Update 97-07-11 (0.6.9 release 12)

- removed password protection
- fixed a DirectDraw race condition
- added Italian translation of readme.txt (thanks, Enrico Antongiovanni!)

Update 97-06-29 (0.6.9 release 11)

- carefully alignment-tuned this version for maximum speed on Pentium CPUs -
  my standard benchmark now yields exactly the same results as under 0.6.8
  release 14 (with -w 7, of course)
- fixed the -J default if the option appears at the end of the command line
- re-fixed POTINP - it clears bit 12 and 14 unless there is a hardware
  joystick configured for port 1. If you use -J, the Mad Monks intro doesn't
  exit prematurely any more. (Goonie, this was a botch job!)
- some code cleanup and cosmetic fixes

Update 97-06-28 (0.6.9 release 10)

- added stereo sound (-S 2:s)
- added file "todo.txt" with the status of planned enhancements

Update 97-06-28 (0.6.9 release 9)

- with the time saved after password-protecting UAE, I was finally able
  to attack one of the major weaknesses of WinUAE: Sound. Sorry, you
  still won't get contiguous output on your 133 or 166 MHz machines, but
  I managed to seriously reduce the hitherto intolerable latency as well
  as the CPU usage of the output routines.
  This is a complete rewrite of the sound code with the following
  user-visible/audible changes:
  - the buffer size (-B) parameter is gone
  - sound will loop instead of pausing if UAE can't keep up with real
    time (sorry, this is an inherent 'feature' - I cannot change it).
    If your machine is too slow, things used to sound horrible and will
    continue to do so (instead of gaps, you'll get looping hiccups :-).
    Sometimes, it takes a couple of seconds for the sound to stabilize.
  - the entire timing is extremely fragile, but works fairly well on my
    configuration (NT 4.0, AWE64Gold). I haven't tested this under Win95
    or with other sound cards. Please email me if it breaks on your
    machine.

Update 97-06-25 (0.6.9 release 8)

- due to the unbelievable amount of RTFM problems some UAE users seem to
  have (resulting in the annoying flood of their dumb questions cluttering
  our inboxes), I had to implement a feature that I didn't expect to be
  necessary when I first started contributing to the UAE project: When you
  run release 8 for the first time, you will be prompted for a password
  hidden in the documentation.
             DON'T YOU DARE TO EMAIL ME ASKING FOR THE PASSWORD!
  My apologies to all _intelligent_ UAE users.
- renamed "linuxdocs" to "docs" (requested by Bernd Schmidt).
- changed default of the -w parameter to 4 (requested by Bernd Schmidt).
  This eliminates the need of using -w 4 to make numerous programs run
  under WinUAE, but will make graphics-intensive programs run slower.
  Try -w 7 or higher if you feel that your software was running faster
  under release 7.
- made floppy speed configurable (-d parameter). You normally won't need
  this.
- increased relative priority of the file system thread. It was not a good
  idea to have it running at the same priority level as UAE itself (this
  doesn't affect SMP machines).
- fixed completely broken logic in the 24/32 bit address space selection.

Update 97-06-19 (0.6.9 release 7)

- enabled the experimental gfx lib replacement (-g)
- added some buffering to reduce system call overhead in the serial port
  emulation. Term works fine if you change the "CONNECT" response in the
  modem settings to an empty string.
- increased 24 bpp efficiency

Update 97-06-17 (0.6.9 release 6)

- implemented Stefan Reinauer's latest serial patch (thanks!)
- some serial enhancements and optimizations. The serial support can now
  be called 'usable'. JR-Comm works great. Term still hangs after a
  connect.

Update 97-06-16 (0.6.9 release 5)

- fixed floppy speed to increase compatibility with buggy track loaders
- fixed disk change (argh)
- chipmem and fastmem settings are now checked for validity (> 2 megs
  of chipmem and fastmem are mutually exclusive)
- increased 32 bpp efficiency (Stephan Schaem)

Update 97-06-15 (0.6.9 release 4)

- re-enabled all of the 0.6.8r14 optimizations, resulting in a 10..20%
  performance gain on Pentium CPUs. If you have a PPro, K6 or P2, the
  impact may be somewhat weaker.
- CPU prefetch queue length reduced from 16 to 4 bytes
- implemented backwards compatibility mode for the -J option, which, if
  used without an argument, now defaults to -J Ma.

Update 97-06-12 (0.6.9 release 3)

- file system now multithreaded
- some major bug fixes

Update 97-06-12 (0.6.9 release 2)

- upgraded to 0.6.9b:
  - new option -C for controlling the copper timing. -C 224 fixes the
    dreaded 16 color hires overscan bug, but breaks almost everything
    else. If you omit the -C parameter, the old copper routines will be
    used.
- parallel port enabled (due to Windows' print job concept, it is
  impossible to provide a 1:1 mapping of the parallel port. The current
  implementation starts a new RAW print job after 2 seconds of inactivity
  on the printer port).
- mousehack enabled (if you start the program "mousehack" [contained
  in the "amiga" subdirectory], preferably from your startup-sequence,
  the Amiga pointer follows the PC pointer. This makes UAE compatible
  with non-deltamove mouse replacements such as graphics tablets.)
- (some) more speed
- the icon is back! :-)
- fast 68000 (uaef.exe), compatible 68000 (uaec.exe) and 68020 (uae020.exe)
  versions now distributed in one package

Update 97-06-07 (0.6.9 release 1)

- upgraded to 0.6.9. This is an experimental quick & dirty port and
  intended as a public beta, not a new production release. It has all
  of the CPU compatibility settings, specifically instruction prefetch
  and exception 3 emulation, compiled in and is therefore up to 30% (!)
  slower than 0.6.8 release 14 (you'll enjoy this one only if you have
  a high-end PC). However, it finally runs SOTB, among other things.
  Be warned, though: With this release, Beast requires a PentiumII-300
  to run 100% smoothly. :-)
  Starting with the next release, I will distribute the 68000 version
  as two different binaries: One compiled for maximum CPU compatibility
  and another one for maximum speed.
- The complete Linux documentation is now included in the WinUAE package.
  If you have been unable to follow the link to the UAE main page from
  the WinUAE home page and to unpack and read the documentation contained
  in the generic distribution, this is your chance to learn about the
  basics and to ask fewer trivial questions.
- The CPU's clock frequency is now read directly from the registry under
  NT, saving one second of startup time. Reportedly, Windows 95 does not
  provide this entry. I'm neither surprised nor amused.
- .uaerc has been renamed to uae.rc
- fixed minor bugs in the DirectX interface

  New in the Linux version and implemented in this release:  

- enhanced joystick support:
  - up to two joysticks with up to three buttons each
  - choice of three different keyboard replacements
    (see readme.txt / -J for details)
- serial support working (well, sort of...)
- inter-sprite collision detection
- new sprite code (again!). Unfortunately, it breaks a lot of old
  software that worked in 0.6.8, but it also fixes some problems.
- lots of bug fixes and optimizations (see linuxdocs/ChangeLog for details)
- 4 and 8 MB of chip memory

  New in the Linux version, but not enabled yet:
  
- mousehack
- multi-threaded file system

  Known problems:

- disk change not 100% reliable
- due to a linker bug, no icons can be added to the executable at this time

  Please report anything that worked in version 0.6.8, but doesn't in 0.6.9.
  Thanks.

Update 97-02-27 (0.6.8 release 14)

- well, of course, the 6x86 doesn't implement RDTSC, so release 13 is
  of little use to Cyrix owners. And, surprisingly, there are people
  who use UAE on a 486. If RDTSC doesn't work for you, use the new
  option -Q-1 to revert to QueryPerformanceCounter(). Cyrix experts
  out there: Does the 6x86 have a counter register with an accuracy
  of least 1 s that can be queried in user mode? If so, what is the
  opcode of the related instruction?
- fixed an old race condition in the sound toggling code (Pause key)

Update 97-02-27 (0.6.8 release 13)

- using RDTSC instruction instead of QueryPerformanceCounter(). This
  saves a _tremendous_ amount of system call overhead. The faster your
  machine, the more you'll profit from this. Thanks, Bernd!
  If you have a non-Intel CPU and RDTSC breaks on your machine, please
  email me. There's a new parameter (-Q) for overriding clock measuring.
  See readme.txt for details.
- cleaned up a lot of old garbage code (the handling of changes to the
  window's show state used to be an awful hack)
- good news for the fans of pseudo-"operating systems": Microsoft is
  seriously trying to iron out some of the quirks in Windows 95 that
  -X demonstrates so beautifully. A user reports that it *almost* works
  under Memphis Build 1387 (I don't care... why wait for Windows 97
  when you can start up NT 4.0 today?).

Update 97-02-24 (0.6.8 release 12)

- further Windows 95 fixes. Directory mount paths with trailing slashes
  or backslashes are now supported.
- 1081/1084 monitor can't be moved entirely off the screen any longer
- minor overall fixes and optimizations while completing code cleanup

Update 97-02-23 (0.6.8 release 11a)

- fixed Windows 95 mouse buglet. Don't bother upgrading to this release
  if you are using a real OS.

Update 97-02-22 (0.6.8 release 11)

- attempting to cd into a file now produces a meaningful error message
  instead of "Error code 83886091"
- opening read-only files on a read/write file system for reading no
  longer fails with "write permission denied"
- opening UAE in window mode with the client area being invisible at
  startup time no longer quits the emulator
- changed alignment of 24 bpp conversion - quadrupled drawing speed on
  24 bit desktops (24 and 32 bit are now equally fast). I do not quite
  understand the mechanism behind this yet. Please try it on your machine
  and notify me if your results are different from mine.
  See "Performance Issues" in readme.txt for further information.
- fixed detection of window title double click

Update 97-02-21 (0.6.8 release 10)

- added emulation of a Commodore (Philips) 1081/1084 PAL monitor.
  This used to be the standard Amiga screen in Europe. Sorry, American
  and Japanese users: I never liked the design of the NTSC variant.
  To activate, add
        -O 672:550:xyd -X (hi-res) or
        -O 332:275:xyl -X (lo-res) to the command line.
  If you're _still_ using Windows 95: It's about time for you to switch
  to a real OS, too! Few things demonstrate the inferiority of Win95 as
  pithily as this feature. I was about to implement a check that disables
  -X under Win95, but then, this would have just saved the company of the
  richest man in history the well-deserved embarrassement.
- added program icon (design: James Weatherall, Cambridge, UK - thanks!).
  Feel free to use it as the default icon for the file types .ADZ and .ADF.
- fixed '.' / '..' directory name misfeature

Update 97-02-20 (0.6.8 release 9)

- activated can_double (significantly more efficient drawing, EHB and
  dual playfield acceleration actually enabled)
- fixed some minor bugs
- some code cleanup for source tree merging in version 0.7.0

Update 97-02-17 (0.6.8 release 8)

- upgraded to Bernd's intermediate patch 0.6.8c. Changes include:
  - official sprite fix (probably functionally equivalent to my kludge
    in release 7)
  - assembly routines for EHB and dual playfield
- right control key acts as joystick button in fake joystick mode
- base priority configurable with -P. See readme.txt for details.
- double-clicking the window title now correctly enables full-screen
  mode

Update 97-02-14 (0.6.8 release 7)

- finally eliminated OS 2.x/3.x display quirks. Disabled fake ECS Denise,
  big blits still work. It is unlikely that this will break anything.
- lowered base priority even further (to IDLE_PRIORITY_CLASS). There won't
  be a fix to the ClipList problem described below. Microsoft says that
  "this behavior is by design" - flawed by design! Bah.
- tightened DMA fix, so "Disk DMA aborted" warnings will appear less
  quickly.
- added handling for DirectDraw "surface busy" error
- implemented highly experimental sprite fix. Many problems are gone,
  even more remain.
- some changes to the windowing code. Please complain if something stopped
  working.

Update 97-02-12 (0.6.8 release 6)

- maximize button maximizes window again
- GNU-ld broke some routines after reordering .o files in 0.6.8 release 5
- lowered base priority of UAE to 10. This is purely experimental, but
  seems to actually improve responsiveness in window mode until I find
  a better solution. Please report any unusual problems.

Update 97-02-11 (0.6.8 release 5)

- final chapter of the file name nightmare? Names that contain '*', '?'
  and '|' are now handled correctly.
- fixed minor centering bug
- restructured documentation

Update 97-02-09 (0.6.8 release 4)

- changes to the display settings while UAE is running in window mode are
  now allowed (but don't switch to 8 bpp - UAE will quit with a nasty
  message).
- "Great Giana Sisters" disk DMA fix. Due to a modification in UAE 0.6.8's
  MFM routines, -w 5, which is required to get this piece of quality
  software to run, broke the loader routine ("Disk DMA aborted").
- Mad Monks intro POTINP fix
- Ctrl-Win-Win doesn't kill sound any more (but is still not 100% reliable)
- more comprehensive documentation (see below)

Update 97-02-07 (0.6.8 release 3)

- corrected some Win32->AmigaDOS error code translations. This enables
  multi-assigns and fixes some other glitches (Workbench 3.1 and MUI
  can now be installed in a mounted regular Windows directory)
- you can eject disks by pressing Shift+End+F1...F4
- F12 toggles between full screen and window mode
  Side note: I'm trying to find out why IsClipListChanged() returns
  TRUE all the time even though the clip list has _not_ changed.
  If I succeed, user input will be processed much more smoothly in
  window mode.
  If you like colorful mouse pointers (the ones that the graphics card
  cannot display in hardware), you'll have to live with the fact that
  they get overwritten by UAE's display update. There's no clean way of
  preventing this. Microsoft, are you listening?
- Help key works (PgDown - this is for all the Dungeon Master freaks
  out there)

Update 97-02-06 (0.6.8 release 2)

- the file name nightmare continues: (aux|prn|con|nul).#? is illegal -
  the mapping routine is growing fast.
- disk change works again
- re-implemented sprite fix in a more moderate form - if you see any
  sprite irregularities, email me!
- in case you have been wondering why Marble Madness freezes while
  loading: -w 8 does the trick.

Update 97-02-05 (0.6.8 release 1)

- upgraded to UAE 0.6.8
  Unfortunately, not all of the 0.6.7 bugs appear to be fixed (most
  notably, the vertical display shift by one pixel under OS 2.x/3.x
  remains). This release is pure 0.6.8 code. None of my own patches
  have been applied. Expect some disappearing sprites.
  The good news about 0.6.8: Bernd managed to slightly increase overall
  performance and to dramatically increase display compatibility. An
  old intro of mine that displayed garbage beyond hope on all previous
  versions of UAE runs flawlessly now. Congratulations!
  For some weird reason, the executable's size dropped by 76 K (14%).
  If anyone feels that this version has important parts missing, please
  email me. :->

Update 97-02-04 (0.6.7 release 14)

- as an alternative to full-screen mode, UAE can now open its display
  as a regular window on your desktop (command line parameter: -W).
  Performance depends on the bit depth of your display:
  - 16 bpp: same as full-screen mode (recommended)
  - 24 bpp: unfortunately, this one is dog slow. Packed pixel mode
            is tough to handle. Are there any video cards out there
            that can do 32 to 24 bit conversion in hardware?
  - 32 bpp: this is slower than 16 bit, but much faster than 24 bit.
  The Amiga mouse is inactive until you click on the UAE window (I did
  not implement the kludgy "follow-sprite" hack). Once the mouse has
  been captured by UAE, you can get it back by pressing Alt-Tab,
  Ctrl-Esc or one of the equivalents.
- drag and drop support. You can insert a disk into drive df0: by
  dropping its corresponding .ADZ or .ADF onto the UAE window.

Update 97-02-03 (0.6.7 release 13)

- major file system fixes. lha/lzx directory creation incompatibility
  resolved. Implemented workarounds for some more Microsoft file system
  braindeadnesses (you can now create files and directories called
  "aux", "prn", "con" and "nul" - sheesh!).

Update 97-02-01 (0.6.7 release 12)

- minimalistic disk change support. Pressing End+F1...F4 pops up a
  file requester that allows you to select the .ADF or .ADZ file to
  be inserted into the corresponding drive. At 320x200 and 320x240,
  the Windows file selector exceeds the size of the screen, but if
  you drag it around a bit, you should be able to locate the desired
  file anyway. :->
  Now you can finally play your favorite multi-disk single-drive
  games (Turrican etc.).

Update 97-01-31 (0.6.7 release 11)

- minor keyboard fixes (CAPS LOCK and Del finally work)
- file names that consist of spaces (Win32 does not permit those) are
  now remapped to a sequence of ASCII chars 160. Prior to that, a MUI
  installation to a mounted Windows file system would have stumbled
  over stuntzi's cute little smiley utility.
- the mystery of "Great Giana Sisters" freezing at the initial screen
  is solved: Lame programming and -w 7 timing interact with each other
  and cause a vertical beampos of 244 never to be reached. It works
  perfectly with -w 5.

Update 97-01-29 (0.6.7 release 10)

- code efficiency should now be roughly equivalent or superior to the
  Linux version in every aspect
- UAE 0.6.7's sprite problems hopefully fixed completely - ProjectX and
  Battle Squadron reportedly work. Please do report any type of unusual
  sprite behavior.
- New parameter (-w n) allows adjustment of the ratio CPU cycles/hardware
  cycles. This is very important to optimize UAE's behavior on machines
  that are unable to complete one frame in less than 20 ms. General
  rules:
  -w 4 is ideal for speeding up things while working with applications
  (use lower values at your own risk, as they seem to break some things).
  -w 7...12 is good for demos and games. If you get choppy sound, pick
  higher values. Fine-tuning this value is a science in itself. Normally,
  -w 7, -w 8 and -w 9 are ideal. Default is -w 7.
- If sound is enabled, you can now toggle it using the Pause key.

Update 97-01-27 (0.6.7 release 9)

- a few mysterious gcc inefficiencies eliminated, some more remaining

Update 97-01-25 (0.6.7 release 8)

- because I'm still unable to locate Watcom 11.0 in Germany, I decided
  to try the choice of a GNU generation. The result runs about 20%
  faster internally than its MSVC++-compiled counterpart; the drawing
  speed is not majorly affected. Sorry about not having enabled full
  optimization in this release, but I'm still quite new to gcc under
  Win32.
- ^C should work properly again
- AmigaOS now sees the _local_ file time instead of UTC. :->
- dual playfield palette bug fixed. For now, I'm leaving the sprite fix
  as it is, because it enables you to play Marble Madness. Reports
  indicate that it kills sprites on some other games. As Bernd is
  rewriting the sprite code for 0.6.8, there is reasonable hope that
  we'll see full compatibility soon.
  
Update 97-01-23 (0.6.7 release 7)

- major optimizations (drawing speed increased by about 30%)
- 50 Hz frame synchronization can now be disabled with -S-1. If you
  want to compare UAE-Win32/DirectX with other versions, you need to
  use this (beware: MSVC++ generates code that runs exactly half as
  fast as gcc's).

Update 97-01-23 (0.6.7 release 6)

- minor optimizations (drawing speed increased by about 5%)
- mounting root-level directories works (e.g. -mCD0:d:\)
- changed default sound buffer size to 8192 (4096 bytes and less tend
  to cause hiccups with some sound drivers)

Update 97-01-22 (0.6.7 release 5)

- fixed sprite bug (Marble Madness!)
- no crashes any more when dragging screens

Update 97-01-21 (0.6.7 release 4)

- rewrote sound output again (sorry) - now much simpler than before,
  with about the same performance. The latency is still intolerable -
  I'll probably switch to DirectSound sooner or later.
- Bernd found the reason for Paul van der Valk's player routine
  generating nothing but silence. This positively affects Hybris and
  Imploder (and probably more).
- some minor fixes.

Update 97-01-19 (0.6.7 release 3)

- experimental joystick support (untested, as I don't have Windows 95)
- zfile (.adz/.roz) support for gzip'ed disk and ROM images. Sorry,
  no other formats (ZIP/LHA) implemented right now. Make sure you have
  a _32 bit_ gzip.exe installed if you want to use this!
  (ftp.lcs.mit.edu/pub/pc/fileserver/utility/unix95_6/32-bit/gzip.exe)
- sound code rewritten (won't affect machines below 200 MHz).

Update 97-01-17 (0.6.7 release 2)

- FASTER! video output ("DirectDirectX")
- re-enabled 16 bit sound and Ctrl-Amiga-Amiga detection

Update 97-01-17 (0.6.7 release 1)

- quick port of UAE 0.6.7, essentially untested
  NEW COMMAND LINE OPTIONS FOR SETTING THE GRAPHICS MODE!
  Type uae -h for details.

Update 97-01-15 (release 3):

- major keyboard fixes (won't drop characters under OS 2.x/3.x any more!)
- file protection bit semantics are retained as far as possible now
  (sorry, s, p, r and e have no equivalent in the Microsoft scheme).

Update 97-01-14 (release 2):

- pure 0.6.6 code now, no 0.6.5 artifacts remaining. Be prepared for
  minor display differences.
