the flashrom project a JEDEC chip needs to be emulated here. Tests with BIOS
images designed for the i440FX chipset are okay, but for the i440BX chipset
some more commands need to implemented.
- Store BIOS ROM start address and use it instead of hardcoded BIOS_MASK.
- Added stub for BIOS ROM access switches.
- Verify checksum of legacy Bochs BIOS, too.
selected PCI chipset. When testing Bochs with a real BIOS image the programming
sequence for the ESCD unexpectedly overwrites some BIOS locations and it fails
to reboot. For the i440BX AGP chipset development it is required to test with
a real BIOS image, but currently reporting "unknown flash type". For now I have
added two stubs with unmodified behaviour.
- Access to VGA registers via Banshee address space is not memory mapped.
- Added BX_ERRORs for unimplemented unaligned access to i/o registers.
- Access to the unimplemented YUV address space now causes BX_ERROR.
- Removed some TODO comments.
- Shortcut parser sets a flag in case of error.
- String parameter handler resets to old value if the new one is invalid.
- User button handler shows a message box if the parser signals an error.
- Added support for split screen mode and horizontal pel panning.
- Text mode cursor now correctly visible independent from blink state.
- Workarounds for the existing text_update() API.
set accordingly.
TODO #1: Text mode features split screen and h/v pel panning missing for now.
TODO #2: Show error messages related to dialogs in message boxes.
- New attempt to fix the host-to-screen bitblt issue with monochrome source.
The specs are wrong at this point. The rules for calculating the real source
pitch are still unknown. These hacks should fix most the glitches.
- Added source colorkeying support for host-to-screen blt with pattern.
- Disable i/o and memory access in PCI command register before probing and
setting up base addresses and re-enable it when complete for selected device.
- Legacy BIOS should not enable busmaster function in PCI init.
- Banshee: Added colorkey support for host-to-screen blts without pattern.
- Clearing base address registers in reset is not okay. Maybe we have to
unregister i/o and memory regions here. This affects other devices, too.
- Some other small changes and typo fixes.
- Added support for "byte packed" monochrome source data.
- TODO #1: All modes except 16 bpp color source untested (based on specs).
- TODO #2: byte / word wizzling, other color orders, 24 bpp source dithering.
- Added support for undithered 24 bpp source data.
- Fixed "standard monochrome" mode in case of transparency.
- TODO: byte / word wizzling, other color orders, 24 bpp source dithering.
- Implemented screen-to-screen blt function (untested).
- Fixed and improved "Rectangle fill" function.
- "SGRAM fill" function rewritten similar to the other functions.
- TODO: cpu-to-screen function (handle data written to bltData register).
- Implemented destination colorkey support for the remaining functions.
- Some other small changes in the Banshee code.
- TODO: Source colorkey support for host-to-screen blts.
- Fixed SGRAM fill command (y start and size have one more bit - spec is wrong).
- CMDFIFO packet type #5: texture address is a 32 bit register number.
- Setting clut_dirty is required when writing to most of the fbiInit registers.
- Added colorkey support for screen-to-screen pattern blt, rectangle fill and
pattern fill commands.
- Consider colorkeying enabled when branching to normal or pattern blt functions.
- Changed 2D TODO messages to BX_ERROR and some other minor changes.
feature of packet type #5. Added some W2 cases for texture writes, modified
code for linear LFB writes and added BX_ERROR messages for the WN cases. Some
more test cases are needed for a complete implementation.
- Real fix for the Voodoo2 DirectX issue after CLUT pens implementation.
- Banshee hardware cursor now uses the CLUT pens for the background.
- Banshee hardware cursor now uses the same 3D start address as update().
- Set clut_dirty flag when writing to fbiInit registers. This fixes a Voodoo2
issue when using the DirectX driver.
- Report another unimplemented case in the texture code.
