Mac99 v0.03

• GDI driver: perform bounds checks when doing bitmap copies from/to framebuffer, fixes bugcheck in NT4 when rendering part of a window off the screen
• stage1 (nvidia): set the framebuffer offset to 0 like OS9 NDRV does, should fix the framebuffer rendering issues

Mac99 v0.02

Two bugfixes:

• Implement low level ATI modeswitch code, should fix clamshell;
• Invalidate data & instruction caches after writing exception handlers, should fix systems with an ATA-6 controller and PowerPC 7455 or prior.

Mac99 v0.01

Initial release for Mac99/Uni-North systems.

Only supports laptops due to lack of working USB drivers.

With the current HAL, no drivers for PCI devices will work unless they have been written to handle the big-endianness of the PCI bus.

Tested on an iBook G3 Snow (PowerBook4,3) and a PowerBook G4 (PowerBook5,2), both of which dualbooting NT 4 and NT 3.51.

Driver changes:

• halunin: initial release
• offbdll: initial release. This is a fork of framebuf.dll but with additional code to handle the big-endianness of the PCI bus on uni-north systems; a framebuffer is allocated out of main memory, and when operated on is then copied to the real framebuffer with correct MSR_LE swapping
• atapimio: added support for Apple's ATA-6 controllers elsewhere on the PCI bus
• adb: fixed bug with Ctrl-Alt-Backspace handling in which the Ctrl-Alt-Backspace combo would only work once per boot

v0.07

• arcfw: fix iso9660 implementation to allow ISOs where extensionless files have the filename with single dot at end
  • this fixes the original dump of NT 3.51 PMZ

v0.06

If you are running under emulation, get the latest version of dingusppc-nt to fix the broken framebuffer

• arcfw: implemented SCSI (MESH) driver
  • (still no NT driver for MESH yet)
• stage1: attempted to fix issues with physical memory enumeration
• stage1: fixed getting framebuffer address for certain cards
  • this fix breaks older versions of dingusppc-nt which was relying on the incorrect behaviour, get the latest version
• arcfw: when reading/writing ATAPI only send test unit ready
  • should fix read failures / slow read speed with certain IDE optical drives
• HAL: PCI slot interrupt fixes
• drivers: rebuild for supporting (still unreleased) uninorth port
• offrmdrv: fix text mode setup on systems with big endian PCI bus (like uninorth)
  • this is done by allocating a framebuffer in main memory for setupdd to use, and correctly copying it to the real framebuffer every 100ms (10fps in text setup should be fine, right?)
  • the framebuffer allocated is 640x480, copied to centre of the screen, rest of framebuffer is filled black
• offrmdrv: pass flag to GDI drivers for systems with big endian PCI bus (like uninorth)

v0.05

Now compatible with NT 3.51 (build 1057 and above, untested on real hardware for the time being)! Also seems to fix a lot of bugchecks in GUI setup, done lots of installs under emulation with no bugchecks whatsoever.

• arcfw: fixes "sticky keys" on ADB keyboard
• arcfw: fixes bug when formatting ARC system partition, should resolve the "no system partition" issue
• arcfw: now has a system identifier, fixes compatibility with NT 3.51 PMZ
• arcfw: relocate out of low 8MB, allows earlier NT builds without relocations in kernel to try to boot
• HAL: delegates arc environment read/write to a worker thread, increases stability
• HAL: adds NT 3.51, NT 3.51 PMZ, NT 4.0 beta 1/2 compatibility
• atapimio: heavily lowers delays when resetting controllers; not needed even on real hardware and fixes NT4 checked build
• drivers: fix compatibility with NT 3.51

v0.04

• arcfw: Fix instruction patcher bug for compatibility with PowerPC CPUs designed after 1996 (#14)

v0.03

• loader: Fixed 32bpp modeswitch on ATI Rage4/Rage128
• arcfw: Added support for trying to load drivers ramdisk again (if initial load failed for whatever reason)
• HAL: Fixed PCI bus enumeration on Yosemite
• HAL: Cuda driver works on real hardware

v0.02

• ATI Rage128 (Rage4) has registers in different locations, added code to attempt to get this working;
• One ATI DSP timing register write was writing to the wrong half, rewrite to what is actually happening (on Lombard)

v0.01

Initial release, for Gossamer-derivatives only.

Only the bare minimum hardware support, and instability has been noted (expect bugchecks!)

Notably: as far as I can tell, you WILL bugcheck at the end of graphical setup, and then you WILL bugcheck at first boot after that.