apple2ix/PROBLEMS

116 lines
5.1 KiB
Plaintext
Raw Normal View History

Known issues with the emulator:
Emulation Fidelity:
- Disk emulation. The emulator is not very realistic. It handles almost
all non-copyprotected disk access okay, but copyprotected or diagnostic
programs may be confused by our drive, which magically spins at precisely
the optimal speed. (see Specific Programs, below).
- Medium Resolution graphics. This is a rarely used //e mode that is to
Low-Res what 80 Column text is to 40 column text. We don't support it as
yet, although it should be relatively simple.
- Joystick. The apple joystick hardware is read by measuring the time it
takes for a specific softswitch to clear. The emulator cannot as yet
compensate for the speed of the loop used to measure the joystick --
applications using ROM routines will be fine, but custom joystick code
may become mistuned. The Range configuration feature can be used to
compensate for this.
- We don't emulate the //e's vertical blanking interval detection feature.
Graphics:
- Composite graphics artifacts are not emulated. This is not a big
interest to me (Michael), but the previous programmers have suggested it.
- B/W color setting does not apply to lores or double hires. This
generally is not an issue in practice though, as it's really only needed
to avoid color fringing in b/w hires images.
- Double Hires mode is always 140x192 color. Some applications use it as
a 560x192 b/w display however. Note that most applications indicate
which mode they want using the high bit of the dhires data bytes, so it
wouldn't need to be a preferences setting.
- If an 80-column mode is selected in the low-res emulator, nothing will
be written to the display -- the image from the last video mode will
remain on the screen. If a menu is brought up it will `stick'. This may
make people think the emulator crashed, although it will recover if the
application returns to 40-column mode.
- There is no 80-column mode for svgalib apple2. It would be possible on
standard VGA, although there are not enough colors to do flashing letters
by palette tricks as we do now.
- In X, the emulator window takes over the palette, causing all the other
windows to be miscolored. We only need 20 or so colors, so it would be
better if we could share the common palette.
Keyboard:
- Key Handling is a little messy, especially with the way the X support
is bolted on top of core code originally designed for SVGA scancodes.
- Presently, the Backspace key is interpreted as Left-Arrow (Code 0x88).
It could be argued that it should be interpreted as Delete (Code 0xff)
instead. Real Apples had no seperate Backspace key, but the //e's Delete
key was in an analogous position to the PC's Backspace). The PC
keyboard's Delete is assigned to 0xff (in //e mode).
- If you're using Apple ][+ mode, the keyboard may appear broken. That's
a feature -- the program is imitating the historical ][+ keyboard layout.
Press F5 to see a map. Maybe we should add a means to turn it off.
- On my X system, without explict xmodmap or xkbcomp intervention,
Break (reset) won't work with XKB enabled, and PrintScreen (reboot) won't
work with it disabled.
This is X's fault, and I've sent them a bug-report, but maybe we should
consider moving the keys (again...)
Miscellany:
- Older versions of this file note problems with scandir() in the
presence of orphan symlinks in some versions of the C library. Hasn't
been a problem for my glibc 2.1.2 system.
- Presently the programs are not automatically set SUID on install.
- The emulator requires ROMs for both ][+ and //e even if only used to
emulate one of them.
- In theory, typing "Ctrl-Left Alt-Pause" should be equivalent to
"Ctrl-Open Apple-Reset" and make the //e reboot. This doesn't work
properly in practice. Somehow CXROMINT is not cleared at disk-interface
detection time, so the disk won't boot.
(Ctrl-Right Alt-Pause activates diagnostic mode, as expected.)
Specific Programs:
- Some programs (Computist's Nibbler, Sword of Kadash Master copy for
example) lock up. It appears (in debugger) that they are reading the disk
with the motor off. Perhaps they pulsed the real Apple's drive motor to make
it turn slower?
- ProDOS will refuse to format disks, claiming that the disk is too slow.
- ``Alternate Reality: The City'' seems to get jammed, rapidly changing
the video mode. I'm not sure if this is a real failure, or just a special
effect that takes longer than I'm willing to wait to finish (mode switches
would be much faster on a real Apple.) I can get into the program
with some nontrivial debugger manipulation to `short out' the offending loops.
- ``Moon Patrol'' will crash in ][+ mode at the end of the first level
(ie: Point E). This is caused by a "02" opcode at address 1E15, an
undocumented HANG instruction. //e mode works, since the 65C02 treats
this instruction as a no-op.
Moon Patrol may actually be //e-only. Yet this seems a gratutious
incompatibility, since the fact you can play the first level (and further
ones using the debugger) under ][+ mode indicates it doesn't need any //e
features. Perhaps the bug was introduced by the people who cracked it's
copy protection.