@The Solutor: Does DOS Navigator appear, or does it stop on the boot messages? If it
stops, could you photograph the last messages?

My laptop’s built-in keyboard uses our xHCI USB driver, and I’ve also had a USB mouse
working on bare metal, so some support is already there. One current limitation is that
RetroOS initializes only the first xHCI controller it finds; your machine has three.

I’ve attached a smaller diagnostic image that automatically saves the kernel boot log
to C:\KLOG.TXT before starting DOS Navigator. If boot completes but input doesn’t work,
that log should help explain what happened during xHCI discovery. To retrieve it after
reboot, select Persistent disk and have a supported internal FAT or ext4 data volume
available. It cannot save back to the USB stick yet.

@LSS10999: Ideally, I’d like RetroOS to handle the motherboard setup and ISA PnP
initialization itself, so configurations like yours wouldn’t require DOS utilities
before booting.

Your HDA should likely work also, retroos would emulate the soundblaster there.

The problems addressed by EGAFIX and NEWAX are also worth investigating. RetroOS
already provides its own BIOS services for DOS programs, so some equivalent fixes might
fit there, but I’d need to examine what those utilities change before judging the work
involved.

Just to confirm: does the new image still hang at exactly the same point?

gsos wrote on Today, 07:16:

@LSS10999: Ideally, I’d like RetroOS to handle the motherboard setup and ISA PnP initialization itself, so configurations like […]
Show full quote

@LSS10999: Ideally, I’d like RetroOS to handle the motherboard setup and ISA PnP
initialization itself, so configurations like yours wouldn’t require DOS utilities
before booting.

Your HDA should likely work also, retroos would emulate the soundblaster there.

I know of HDA support which may be useful for some of my other systems on which it's being actively used.

But on that target system I need to disable it because it's the only way for me to free IRQ7 so the CT2950 card behind dISAppointment can use it.

gsos wrote on Today, 07:16:

The problems addressed by EGAFIX and NEWAX are also worth investigating. RetroOS already provides its own BIOS services for DO […]
Show full quote

The problems addressed by EGAFIX and NEWAX are also worth investigating. RetroOS
already provides its own BIOS services for DOS programs, so some equivalent fixes might
fit there, but I’d need to examine what those utilities change before judging the work
involved.

The source code for these are available and are written in assembly.
I think they're specific to affected nVidia cards and not universal.
Other vendors may also have issues with certain VGA/VESA functionalities, but then they'll need their own methods to fix them.

gsos wrote on Today, 07:16:

Just to confirm: does the new image still hang at exactly the same point?

Yes, it still hangs at the same point, at least with no visible difference.

gsos wrote on Today, 07:16:

@The Solutor: Does DOS Navigator appear, or does it stop on the boot messages? If it
stops, could you photograph the last messages?

Yeah the system reaches the DN , so I don't have anything of interesting to photograph, i'll try the attachment ASAP

LSS10999 wrote on Today, 05:01:

Though I do agree with supporting Btrfs because my current Linux use cases mostly use Btrfs for "/home" partition while using ext4 for system.

I use it for everything, because the compression and the huge space it can save, then albeit I don't use them, the BTRFS subvolumes help to keep things separated for people used to have /home in its own partition. Again sparing wasted space, given the partition are fixed while subvolumes aren't

I need to keep at least the "/boot" part as ext4 in a BIOS scenario, as that's what GRUB4DOS could handle when multibooting.

Which is what everyone did when grub2 had yet to support BTRFS and what I'm doing right now with bcacheFS (which is proving really good from first months of tests)

In case of grub4dos the simplest way is to put the kernel and initrd in the same fat/fat32 partition used as grub4dos home, but a lot of variations are possible.

@gsos

Update.

Tested the diagnostic image. it boots and all but shows that klog is created in the same directory where are /retroos /proc and so on, not on another fat partition.

Possibly hdds aren't detected just like usb, I don't know.

I tried to add a fat32 partition to the pendrive, not knowing the logic of log saves, to see if helped. It didn't

Perhaps the partition table is still broken, which make the pendrive unaccessible from windows, but linux fdisk blames for errors as well, although Linux is more tolerant and mounts the partitions anyway.

You should really fix that part because leads to wastes of time on each test, perhaps why not add a further partition to the pendrive, like I did, and save the log there, by design?

In the meantime could be a good idea to disable the commander, and redirect the output to the video at least we can have a clue and a photo could be taken.

I conducted some tests using your ISO image written to a USB stick on a different system booting via UEFI, and so far everything seems good:
- Sound Blaster emulation works as expected through that system's onboard HDA.
- GOP Framebuffer works and everything looks smooth enough.
- Tested a few bundled games (Wolf3D, Duke3D, Skyroads, Quake) and they all look good.

Some caveats, however:
- As my monitor is widescreen, GOP Framebuffer seems to maintain aspect ratio for DOS stuffs. I wonder if there can be an option to stretch it to the entire screen.
- The active DOS window (e.g. DOS Navigator) moves to the upper left and resizes when invoking the F12 menu, and returns to previous position upon exit.
- Minesweeper seems to open itself on the upper left and the game's timer feels bugged that it reaches 999 very quickly.
- Areas not covered by the active window remain untouched without any cleaning (blanking) taking place. Since the DOS window maintains aspect ratio, the border areas retain whatever that were there and could end up a bit messy.
- Is it possible to manually launch a command window? Actually, the initial DOS Navigator (DN) window can be killed, leaving the system in a state kind of like how I used to end up killing everything (including Explorer) in Win9x leaving nothing but a background.
- For some reasons, HDA Sound seems to default to "Headphone" instead of "Jack" so initially I wasn't able to hear any sound until I toggled this.
- Is there a guide on how to mount my (AHCI/NVMe) hard disks? From the Disk tab there's only A:, B: and CD. Not sure how I could mount and use my own partitions, or maybe my disks weren't even detected at all.
- Looks like there's no way to perform shutdown/reboot yet (either via APM or ACPI).

Anyway, for very modern systems this is already promising enough to get a near-native DOS experience.
I wonder if it's possible for me to customize the USB image somehow so I can bring my own stuffs for testing...

EDIT: Interesting... this time the framebuffer is properly stretched (without borders) so DOS programs look natural, but text mode font looks a bit messed up although still readable...
That system's DisplayPort output was a bit unreliable so I swapped a cable which seems to work a bit better and I'm again in a state where it maintained aspect ratio (with borders).
I guess this might be a limitation for GOP framebuffer after all... the behavior differs depending on what mode the UEFI BIOS could set.

@The Solutor

A new image is attached, it has a boot option to not start DN. It also has a possible fix in xhci discovery, and also a more robust discovery for selecting a partition as C: instead of RAM overlay. Maybe it fixes your problem, otherwise it might find a usable HD fat partition to copy to, and possibility to photograph

gsos wrote on Today, 14:25:

@The Solutor
It also has a possible fix in xhci discovery

Short answer, yessss, it works.

Longer answer: it works on one of my keyboards, while doesn't work on the other

(which proved problematic in older versions of CSMWwrap as well, you may find the related bug interesting https://github.com/CSMWrap/CSMWrap/issues/87)

Anyway also the working keyboard seem dependent on which port is connected, it worked at the first run, by luck it was connected to the "right" port, but swapping it with the problematic one lead to both keyboards dead.

As for the fat partition I just realized that I have one, but is on the NVME disk, I assume NVMEs aren't supported yet, I'll try making one on the SATA disk when I have time, but still I can't understand why an "exchange" fifth FAT partition on the downloadable image isn't taken in account.

If you need further info, I'm here.

NVME is supported, at least that is what I'm running on my machine. I'm not able to read/write usb storage yet, so need to a hd. I'm using grub to copy a ext4 image to ram and use that as a ramdisk. Opening proc/klog might give some log lines about NVMe and/or why its not selected

gsos wrote on Today, 16:28:

NVME is supported, at least that is what I'm running on my machine.

And I guess, its partitions should be visible, w/o any, linux like, mount action. Right?

I'm not able to read/write usb storage yet, so need to a hd. I'm using grub to copy a ext4 image to ram and use that as a ramdisk.

Ok, got it

I use a similar approach using grub4dos, but in my case the ram thing is optional, if use the image mounted by grub the image is writable, no matter the status of the parent physical partition. But AFAIK that works only in grub4dos/Grub1, not grub2, which implies BIOS mode not UEFI.

If there is just one non-boot partition it would use either FAT (C:) or EXT4 (/home/retroos = C: or if not existent another home), if multiple partitions it tries one on some heuristics (games / ultramid /config dir).

Ok, I had some minutes to look a bit better.

NVME is indeed detected, and fat partition as well, on DN i see a D: letter but that is empty and not writeable.

If there is just one non-boot

Do you mean that boot partitions are purposely blacklisted ? 😮

Whatever I found MC ported for DOS/Win (here on the forum), and I wanted to replace DN with it to see if it works

I copied the image unpacked it, edited the config and recompressed, but then I realized that he partition on the USB is HPFS hence mounted read only.

How you write it? are you a mac user?

Update, i found that I had 76MB hole between my partitions, so I created a new one there, on NVME and finally I got the chance to copy some SW there.

So I guess that, I understood correctly, the boot partition is blacklisted on purpose (something I didn't expect and frankly I don't understand)

Another thing I didn't expect is to see the SW on that partition shown on the top level of C: rather than inside of a classic DOS letter, a bit too "iphoneish" behavior for my taste, still not a big thing.

Aside that everything worked perfectly, audio and all, no hang, no crash, a really good job. 😀 UX, instructions, partitioning and alike needs a bit of work, though.

my two eurocents.

Yeah so A, B drives are floppy disks and D: is cdrom, which can be loaded with OSD from images in C:\FLOPPY and C:\CD directories.

C: is not a drive in the DOS sense but an alias into a unix-like filesystem that is constructed from pieces, so yes it's a bit more unixy /plan-9y or iphonish. I did it that way so I could use my linux fs as a C: for retroos. I wanted to prevent the EFI System Partition to be selected as C: , so I excluded it, but i didn't mean to cause a legit FAT partition to be ignored. I attached a new image which might solve this.

Thanks for your compliment. I think it has reached a state where it can legit provide some value in this niche retro computing/gaming community. It mitigates some problems of CSMWrap (support for legacy vga tweaks), SBEMU, SBHDA, etc.. in one, that's also why I'm interested in including EGAFIX, NEWAX and UNISOUND.

There’s still work to do on hardware compatibility, different setups, and—as your feedback makes clear—the UX and documentation.