 8x. It's difficult if not impossible for me to swap cards around in that machine; for one because of the physical configuration and then for fear that changing IRQs, DMA channels and other settings might trigger a software incompatibility (registry, private app settings and so on) that may even lead to system lock up or worse. I've had this machine working 24/7 for seven years and I'd hate to break it know despite those browser issues.

Compilation problems may have been addressed in regard to Leyok's repository, however the resulting code does not function correctly - either does not verify, freezes machine or crashes. At least on my side. That while the original 4.5.2 code compiles and functions correctly in the same environment. Besides, unless I'm in the dark, there has been no activity at Leyok's repository for about five months now. Nobody else seems to be skilled and willing enough to take over, so there's practically nothing worthy of announcement in first post or elsewhere other than the project being currently abandoned.

The symptoms you describe regarding video distorsion seem to match mine. There may be a (video) memory leak that quickly brings the system down. Are you by chance using an nVidia video card as I do? In this case it may have to do with the driver; I always get a crash in nvdisp.drv if I let the machine go on instead of rebooting a.s.a.p. There is a chance - at least theoretically - that something in KernelEx leads to that kind of crash; could be a stub, could be a bad/incomplete wrapper,a bad return value... Or it could well be the dreaded .NET that's linked dynamically to the VC8 runtime (msvcm80.dll -> mscoree.dll). Just dunno... I've noticed the video breaks on certain sites such as the AutoHotkey forum board if the CSS is enabled. When disabled (through the Quick Java add-on), there is no crash. I also installed the Web Developer Tools add-on in Firefox 9 and it almost always catches js/css/etc errors such as unrecognized commands or unexpected values. Another problem is with web-loaded fonts. Right on Mozilla's web site where the browser goes on first launch after installation, certain text strings appear garbled if "Allow pages to choose their own fonts..." in Tools > Options > Content > Fonts & colors is enabled. But even so, Unicode characters do not always render correctly, at least not on my system (for example in the Wordpress toolbar and its new replies panel, when logged in). A few times in my life I tried Opera but unfortunately it never managed to grow on me and got to a point where I swore it will never get on any of my machines ever again. I'm a man of my word (most of the times, at least). As for older versions of Firefox, I haven't tried any, not recently at least. I believe they wouldn't be much help anyway nowadays, considering where HTML5 and all other web languages go. Apparently even Firefox 9.0.1 (installed yesterday over 9.0 beta1) lacks a lot of knowledge in recognizing commands and parameters. Had similar video crashes with SeaMonkey long time ago and I just can't remember if I ever tested any Palemoon version altough I do have a few (with sources too) in a folder, somewhere. Anyway, if I got it right, Palemoon is just a fixed/tweaked version of the same Firefox version, so there shouldn't be much to expect from it in terms of 9x compatibility or anything. Almost forgot: don't get any hopes about that SQLite3 wrapper - it's only intended for use in AHK scripts. The wannabe-editor I'm building can only display DB tables and their contents, currently - no editing yet (if ever). And those SQLite3 library versions are quite different, it's hard to keep track of what commands work with which version, not to mention the Firefox library is custom built and if replaced with an official library, it will crash the browser instantly at start. I've just tried to load mozsqlite3.dll in my editor and it failed, therefore I won't be able to test its exported functions. (it's msvcm80.dll that's linked to mscoree.dll, not msvcr80.dll - edited to correct that)

We're kinda shooting in the dark here. What we know for sure is that 9x cannot create certain entries in settings file(s) such as localstore.rdf but it can and will read/modify them once they exist. We still do not know which entries exactly pertain to which setting and we also did not find a way to enable bookmark addition. But we know the SQLite3 database(s) can be edited in 9x. Thing is, the SQLite3 library that comes with Firefox9 is custom built and we don't know if all of its functions work correctly under 9x. Some appear to work - since we can modify and delete bookmarks and history. Bookmark addition does not, although there are at least three ways: - click the star in the address bar - use main menu: Bookmarks > Bookmark This Page (which should open a folder selection dialog) - main menu > Bookmarks > Show All Bookmarks --> right-click in empty space > New bookmark I believe this tells us there's more to it than just some missing entry in a configuration file, although I may be wrong. So far I could not test the custom SQLite3 library since my wrapper is far from being complete. And checking all the sources would require knowledge and a lot of time. In the end, the answer may be: modify this/that and recompile OR it cannot be done.

Not all entries in that new rdf are useful for this particular purpose. If the new file is copied over to an old installation, it will add unnecessary entries and will also change certain current settings. There should be only a few bookmark-related entries in the rdf that would be beneficial for a 9x setup. The rdf file can be opened as a regular text file, provided the editor can handle an eventual UTF8-encoded file. A few copy/paste operations should do.

Uhm, dunno about that - I hate toolbars with a passion, never let any of them display, other than the strictly necessary one. I'll look into it sometime later, when I get some spare time. Later edit: Well, I've looked through the profile, searching for the string "Getting Started", one of the only three I got in my standard bookmarks toolbar menu. One place to find it is in the bookmarkbackups folder in all those json scripts and the other place is in... places.sqlite. So I'm not sure how this all works but the latter seems to be the central place to store bookmarks. If I add a bookmark to the bookmark toolbar, it will be added to places.sqlite with parent ID=3 (I have a SQLite3 viewer script). I have however noticed a dramatic increase in the size of places.sqlite on the 98SE machines as compared to the original profile on XP (10.4MB vs 2.0MB). Seems there's a lot of garbage at the end of the file. Please keep that file under observation.

I've tested yesterday an older version 1.3.5 and latest (at the time) 1.4.4, both being portable versions of QupZilla. While 1.3.5 appeared to crash a little later after fiddling with compatibility in various dependency files, 1.4.4 was fairly prompt in crashing. There's a number of missing API in QtCore4.dll, QtGui4.dll, QtWebKit4.dll. There may or may not be workarounds by using ImportPatcher or similar tools - I have not tried any, for now. Firefox 9.0b1 broke my video driver last night, after having installed Java 1.6.0.27, which may not be a coincidence. I already have Java 1.6.0.7 but it's not recognized by Firefox as a plug-in. After uninstalling 1.6.0.27 and a couple reboots, Firefox is now working correctly - albeit very slow on my 667MHz Pentium III - at least here at MSFN and in a couple Wordpress blogs. Other pages might break it though. Currently, Flash is disabled through an add-on.

I have just installed and tested Firefox 9.0 beta1 on my 98SE and will soon test the stable 9.0.1. Earlier I've tried 12.0, 11.0 and 10.0.2, without success - there were crashes and/or hangs all over the place. But the 9.0 beta seems to work decently here, at least today. Not even video driver corruption as it used to happen with other browsers, but there's still time for that. One thing about bookmarks: they can be used if the SQLite3 database called places.sqlite is copied from another working Firefox installation's profile. I've got mine from a Firefox 13.0 on XP and all links are usable; new links cannot be added but at least an already existing collection of bookmarks can be used. If it doesn't work at first attempt, make sure you set mozsqlite3.dll to Windows 98SE compatibility in KernelEx properties tab. I should clearly state that my 98SE machine has a lot of manually upgraded system files on top of an old Auto-Patcher installation, so if it doesn't work for you, try to find out which files/API you're missing (Dependency Walker may help) and update your files, keeping old versions as backup for safety. Also Problemchyld's SP3 may be of help.

I couldn't find an acceptable solution for 9x browsing. Could be my video driver or anything else in this way-too-much-tweaked system, but all Gecko-based browsers bail out and break nvdisp.drv after displaying a few web pages. Problem is, I can't use a newer/tweaked nVidia driver for this MSI GeForce4 Ti4200 AGP8x video card because it loses Direct3D (or was it AGP texture?) acceleration. I'd say a lot more but it'd be useless and off-topic.

Ah, that's a relief! Since I got no feedback at all, didn't know what to think. Well, glad to be of service! Hopefully missing translations will pop up soon.

Has anybody tried that translator application or is it that bad...?

Well, if we could create a modern browser that could work natively in 9x and could handle HTML5 and all other web "enhancements" (CSS, PHP, JS, Java, Flash, etc) - allowing, of course, for any of them to be disabled at will without requiring tons of add-ons - then we wouldn't have to put up with all that bloat in Firefox, Opera, SeaMonkey and whatever other browsers people may desperately try to operate in a 9x environment. Apologies for thinking out loud, sometimes I'm daydreaming too much...

Ah, that's why I didn't see it - actually I saw nss3 but never looked through its exports. They sure make it harder and harder. I miss old times... Well, it's almost 12 hours past my bedtime, I should get some sleep. Maybe the new library will be ready when I wake up. Good luck!