jumper

I would try lowering the browser.cache.memory settings or disabling it altogether. Extensions CacheToggle and CacheSwitch can flush the memory cache out to disk.

https://learn.microsoft.com/en-us/windows/win32/api/winnls/nf-winnls-lcmapstringex : LCMapStringEx is a z9e. LCMapStringW is a z6e with identical first six parameters, so might work. After substituting with ImportPatcher, loads and immediately runs into an Illegal Instruction (SSE+) on my Pentium 2 / 98fe with KernelEx in XP mode.

Bloated sites can require an increase to 4M to avoid errors. I recommend you continue cutting it in half until you see problems on any sites.

That version runs and plays some videos on my 98fe Pentium2 and is compatible with the 95 kernel32.dll. What exactly is the problem you have encountered on 95?

0xe5e5e5e5 seems to be a JIT pseudo-handle related to CVE-2020-26950. Search DLL's Mozjs and Xul.

Good! So v1.90 RC2 is the latest that runs. I'll compare RC3 and Final to it and look for changes that might cause the Page Fault.

Not IsDebuggerPresent, but IsProcessorFeaturePresent. (IsDebuggerPresent takes no parameters.)

Did you try https://web.archive.org/web/20220521054008/http://www.oldversion.com/windows/xnview-v1-90-beta-1 ? If that works, there is also 1.90 beta 2-6 and rc 1-3 before 1.90 final.

Older versions of MinGW support older versions of Windows, but like all other older compilers don't support all recent source code syntax changes.

In File System Properties, you can increase the read-ahead buffering on HDDs and toggle Write Caching on removable drives. I just measured 877000 Bps with System Monitor on a 30MB copy from PNY USB3 32GB in 4-port mini-hub to IBM 6GB PATA HDD over USB1.1(UHCD) on Win98fe. There is also a PNY USB2 64GB, another 16GB, and a mouse in the hub. I have a USB2 PCMCIA card, but don't use it--I'm happy enough with what I have. The hub acts as a docking port and backups (with write caching enabled) and searches are fast enough. Another copy from the PNY USB2 64GB had the same peak speed, but fluctuated slower. HDD has DMA enabled and all are Int 13 units: "If this feature is enabled for removable drives, the media must be in the drive when you start your computer." (But the 16GB works just fine when hot-plugged.) All three thumb drives use Lexar JumpDrive2 drivers, but the 16GB is labeled: "Non-Lexar USB ThumbDrive". I think this one is rloew's mod. Defragging the HDD can also help. Also test searching for text on the USB drive; this will read without any writes to the HDD (a bit faster than copying here).

What non-https links are you testing with? Any ftp?

The bottleneck seems to be the thumb drive itself. 150/50*8=24 Mbps is about right for most budget USB2 flash drives (and doubles USB1.1). Try a USB3 drive to get better performance.

I was unable to get any browser later than FF28 to load (and even that crashed at exit), so I wrote a small test app to raise various exceptions. I've added to Finesse basic support for several new continuable exception types, but still need to add detail reporting and make it all more robust.

The path to the LAPTOP shared folder has changed. Update any shortcuts to it on DESKTOP.

Select ntoskrn8.sys and post again.