Chip 2001 Mobile/Chip_Mobile_2001.iso/palm/system/_palmemu/palmemu.exe/Docs/_OldNews.txt

Text File | 2000-08-14 | 152.1 KB | 3,744 lines

========================================================================
News (latest changes) archive for the Palm OS Emulator
Copyright (c) 1998-2000 Palm, Inc. or its subsidiaries.
All rights reserved.

Please send bug reports, comments, suggestions, etc. to [e-mail]
========================================================================


Changes for 3.0a6 (4/19/00)
----------------------------
*    (Unix) Support POSER_DIR environment variable. If this variable
    exists, Poser uses it as the root for all files it uses (including
    preferences, Gremlin States, log files, and skin files). Otherwise,
    it uses HOME.

*    External skin support. A "skin" is something that defines the
    appearance of an application. Palm OS Emulator uses skins for each
    device that it emulates. Until this release, skins were built into
    Poser and couldn't be changed without recompiling the application.
    Now, skins are stored in external files that Poser locates and uses.

    When starting up, Poser looks for a directory starting with the text
    "Skins" (that is, it may be "Skins", "Skins.Palm", "Skins.Personal",
    etc.). Poser looks for all such directories in its own directory
    (that is, the directory in which it resides on Mac or Windows, or
    POSER_DIR on Unix), as well as that directory's parent and
    grandparent directories.

    For each such directory it finds, it scans the contents for files
    ending with ".skin". These files are text files that describe the
    skin. Associated with each .skin file are two graphics files that
    define the appearance of the skin (in single and double size mode).

    All of Poser's previously built-in images are now provided
    seperately in a Skin archive available on Poser's Web pages. This
    archive includes a ReadMe.txt file that describes how to create your
    own skins.

*    Internal change: as part of the skin implementation, a number of
    file handling and streaming classes were replaced with a new, more
    consistant set. The following classes:

        FileReference
        StreamHandle
            BufferHandle
            FileHandle
        PPStream
            ChunkStream

    Have been replaced with:

        EmDirRef
        EmFileRef
        EmStream
            EmStreamBlock
            EmStreamChunk
            EmStreamFile

    Testing note:

        All File I/O is affected by this change. The following places
        need to be tested to make sure they're not broken:

            Reading
                AutoLoad, AutoRun, AutoRunAndQuit directories
                Files specified to be loaded via cmd line.
                Loading PDB, PRC, and PQA files.
                Loading skin files (.skin and .jpg files).
                Loading Preference file
                Loading PSF files
                Loading ROM files

            Writing/Creating
                Exporting PRC, PDB, and PQA files (both via the menu
                    item and via HostExportFile).
                Saving screen shots.
                Saving PSF files
                Saving preferences
                Saving downloaded ROM file.
                Saving Logging files
                Saving Profile files.

*    Our patch to SndDoCmd now initializes the return value. [David
    Slotter]

*    (Unix) AutoLoad, AutoRun, and AutoRunAndQuit directories now
    supported.

*    (Windows) Binder utility now merged into Poser directly (see the
    "Save Bound Emulator..." menu item).

    "Binding" Poser means to create a new version of Poser based on the
    settings in the currently running Poser. A new executable is
    created that includes a ROM image, configuration settings, skins,
    and optionally a RAM image (the stuff that goes into a .psf file).
    All of this information is taken from the current settings in the
    running Poser. The result is a new Poser that is self-contained and
    runs stand-alone.

    When binding Poser, you'll be presented with two options:

        * Create a version that always creates a new session based on
          the current emulator settings

        * Create a version that restores the current session

    Selecting the first option creates a version of Poser bound with a
    ROM image, configuration settings, and appropriate skins. When
    launched, this version of Poser will always create a new session
    based on those settings.

    Selecting the second option creates a version of Poser that is also
    bound with a RAM image. When launched, this version of Poser will
    always restore the state that is stored in that image. This is
    effectively the same as reloading a .psf file.

    In both cases, Poser functionality is abbreviated. Because much
    information is now specified by the bound items, menu items used to
    create, save, and load sessions and session files are removed. And
    since bound Posers are intended to be used in demo, kiosk, or other
    closely controlled environments, most developer-related menu items
    are also removed.

*    Fixed a problem where NetLibReceivePB could return netErrParamErr
    instead of a more informative error code. [Frank Yellin]

*    Fixed a problem where NetLibReceivePB where an in/out parameter
    passed to recvfrom was not initialized first. [Frank Yellin]

*    When emulating NetLib calls by turning them into host TCP calls,
    and an error occurs, add a log message (if NetLib logging is turned
    on) that describes how the host error code is translated into a
    NetLib error code.

*    Implement the documented -skin and -silkscreen command-line
    options.

*    Got rid of problem that prevented Poser from running on Macs with
    more than 256 Meg of RAM (and which could have lead to problems on
    other platforms as well).

    Details:

    (First, some terminology. In the discussion below, "host address
    space" refers to the memory that any Windows, Mac OS, or Unix
    application sees and manipulates. "Emulated address space" refers
    to the memory that the Palm OS or a Palm OS application sees. It is
    up to Poser to make the host address space look like emulated
    address space.)

    Poser emulates memory by allocating some buffers in the host's heap
    to "back" the various memory ranges utilized by the Palm OS. For
    instance, it creates buffers to represent RAM, ROM, and the
    Dragonball registers. Whenever a memory access is performed, Poser
    determines which of these buffers is required, calculates the
    correct offset into that buffer, and fetches or stores as
    appropriate. These buffers are used to create the illusion of the
    emulated address space.

    On occassion, it is necessary for Poser to magically make more
    memory "appear" to the emulated process. That is, the emulated
    process needs to be granted access to memory other than that
    normally needed to emulate a true Palm OS environment. A chief
    example of this is when loading a .prc file. Poser performs the
    install process by creating the database, creating the records in
    the database, and then copying the record contents from the .prc
    file to the records in the database it created. This copying is
    done by calling DmWrite. But the source of the DmWrite is a buffer
    that Poser allocated to hold the contents of the .prc file. This
    buffer is not one of the standard buffers Poser uses to emulate the
    any of the standard memory ranges that a Palm OS process can access.

    In order -- in this example -- for DmWrite to access the bytes in
    the file buffer Poser allocated, Poser temporarily "maps" arbitrary
    ranges into the emulated address space. Ranges that are mapped in
    can be accessed by Palm OS processes. However, this mapping did not
    undergo any address translation. That is, if the buffer holding the
    .prc file happened to have been allocated at memory location
    0x01234560 by Windows, the Mac OS, Unix, or whatever, then Poser
    would map that buffer into memory location 0x01234560 in the Palm
    address space.

    This was OK, except when memory ranges mapped into the emulated
    space overlapped with standard memory ranges in the emulated space.
    For instance, on Macs with lots of RAM (greater than 256 Meg, at
    least), Poser could easily make a memory allocation that return an
    address > 0x10000000. If the buffer that was allocated were actually
    in the range 0x10C00000 - 0x10E00000 (or whatever the ROM range
    was), then when Poser attempted to "map in" that buffer, it would
    fail because of the overlap.

    Now, finally, this problem is fixed. Poser now performs memory
    address translation. Segments of memory mapped into the emulated
    address space are now mapped in so that they appear at 0x60000000 -
    0x6FFFFFFF.

*    When checking for the validity of calls to shared library
    functions, don't allow a refnum of zero. Also, display a specific
    error message when an invalid library refnum is detected. [Frank
    Yellin]

*    (Unix) Fixed problem with crashing on Sparc systems when loading
    .prc files. [Frank Yellin]

*    Silently ignore a bug in an internal Palm OS 3.3 function
    (PrvConvertDepth1To2BW) that causes it to walk off the end of an
    allocated memory chunk by a single byte when reading from it.

*    Slightly more rigorous PC validation. Previously, I'd just check
    to see if the PC were in a valid address space (RAM, ROM, Dragonball
    register, etc.). Now only RAM or ROM addresses are allowed. [Frank
    Yellin]

*    HostFFlush (HostLogFile()) now writes the log file to disk.

*    Re-hooked up the "Logging Options..." button in the New Gremlin
    dialog box. In order to do this, utilized some cross-platform
    dialog handling routines that are slowly being phased in. As a
    result of these new routines, the dialog box has been changed from a
    multi-panel dialog on Mac and Windows to a non-panel dialog box (as
    was on Unix).

*    (Unix) No longer hide and reshow the main window when establishing
    a new skin (either as part of the Skins dialog, or as part of
    loading/creating a session). This process was an attemp to work-
    around an FLTK problem where the X frame window didn't properly
    resize itself. This workaround didn't work, and another one has
    been put in place, so the old one's taken out.

*    (Unix) Change the filter button to use only lower-case suffixes.
    With both upper and lowercase, the text in the filter button would
    overflow its bounds. [Jon Aslund]

*    (Unix) When opening a saved session file while a current session is
    already running, Poser would attempt to close the first session's
    thread twice, and then attempt to start the second session's thread
    twice. This would often lead Poser to hang. Now each of those is
    done just once. [Jon Aslund]

*    (Unix) Instead of crashing, report errors that occur while creating
    a new session or loading an old one. [Jon Aslund]

*    (Unix) Added BSD/OS 4.1 support. [Bert Driehuis]

*    Allow TsmGlueGetFepMode and TsmGlueSetFepMode access to system
    globals.

*    (Windows) Fixed bug where creating a new session in Poser, quitting
    without saving, and then starting Poser using the -psf command line
    option would result in a new session being created again instead of
    loading the specified session file.

*    ROM Transfer project updated to use SDK 3.5.

*    Internal change: previously, when Poser had to manipulate data in
    Palm OS format, it used the structs in the Palm OS headers directly.
    However, that assumed that we could convince all the compilers used
    to build Poser to lay out those structs correctly (that is, with the
    same size and alignment as the 68K compiler used to build the Palm
    OS ROM and Palm OS applications). This assumption is almost true,
    but not quite. Therefore, with this release, a new mechanism for
    manipulating Palm OS structs is used. This mechanism (see the
    EmPalmStructs.xxx files) allows Poser to access structs in a
    platform-, compiler-, and endian-independent fashion.

    Currently, this new mechanism is used only in the area of sending
    and receiving packets from external clients (debuggers and scripts).
    Eventually, it will be used in all aspects of Poser operation. Once
    this switchover is complete, we'll be mostly independent of the Palm
    header files, and will probably be able to get rid of most of
    contents of the the Byteswapping.xxx files.

*    Fixed problem with our patch to NetLibReceive. The exact problem
    being solved was one where NetLibReceive was called, received some
    data, the application modified that data in the buffer that received
    the data, and then call NetLibReceive again. No bytes were
    available, but the contents of the buffer were updated with the
    contents of the previous receive, thus wiping out the modified
    buffer contents. The application was relying on the buffer not
    changing, and so failed when it actually did.

*    Fixed problems with ROM Transfer not running on Palm OS 1.0.
    [Vladimir Amarante]

*    Fixed problem with trying to boot a ROM from a Palm V device that
    had originally been installed with Palm OS 3.1 but was lateer
    upgraded to Palm OS 3.3. [Douglas R. Shefsky]

*    (Windows) Fixed a problem with not reading the entire preferences
    file.

*    (Unix) Fixed problem with clicking in a non-active area of the
    window (that is, outside of any button or LCD area).

*    (Unix) Added ROM Transfer project (with prc-tools 2.0 compatible
    Makefile) and pre-built ROM Transfer.prc application that can
    be installed with something like pilot-xfer.

*    (Unix) Fixed problem that could lead to unexpected results when
    running certain dialogs after other certain dialogs (FLTK wasn't
    clearing out a reference to a widget in the first dialog,
    causing the second to crash).

*    (Unix) Remapped EINPROGRESS (as returned by sockets functions)
    to netErrWouldBlock instead of netErrSocketBusy (as NetMgr.h
    would suggest). This change allows Network HotSync to work
    between a Unix box and a Windows box.


Changes for 3.0a5 (03/06/00)
----------------------------
*    Interim internal release dates:
        3.0a5e1        -    02/25/00
        3.0a5e2        -    02/25/00
        3.0a5e3        -    03/01/00
        3.0a5e4        -    03/03/00

*    Internal change: major change in the way the Dragonball and
    DragonballEZ registers are handled for different devices.
    Previously, there were two sub-systems: one that handled Dragonball
    registers, and one that handled DragonballEZ registers. Any device
    differences withing those two categories were handled with a set of
    conditional statements. That mechanism started breaking down (the
    conditionals were getting too twisty), so now device support is
    broken down with a set of classes having the following hierarchy:

        EmRegs - fairly abstract base class
            EmRegs328 - mildly abstract base class implementing Dragonball emulation common to all 328 devices
                EmRegs328Pilot, EmRegs328PalmPilot, etc. - concrete base class filling in the missing, custom pieces for a specific device
            EmRegsEZ - mildly abstract base class implementing Dragonball emulation common to all EZ devices
                EmRegsEZPalmIIIx, EmRegsPalmV, etc. - concrete base class filling in the missing, custom pieces for a specific device

*    Added Symbol 1700 support. Note that Symbol-specific hardware is
    not yet emulated. Attempting to use scanner or Spectrum facilities
    will hang the device. [Ron Kupke, Darren Kropp, John Duhart, Harini
    Bharadvaj]

*    Changed the way serial emulation is performed a little bit.
    Before, I'd look at the UART-enable bit to determine whether or not
    to open or close the host's serial port. However, that would lead
    to problems on Unix when using pseudo-devices. While opening the
    serial port, the Palm OS would enable and disable the UART several
    times as it was configuring the communications. This would lead to
    the host serial port being opened and closed several times. But
    Unix pseudo-devices don't react to that usage very well. So now,
    Poser looks at the serial line driver enable bit. This bit is
    toggled just once at the beginning and end of a serial port session,
    just like we want. [Eskil Heyn Olsen]

*    Fixed problem with serial emulation not working on Palm VII EZ.

*    Fixed problem where reloading a .psf file that had a Gremlin
    running would not correctly install the Gremlin-specific logging
    options.

*    Fixed problem affecting Gremlin reproducability. There was a patch
    on KeyCurrentState to randomly set key bits when a Gremlin was
    running. However, the RNG used to set the bits was in no way
    synchronized with the RNG used to generate Gremlin events, nor was
    it ever reset in any way when a Gremlin was started or restarted.
    The effect was that any application that was sensitive to the
    results of KeyCurrentState would react randomly. One example of
    this was the Address Book, which looked to see if the Address Book
    hard key was pressed in order to see if the user wanted to beam
    his/her business card.

*    Fixed problem affecting Gremlin reproducability. Some applications
    are sensitive to time and timing issues. Gremlins now tries to make
    sure that all forms of time (including the DragonBall cycle counter,
    the kernel's tick counter, the Palm OS tick counter, and the
    Dragonball real-time clock) are all set to consistant values when a
    Gremlin is started and restored when a Gremlin-active .psf file is
    reloaded (the cycle and tick counters are set to zero, and the RTC
    is set to 3:00pm). Also, when a Gremlin is running, the real-time
    clock is unhooked from the host's clock (normally, the values read
    from the Dragonball real-time clock are generated from the host's
    clock functions). Instead, the RTC is incremented assuming that the
    average instruction takes 12 cycles, and where the cycle time is
    based on the Dragonball's PLL registers.

*    Set the chip and mask IDs. This is set to the 1H58B part for the
    328, and 1J83G for the EZ.

*    (Unix) Fixed problem with Logging Options dialog hanging.

*    Check PC and stack pointer when doing a TRAP #$F dispatch.

*    More up-to-date instructions for Network HotSyncing with Poser.
    [Brian Mathis]

        These instructions are intended for developers or other POSE users
        who want to synchronize the Palm OS Emulator with the same PC they
        are running POSE on. Modifications to these directions may be
        needed if a user wishes to synchronize with a PC on an attached
        network. They were written on a Windows machine, and probably have
        Windows specific instructions in them. It should be pretty painless
        to adapt them to other platforms that support hotsyncing.

        These are current for version 30a3 and 30a4 of POSE emulating a Palm
        V, OS 3.1, and Hotsync version 3.0.1/3.0.4, Windows 98 & NT.

        ----Configure Hotsync on your Desktop PC

        1. Right-Click the hotsync icon in the system tray

        2. Enable "Network". A check mark will appear next to it if it's
        already enabled.

        ----Configure POSE:

        1. Right-Click on the POSE window to bring up the Menu. Choose
        "Settings/Properties".

        2. Check the box next to "Redirect NetLib Calls to TCP/IP".

        3. Click "OK"

        ----Configure Hotsync on the Palm

        1. Open the "HotSync App".

        2. "Menu/Options/Modem Sync Preferences" - set to "Network"

        3. "Menu/Options/LANSync Preferences" - set to "LANSync"

        4. "Menu/Options/Primary PC Setup" - enter "127.0.0.1" under
        "Primary PC Address" (the middle one)

        5. Under the "Modem Sync" icon on the main Hotsync screen, tap
        "Select Service". Then tap "Done". You don't need to change any
        info, you just need to have something selected. You can create a
        new profile if you like.

        6. Tap the "Modem Sync" button. Your PC should now kick into
        action, and will probably ask you to choose an account, or create a
        new one. It is highly recommended that you create a new account
        just for POSE.

    See also the following KnowledgeBase article:

        <http://oasis.palm.com/devzone/knowledgebasearticle.cfm?article_id=1674>

*    Information from TRGnet on TRGpro support. [Mike Walter]

        TRGpro support for POSE

        Version 3.0a4 includes limited support for TRGpro emulation. This
        version emulates the FAT file system API provided on the TRGpro. It
        does not support the lower level CompactFlash API or the enhanced
        audio API on the TRGpro.

        In order to emulate the TRGpro Fat file system, do the following:

        1.    Obtain a PalmOS 3.3 image from any device. (PalmIIIx, PalmV,
            etc)

        2.    Modify file "Palm OS Emulator.ini" in your Windows
            subdirectory. Change to line FfsHome to point to a subdirectory on
            your drive.

                Example   "FfsHome=C:\POSE\

        3.    Put desired Palm programs and databases into subdirectory
            specified above.

        4.    Run Emulator.exe.

        5.    Install the programs ffs_pose.prc and CFPro.prc onto emulator.

        6.    Run CFPro to move files back and forth between Emulator and
            PC's subdirectory.

        Developers writing code for the FAT file system can now debug using
        POSE.

        More documentation will be provided later, specifying exactly what
        is and is not supported with the emulator. For example, card
        insert/remove messages are not supported with this version.

        For those of you interested, the files modified/added to the POSE
        software base were, Platform_FfsLibWin.cpp and
        TrapPatches_FfsLib.cpp.

        Keep in mind that this is a first beta release, if you have any
        questions or problems, email TRGnet directly.

        Zip file containing ffs_pose.prc and CFPro.prc can be downloaded
        from:

            http://www.trgnet.com/download/pose/fat.zip

        POSE version 3.0a4 or greater can be downloaded from:

            http://www.palmos.com/dev/tech/tools/emulator/>

*    "Pen" events are essentially created as needed and inserted into
    the Palm OS on demand from mouse events by way of a patch on the
    SysEventGet function. However, until now, there was no
    corresponding patch to EvtSysEventAvail to make that function return
    true if the host mouse was down. Now there is. [Michael Nordstrom]

*    Added back long paths for MRU menus. [Chris Antos]

*    (Windows) Made F11 minimize the window. [Chris Antos]

*    (Windows) Swapped positions of About and Exit menu items. [Carlton
    Craighead]

*    Simplified Skins dialog.

*    Simplified Gremlins dialog. Instead of entering magic numbers like
    zero and -1, use checkboxes to enable desired features.

*    Added Palm IIIc skins. Renamed "Color Device" to Palm IIIc.

*    Added UsingPOSE.pdf.

*    (Unix) Implemented "File/Save" menu item. [Dan Poirier]

*    (Unix) Fixed problem where UART::Initialize would get called twice
    when reloading a .psf file, triggering an assert. [Dan Poirier]

*    (Windows) Fixed problem with not setting the skin palette when
    using the generic skin on 8-bit monitors. [Andreas Linke]

*    Removed SndInit patch that would set the system volume level to
    zero, mirroring what happens in the simulator. However, this
    caused problems with people not hearing sounds when they expected
    to. So, removed the patch. If you don't want sound, turn it off
    in the Preferences/Properties dialog. [Frederic Paolucci]

*    (Unix) Building under Solaris 2.6 (with gcc installed) is now
    possible. It's probable that building under other SunOS environments
    won't work.

*    (Unix) The window resizes to match skin selections.

*    Export Database dialog now primarily lists the actual database name,
    with the application icon name in parentheses if it differs.


Changes for 3.0a4 (1/28/00)
---------------------------
*    Interim internal release dates:
        3.0a4e1        -    12/02/99
        3.0a4e2        -    12/30/99
        3.0a4e3        -    01/07/00
        3.0a4e4        -    01/23/00
        3.0a4e5        -    01/26/00
        3.0a4e6        -    01/27/00

*    Yow...time to start updating those copyright statements...again...
    and this time there's a lot more files!

*    When installing a .prc file, check for duplicate resources.

*    Provide better error messages when something goes awry while
    installing an application or database.

*    Added support for Palm IIIe and Palm Vx:

    - Added as items to Device menu.
    - Added code to save those devices as preferences.
    - Added as -device parameters on command line ("PalmIIIe" and "PalmVx").
    - Added and mapped in appropriate skins.
    - Added new internal type IDs.
    - Corresponding tweaks to Binder application.
    - Made Palm IIIe return same device ID as Palm IIIx.
    - Made Palm Vx return same device ID as Palm V.

*    (Mac) Updated to GUSI 2.0.1. This has the following benefits:

    - Adds support for socket options.
    - Based on Open Transport instead of Mac TCP.
    - Fixes problem with menus sometimes getting disabled (the
      old GUSI would eat up "resume" events, so PowerPlant didn't
      know to re-enable the UI).

*    Fixed problem where Poser's heap walking routines were vulnerable
    to odd master pointers. [Mike Chen]

*    Gremlins wasn't generating and posting hard button events properly.
    [Steve Lemke]

*    (Unix) Fixed problem with trying to delete a skin with a stale
    pointer. [Ben Darnell]

*    Updated ROM Transfer.mcp to CodeWarrior for Palm OS R6.

*    Support for the gdb debugger has been built-in for quite some time
    (since 2.1d22 or so). Here, finally, are some instructions for
    interacting with it. Note that you no longer need to use gdbpanel
    or gdbplug if you were using those with earlier versions of Poser,
    Copilot, or xcopilot. [Ben Thomas]

    -    Build your app. Be sure to _both_ compile and link with -g
        (e.g., "gcc -g ..."). On the compile pass, this generates the
        necessary symbol information; on the link pass it forces the
        inclusion of a debug runtime code that results in a breakpoint
        being installed in PilotMain when the app starts.

    -    Start Poser.

    -    Load your app into Poser.

    -    Start gdb, loading your application's symbol table (e.g., "gdb
        myApp"). Note that the file to be loaded is the "myApp" file
        created by the GCC link pass, not "myApp.prc" created by
        buildprc.

    -    Type "target pilot localhost:2000" to gdb. It should respond
        with a message along the lines of "Remote debugging under PalmOS
        using localhost:2000. Waiting... (Press Ctrl-C to connect to
        halted machine)".

    -    Start your app on Poser.

    -    Wait for gdb to see the initial breakpoint and prompt you.

    -    Start debugging.

*    Tweaks to the scripting samples: [Flash Sheridan, Steve Haneman]

    - Wait() no long automatically resumes.
    - Added seperate Resume function
    - Since wait no longer resumes, added TapPenSync, and TapButtonSync
      which automatically call Wait and Resume at the beginning and end.
    - Fixed some warnings that showed up when running Perl -w.
    - Added MemPtrSetOwner, WinDisplayToWindowPt, WinWindowToDisplayPt.
    - In TapButton, call WinWindowToDisplayPt.
    - In EmRPC, added support for Coord type.

*    Internal change: encapsulated all device-specific information in an
    EmDevice class. This should make it easier to add new devices,
    since all the information is now in one place.

*    Stack overflow checking is back (cf 2.1d27)!  Whenever it looks
    like something is assigning a value to the stack pointer register,
    Poser marks the memory block that pointer is in as a stack. Then,
    when the stack pointer is decremented for any reason, it is compared
    to the low end of the memory chunk. If it falls below, Poser
    displays a fatal error dialog. If the stack pointer is just getting
    close without actually going over, Poser will display a warning.

*    Fixed another case of Gremlins going to sleep on the job. Gremlins
    normally posts new random events when the system calls EvtGetEvent
    and there's nothing to return, resulting in SysDoze being called.
    Most of the times, Gremlins will post an event and return from
    EvtGetEvent. In the cases it doesn't, it forces EvtGetEvent to
    return a nil event. However, if Gremlins was posting a Shakespeare
    quote to a form field, and the form field was full, and there was a
    non-empty selection, there was a 33% chance that neither an event
    would be posted nor a nil event returned. This resulted in the
    application having nothing to do (and so SysDoze was called) and
    Gremlins thinking its job was done for the moment and putting itself
    into a "do nothing" state. [Scott Maxwell]

*    (Windows) When reloading a .psf file, fixed a bug that prevent the
    Gremlin Control Window from being reshown if the session file had
    been saved while a Gremlin was running. [Scott Maxwell]

*    (Windows) Fixed a problem with relocating the ROM file that a .psf
    file reference when that ROM file wasn't where it was supposed to
    be, but it *was* in the same directory as the .psf file.

*    (Unix) Rolled in modifications for FreeBSD. [John Ludwig]

*    (Unix) Properly recognise X options such as -display and -geometry.
    [Eric House, John Marshall]

*    Reset the Dragonball timer registers to a consistant state before
    starting a Gremlin. This gives more consistant results. HOWEVER,
    any application that is dependent on the time in any fashion may
    still have consistancy problems when running Gremlins (that is,
    starting Gremlin #0 at 3:00pm may give different results than when
    it is started at 5:00pm). This is an issue that we'll try to
    address in a future release.

*    Switch from obsolete xxxChr names to new vchrXXX names in logging
    output. Also, if a winExit/EnterEvent is associated with a form,
    print the name of that form in the logging output if event logging
    is turned on.

*    (Mac, Windows) Added Tracer support. This is a facility for
    sending logging information to an external logging application. Palm
    applications use the following API for doing this:

    (All HostTraceOutput functions take a module identifier as their
    first parameter. This parameter allows filtering out traces
    according to their origin. Recognized modules are listed as error
    classes in SystemMgr.h. Application should specify appErrorClass.)

        void HostTraceInit(void);
            Call once at the beginning of your application to
            initialize the logging facility. Initiates a connection
            to the "tracing port" (specified in Poser's Tracing
            Options dialog box).

        void HostTraceClose(void);
            Call once at the exit of your application to release
            any host logging resources.

        void HostTraceOutputT(unsigned short, const char*, ...);
            Log a string of text to the tracing port via a
            "printf-like" facility.

        void HostTraceOutputTL(unsigned short, const char*, ...);
            Same as HostTraceOutputT with an additional line break.

        void HostTraceOutputVT(unsigned short, const char*, char* /*va_list*/);
            vprintf-like variant of HostTraceOutputT.

        void HostTraceOutputVTL(unsigned short, const char*, char* /*va_list*/);
            vprintf-like variant of HostTraceOutputTL.

        void HostTraceOutputB(unsigned short, const unsigned char*, unsigned long/*size_t*/);
            Send binary data to the tracing port.

    The functionality for sending the logging information to the
    external application is provided by a shared library (e.g.,
    "PalmTrace.dll" on Windows). The external application can be any
    application that supports the open packet convention. The one we
    provide is "Reporter". At the time of this writing, the shared
    library and external reporting application will be provided in a
    seperate archive. [Patrick Porlan, Regis Nicolas]

*    Internal changes: massive movement to make more source code cross
    platform. New classes include:

    EmTransport: abstract base class representing a "pipe" between two
    entities.

    EmTransportSerial: a pipe based on serial communications. Serial
    port emulation (the facility that allows Palm OS Serial Manager
    calls to pass thru to and from the host serial port) rewritten to
    use EmTransportSerial.

    EmHostTransportSerial: class with different implementations on
    different platforms in order to provide platform specific
    functionality.

    EmDlg: class to manage dialogs in a cross-platform fashion.

    EmROMTransfer: class that implements a rewritten ROM download
    facility, based on the new EmTransport and EmDlg classes. Because
    the new ROM transfer code is based on EmTransport and not
    EmTransportSerial, this means that it may be possible to transfer a
    ROM using something other than the serial port in the future.

*    With the help of the previous changes, the ROM Download facility is
    now available on Unix.

*    Internal change: rewrote JPEG decompression routines to better
    modularize the source of compressed data and the destination of
    decompressed data. This was done to better support JPEG images on
    disk in addition to those in embedded resources.

*    Added back the old generic case graphics. These graphics are now
    used when the user selects "Default" in the Skins menu and dialog.
    The nicer graphics are now identified with names like "Standard -
    English" and "Standard - Japanese". Eventually, those nicer
    graphics will be moved to external files, reducing the size of Poser
    drastically and reducing download times.

*    Internal change: synced up with the latest shared Palm OS files.
    This change also included switching over to the new standardized
    types (UInt32, Int16, etc.) from the old hodgepodge of types (DWord,
    Long, Int, Byte, etc.).

*    ROM Transfer now downloads the entire ROM instead of just the "Big
    ROM". It now is also a little more robust in light of some invalid
    fields in Visor ROMs.

*    (Windows) Added support for TRGpro devices. This support comes in
    the form of supporting TRGnet's FAT File System library. There is
    no TRGpro menu item in the device menu. Instead, you can
    effectively turn any Palm Computing device into the equivalent of a
    TRGnet device by installing their FFS library -- which is what
    manages access to the CompactFlash slot -- onto the device.
    (Actually, all it needs is a stub library that you can get from
    TRGnet.) Poser will redirect calls from that library to the host
    file system. The root of the emulated FFS directory hierarchy is
    specified by the FfsHome preference in Poser's preference file. It
    can be a path relative to Poser's directory (e.g., "foo") or an
    absolute path (e.g., "c:\temp\foo"). The directory will be created
    if needed. If no preference is specified, or the value is empty, or
    the specified directory could not be created, the root path defaults
    to Poser's directory. There is currently no UI for setting the root
    of the hierarchy; you have to edit the preference file by hand.
    [Mike Walter]

*    Added Visor support. Note that no USB support is provided in this
    release. This means that Poser will not download a ROM from a
    device in a USB-connected cradle, nor will USB operations on the
    emulated Visor be redirected to any host facility.

*    Fixed problem with a too-small name buffer in profiling routines.
    Removed absolute address from generated function names, as that
    annotation interferes with the CodeWarrior Profiler's ability to
    unmangle C++ names.    [Catherine White]

*    (Windows) Removed "Close" menu item from "bound" versions of Poser
    (functionality duplicates "Exit"), and added "Save Screen". [Scott
    Johnson]

*    (Unix) Added temporary hack to get Unix menus working (many would
    result in the message that the system was not responding). The
    solution implemented, however, is not optimal, and may lead to other
    problems. Expect this area to receive more attention in the near
    future. [Eric House]

*    Added sound emulation. Be sure to enable this in the preferences
    if you want it. [William F. Weiher III]

*    Fixed a couple of minor problems in the NetLib redirection code,
    one with accept() and another with the way select() timeouts were
    set up. [bill pitore]

*    Fixed a problem with determining the baud rate established by the
    UART, with the result that serial communications on 20MHz devices
    (like the Palm Vx) didn't work. [Alan Finke]


Changes for 3.0a3 (11/13/99)
----------------------------
*    Added Q&D PC checking. On JSR, BSR, RTS, and RTE, I check what
    we're attempting to set the PC to. Originally, I had a pretty beefy
    check that made sure the PC was in things like allocated, locked
    chunks. However, that's expensive for now, so I backed off to
    checking that the PC was (a) even, (b) in RAM or ROM, and (c) not in
    low-memory.

*    The new UAE 0.8.10 code includes a new way to emulate ROR and ROL
    instructions. Unfortunately, this new way runs afoul of a code
    generation bug in CWCPPC 2.3.1 (from CW 5.2) on the Mac. Reverted
    to the old way of emulating those instructions until Metrowerks
    comes out with a compiler patch. [Catherine White]

*    (Unix) Added tweaks for Debian systems. [Ben Darnell]

*    (Mac) Changed the way Poser determines whether or not to translate
    extended ASCII characters (those >= 0x80) into standard ASCII
    equivalents. The new method should be more consistant and robust,
    based on whether or not the user is running a Japanese Palm OS and a
    Mac OS with the Japanese Language Kit installed.

*    The device could actually go to sleep while running Gremlins!
    Sometimes, Gremlins will input a long series of key events.
    However, key events don't reset the auto-off counter. Normally, the
    Palm doesn't care, since key events are a side-effect of pen events,
    which *do* reset the auto-off counter. But Poser doesn't enter key
    events that way -- it inserts them directly with EvtEnqueueKey. Now
    Poser calls EvtResetAutoOffTimer when it enters key events. [Steve
    Lemke]

*    (Windows) Sped up Gremlin Hordes a little. There were some long
    delays while switching from one Gremlin to another due to trying to
    access the Gremlin Control Window from the wrong thread. There's
    still a mysterious 2 second delay while switching Gremlins, but at
    least one cause of the slow-down is fixed.

*    Fixed problem with Gremlins and Palm OS 3.5 systems. Gremlins
    calls WinGetWindowBounds to get the bounds of a window in which to
    generate pen taps. WinGetWindowBounds returns the bounds of the
    window returned by WinGetDrawWindow. On older systems, if the draw
    window were NULL, WinGetDrawWindow would return the display window.
    On 3.5, this has been changed in debug ROMs; WinGetDrawWindow can
    return NULL so as to force applications to make sure there is an
    established draw window. Gremlins ran afoul of that change, and so
    now calls WinSetDrawWindow(WinGetActiveWindow()) before calling
    WinGetWindowBounds(). [Roger Flores, Steve Lemke]

*    In the event posting/receiving logging code, support the new Palm
    OS 3.5 events.

*    In HostFPrintF, support + and space as flags, and * as a minimum
    width field character. [Patrick Porlan]

*    Added Symbol 1700/1740 skin as an option for Palm III's. This is
    *not* to say that Poser emulates symbol devices. It only means that
    you can make a Palm III look like a Symbol device. [Ron Kupke]

*    Reformatted Gremlin Hordes log file to make it a little easier to
    find the error messages. [Flash Sheridan, Catherine White]

*    Make a stronger effort to get the pen calibrated perfectly. The
    original attempt (cf, 2.1d11) would occassionally only get us very
    close, which could lead to inconsistant Gremlin runs.

*    Fixed a bug in the UAE 0.8.10 code where a corrupt status register
    could get pushed onto the stack during an exception/interrupt and
    the previous instruction that had affected the Z CPU flag was a BCHG
    instruction that flipped the highmost bit from off to on (gee, how
    could we have missed *that*). [Catherine White, Stuart Malone]


Changes for 3.0a2 (11/05/99)
-----------------------------
*    Updated to UAE 0.8.10. We were previously based on 0.8.3. The new
    version fixes a few minor bugs (I never noticed them), perhaps
    speeds things up a little, but most importantly cleans up those 2000
    "unreferenced label" warnings you get when compiling with gcc!  OK,
    so now there are some "unused variable" warnings, but there are a
    lot fewer of them...

*    (Windows) The sequence of events involving saving a session when
    creating or opening a new one was completely broken.

*    (Windows) Added support from Palm Europe for their tracing tool.
    This tracing tool is a seperate application that receives tracing
    information sent via HostControl functions. The documentation for
    these functions is not yet available. Neither is the external
    application. But they soon will be. I just wanted to at least
    mention them so that people don't ask me about them if they see the
    new functions in HostControl.h.

*    (Unix) Modified Perl script that creates ResStrings.cpp. It used
    to add the strings to a map object (so that they can be looked up by
    an ID number) by calling a function with about 2000 calls that
    looked like "map[id] = str;". However, gcc would choke on this with
    optimizations turned on. Now, that map addition is performed in a
    seperate function, and that function is called 2000 times instead.
    [Ben Darnell, Jon Fo]


Changes for 3.0a1 (11/04/99)
-----------------------------
*    Interim internal release dates:
        2.1d29.1    -    08/26/99
        3.0a1e2        -    09/14/99
        3.0a1e3        -    09/15/99
        3.0a1e4        -    09/16/99
        3.0a1e5        -    09/21/99
        3.0a1e6        -    09/24/99
        3.0a1e7        -    09/30/99
        3.0a1e8        -    10/08/99
        3.0a1e9        -    10/14/99
        3.0a1e10    -    10/19/99
        3.0a1e11    -    10/27/99

*    Rolled in Llamagraphics's profiling changes [Catherine White]:

    The motivation for these changes was to fix the crashing problem we
    reported earlier. Here's what we changed:

    -    We rewrote RecursiveAddressToStrings without recursion, and
        named the new function LinearAddressToStrings. The old function
        was recursing over both kids and siblings, which was blowing the
        stack on the Macintosh. Since it doesn't matter what order the
        addresses are converted in, a linear loop is faster and simpler.

    -    Called ProfilerCleanup at the end of ProfilerDump. Since
        calls[i].address is modified during the dump, the calls array
        can't be reused after the dump since it no longer contains valid
        addresses. It seems safer to dispose of it so that it isn't used
        accidentally.

    -    Replaced the two Debugger() calls in ProfilerInit with calls to
        DisposeMemory(). We were encountering these Debugger calls
        while testing our own code, probably because ProfilerInit was
        being called twice somehow. Calling ProfilerInit twice should
        now be harmless, and should no longer leak memory.

    -    Changed the handling of function names so that they are no
        longer limited to 31 characters. The new limit is 255
        characters. This means that most mangled C++ names can now be
        parsed properly by the MW Profiler, which makes interpreting the
        profiles easier.

        In order to accomplish this, we needed to add extra arguments to
        two routines in Miscellaneous.cpp for the buffer capacity. This
        capacity defaults to 32, which means that callers who don't pass
        in the extra argument will get the same behavior as before.

        We also added code in the profiler so that the stringTable could
        grow dynamically. This means that it isn't so important to
        allocate enough memory for the stringTable initially. However,
        the initial size is currently the same as the old size was, so
        reallocation shouldn't occur very often.

*    (Mac) Updated to CodeWarrior Pro 5.

*    Overhauled handling of preferences/properties/settings/options...
    These settings were previously stored as resource-based binary data
    on the Mac, System Registry entries on Window, and an RC text file
    on Unix. Now the information is saved out as a text file on all
    platforms. And -- with a common source code base -- all information
    is saved out consistantly (some platforms previously skipped saving
    some information). Finally, with all information in an accessible
    text file format, it should be easier for users to tweak or fix
    settings (or even specify settings for which there is no UI).

    Note that NO attempt has been made to migrate forward settings from
    the previous formats.

*    In line with that, if the InterceptSysFatalAlert is true (default),
    Poser will intercept calls to SysFatalAlert and display the message
    in its own dialog box. If this setting is false, Poser will let the
    standard Palm OS function display the message. [Ken Krugler, Scott
    Johnson]

*    If a SysFatalAlert dialog is displayed while a Gremlin is running,
    the Gremlin will no longer be halted if the user clicks on Continue
    (it used to be that the Gremlin was *always* halted). [Roger Flores]

*    (Windows) Ask user if they'd like to have a shortcut to Poser added
    to the Start Menu. [Phil Shoemaker]

*    Better validation of .prc files. [Daniel McCarthy]

*    (Windows) Cancel button in file download progress dialog now works.

*    (Windows) Fixed byteswapping bug in PSF writing code. Poser should
    now have a better chance of reading Mac-created PSF files on Windows
    and vice-versa.

*    Fixed memory leak bug that would occur when closing a session. The
    buffers holding the ROM image and auxilliary information weren't
    being deleted. [bullshark (a real person with a real email
    address)].

*    (Windows) Added the Palm OS Emulator Binding tool, an NT-only
    wizard which can bind a Windows Poser executable with (1) a ROM file
    and a device configuration, or (2) a ROM file and a session file.
    This will create a self-contained executable suitable for "kiosk"
    run.

    Poser detects when it is bound, and modifies its behavior in the
    following manner:

    (1)    ROM + device configuration. Poser always starts by booting
        into a new session with the selected device configuration and
        ROM. If the user closes the session and creates a new one, he or
        she will not be prompted for a device configuration.

    (2)    ROM + PSF file. Poser always starts right into the state
        saved in the PSF file. The user interface is abbreviated to
        include only pertinent functionality (no debugging or logging,
        for example) that is appropriate to a "kiosk" run. IF the user
        closes the session, he or she only has the option to restart it
        in the state saved in the PSF file. Posers bound this way share
        a separate set of preferences then unbound Posers.

*    Made the socket port that an external debugger can connect to
    configurable. The old hard-coded port of 2000 is still support for
    backward compatibility, but clients should move away from that.
    Instead, they should use the port specified in the
    "DebuggerSocketPort" preference. (There is currently no UI for
    changing this preference. But it can be changed by hand by editting
    the Poser preferences file, which is in text format.)

*    Added the following functions to HostControl.h:

        HostErr        HostProfileDetailFn            (void* addr, HostBool logDetails)

            Profiles the function containing the given address. If
            logDetails is true, profiling is performed at a machine-
            language instruction level (each opcode is treated as
            its own function).

        HostErr        HostGremlinNew                (const HostGremlinInfo*);

            Start a new Gremlin using the given specifications.

        HostBool    HostGetPreference            (const char*, char*);
        void        HostSetPreference            (const char*, const char*);

            Get or set the specified preference. Preferences are
            identified by name, the set of which can be seen in the Palm
            OS Emulator Preferences file for your platform ("Palm OS
            Emulator Preferences" in the Preferences folder on the Mac,
            "Palm OS Emulator Preferences.ini" in the Windows System
            directory on Windows, and ".poserrc" in your home directory
            on Unix). The values are also specified as strings,
            examples of which can be seen in the preferences file.

        HostErr        HostSessionCreate            (const char* device, long ramSize, const char* romPath);

            Create a new session based on the given parameters.
            Fails if a session is already running.

            * NOTE: Not yet implemented.

        HostErr        HostSessionOpen                (const char* psfFileName);

            Open the given session. Fails if a session is already
            running.

            * NOTE: Not yet implemented.

        HostErr        HostSessionClose            (const char* saveFileName);

            Close the current session, saving it to the given file.
            The session is not saved if no file name is specified.
            This function fails if no session is running.

        HostErr        HostSessionQuit                (void);

            Ask Poser to quit. Fails if a session is already
            running.

        HostErr        HostSignalSend                (HostSignal signalNumber)

            Send a signal (identified by an integer) to any scripts
            with HostSignalWait calls pending. If there are such
            pending calls, Poser then halts waiting to be restarted
            with HostSignalResume. If no one was waiting for any
            signals, then Poser doesn't halt.

        HostErr        HostSignalWait                (long timeout, HostSignal* signalNumber)

            Wait for a signal from Poser. Signals can be predefined
            (hostSignalIdle, hostSignalQuit) or user-defined.
            HostSignalWait returns the signalled value.

            hostSignalIdle is issued by Poser when it detects that it's
            going into an idle state.

            hostSignalQuit is issued by Poser when it's about to quit.

        HostErr        HostSignalResume            (void)

            Restarts Poser after it has issued a signal. By waiting
            to be restarted, Poser allows external scripts to perform
            any other operations first.


    While the following functions are in the HostControl API,
    they exist for external RPC clients to call and will return
    errors if called by Palm OS applications:

        HostSessionCreate
        HostSessionOpen
        HostSessionClose
        HostSessionQuit
        HostSignalWait
        HostSignalResume

*    Opened a socket (on the port specified in "RPCSocketPort"
    preference) for external applications to make RPC calls. By sending
    RPC packets, external applications can invoke any function in the
    Palm OS dispatch table. This set of functions includes the
    HostControl functions, which can be called even when no session is
    running (which normally would mean that there isn't really a
    dispatch table containing functions that can be called).

    Packets sent to the RPC sub-system use the same format as packets
    sent to the debugger sub-system, that is, the Serial Link Protocol
    packet format (also used by HotSync). This format is described in
    Debugging.html.

    When sending command packets to the RPC sub-system, you need to set
    the "dest" field of the packet header to:

        #define slkSocketRPC            (slkSocketFirstDynamic + 10)

    as opposed to slkSocketDebugger or slkSocketConsole, as you would
    when talking with the debugger sub-system.

    The RPC sub-system can accept one of 4 kinds of command pacets:
    ReadMem, WriteMem, RPC, and RPC2. The first three packets are the
    same as those sent to the debugger sub-system and are defined in
    Debugging.html. RPC2 is an extension of the RPC packet in order to
    support a wider range of facilities.

    RPC2 packets are defined as follows:

        #define sysPktRPC2Cmd    0x20
        #define sysPktRPC2Rsp    0xA0

        struct SysPktRPCParamInfo
        {
            UInt8    byRef;        // true if param is by reference
            UInt8    size;        // # of Bytes of paramData    (must be even)
            UInt16    data[1];    // variable length array of paramData
        };

        struct SysPktRPC2Type
        {
            _sysPktBodyCommon;            // Common Body header
            UInt16    trapWord;            // which trap to execute
            UInt32    resultD0;            // result from D0 placed here
            UInt32    resultA0;            // result from A0 placed here
            UInt16    resultException;    // If an exception occured, it's ID is here
            UInt8    DRegMask;            // Bitmasks indicating what registers need...
            UInt8    ARegMask;            // ...to be set on this call.
            UInt32    Regs[1];            // Variable-length array holding register...
                                        // ...values to be set on this call.
            UInt16    numParams;            // how many parameters follow

            // Following is a variable length array of SlkRPCParamInfo's
            SysPktRPCParamType    param[1];
        };

    Most of this is the same as the RPC packet. It differs with the
    addition of the following fields:

    resultException: if the function call failed because a hardware
    exception occured, the exception ID will be stored here. Otherwise,
    this field contains zero.

    DRegMask: a bitmask indicating what D registers need to be set in
    order to make this call. In particular, some functions require that
    a "selector" value be placed in register D2. In this case, bit 2 of
    DRegMask (the 3rd bit from the right) would be set.

    ARegMask: same comments.

    Regs[1]: a variable length array containing the values to be placed
    in the registers to be set. Only the registers that are being
    changed need to be supplied. Most of the time, DRegMask and ARegMask
    will be zero and Regs would not appear in the packet. In the example
    where D2 needs to be set, DRegMask would be 0x04, ARegMask would be
    0x00, and ARegMask would be followed by a single 32-bit value to be
    placed in D2. If more than one register needs to be set, then the
    register values should appear in the following order: D0, D1, ...,
    D6, D7, A0, A1, ..., A6, A7. Again, only values for the registers
    specified in DRegMask and ARegMask need to be provided.

*    Fixed crasher that could occur if "Received Events" logging was
    turned on and one of three KeyDown events was popped off the stack.
    [Ken Krugler, Jeff Yasuda]

*    (Unix) Added serial port emulation, as well as better support for
    other OS and compiler flavors. [Jerry Kirk, Michael Kedl]

*    (Unix) Added support for cursor, page up/down keys. [Michael Kedl]

*    (Unix) Fixed massive memory leak when sessions ended. [Michael Kedl]

*    More evolution of skin support. Skins are now set on a device-by-
    device basis from a new "Skins..." dialog box. Other aspects of the
    device's appearance are also set from this dialog box, such as the
    scaling factor and the LCD background color.

*    Look at the chip selects when determining what range of RAM should
    be protected or not instead of just assuming 128K. [Steve Lemke]

*    Fixed problems with reloading .psf files what were saved when a
    Gremlin was running.


*    Setting "DebuggerSocketPort" or "RPCSocketPort" preferences to zero
    will disable their respective facilities. This should be handy in
    cases where merely opening a TCP socket fires up any sort of
    unwanted Remote Access facilites. [Ken Krugler]

*    (Mac) Fixed bug in new preferences saving mechanism where the
    preferences would not get saved if the file didn't already exist.
    [Mike Puckett]

*    Allow LCD window to be placed in negative coordinates. [Mike Puckett]

*    Addition of Gremlin Hordes. A Horde is basically a range of
    Gremlins to run. You specify the range: the first Gremlin and the
    last Gremlin in the Horde. Poser will start with the first Gremlin,
    and generate that Gremlins event stream until one of four things
    happens:

    - An error occurs (hardware exception, illegal memory access, etc.),
    - The maximum number of events have been generated,
    - A "switch" event occurs. A switch event is when a specified number
      of the maximum number of events have been generated.

    When any of these conditions occur, the current Gremlin is either
    halted or suspended and the next Gremlin in the Horde is selected.
    When Poser has cycled through all the Gremlins in the range, it
    cycles back and restarts any suspended Gremlins one at a time. When
    all Gremlins have been halted (either because they've encountered an
    error or they've reached the maximum number of events), the Hordes
    session is completed.

*    (Windows) Added the following command-line options [Brian Estes]:

    -skin <name>:
            Synonyn for -silkscreen.

    -ram_size <size in K>:
            Synonym for -ram.

    -load_apps <file name list>:
            A comma-seperated list of files (.prc's, etc.) to load after
            startup.

    -run_app <app name>:
            Application to run after startup. Note that this is the
            name of the *application*, not the file.

    -quit_on_exit:
            If -run_app was specified, quit when that application exits.

    -log_save_dir <path>:
            Location for standard log file. Default = Poser's directory.

    -horde <num>:
            Gremlin number to run after session is created or loaded.
            Equivalent to setting -horde_first and -horde_last to the
            same number.

    -horde_first <num>:
            First Gremlin in a Gremlin Horde to run.

    -horde_last <num>:
            Last Gremlin in a Gremlin Horde to run.

    -horde_apps <app name list>:
            A comma-seperated list of applications Gremlins is allowed
            to switch to. Default == entire device (no restrictions).

    -horde_save_dir <path>:
            Location for directory that keeps saved session files and
            logs. Default = Poser's directory.

    -horde_save_freq <num>:
            Gremlin snapshot save frequency. Default = don't save
            snapshots.

    -horde_depth_max <num>:
            Total number of Gremlin events to generate for each Gremlin.
            Default == no upper limit.

    -horde_depth_switch <num>:
            Number of Gremlin events to generate before switching to
            another Gremlin in a Horde. Default == the same number
            specified for -horde_depth_max.

*    (Windows) Added support for Alt-F4; made it act like the Exit menu
    item. [Ryan Robertson]

*    Fixed a problem with profiling calls to Palm OS function made via
    the SYSTRAP_FASTER macro. The function entry would be recorded once
    because it was a system call, and then again because it was made via
    a JSR. Because the subsequent RTS would pop only one of those off of
    the profiler's stack, the stack would get out of sync and overflow.
    [Jameson Quinn]

*    (Windows, Unix) Now stop the CPU thread on almost any UI action.
    Previously, the CPU thread would be stopped only in targeted
    situations. Now, it's stopped whenever the user executed a menu
    command. This change was made when I found out that the profiling
    menu items could mess up the CPU thread if it weren't stopped first.

*    Updated ROM Transfer project to CodeWarrior for Palm OS 5.2 (that
    is, CodeWarrior for Palm OS Release 5 with Metrowerks' two updates
    applied).

*    Fixed the way time spent in the trap dispatcher was recorded when
    profiling; entries for the same function weren't necessarily being
    collapsed into a single entry. [Bob Ebert]

*    (Unix) Added JPEG case graphics.

*    (Windows) Added taskbar menu with Minimize/Move/Close menu items.
    Note that there may still be problems with minimizing/maximizing
    Poser. For instance, Windows NT does not minimize windows without
    caption bars when you select "Minimize All Windows". Windows 98
    will hide Poser when selection "Show Desktop", but will restore
    Poser when restoring another application window. These problems
    appear to be shortcomings of Windows, but if anyone knows how to
    accomodate these foibles, please let me know. [Scott Johnson]

*    More accurate emulation of ROM space. Now the chip-selects are
    used in conjuction with the ROM image file size. [Steve Lemke,
    Jesse Donaldson]

    In general:

    The size of the ROM image is always rounded up to the next power of
    two.

    On Dragonball devices, I assume that the chip-selects are always set
    to specify a 2 Meg ROM (this appropriate for all ROMs so far). On
    DragonballEZ devices, I look at csASelect, bits 1-3 (BTW, the reset
    value for this register is 0x00E0, so there don't seem to be any
    boot-up issues).

    If the rounded ROM image size is larger than the size specified by
    the chip-selects, then it gets truncated; the CPU won't be able to
    access the far end of it.

    If the rounded ROM image size is the same as the size specified by
    the chip-selects, then everything is fine. No truncating, no
    mirroring.

    If the rounded ROM image size is smaller than the size specified by
    the chip-selects, then the ROM image is mirrored for as many times
    as is necessary to cover the range.

*    (Mac) Added Power Manager code to keep PowerBooks awake when Poser
    is doing something and when the PowerBook is plugged in. [Mike
    Puckett]

*    Ensure that windows (LCD, Gremlin Control) are onscreen when
    they're created. Right now, this means that they are brought
    onscreen if any part of them is offscreen. If this is too strict,
    we can revisit the algorithm later.

*    The code to validate the objects on a form has been moved from
    Gremlins-specific code into a patch on FrmDrawFrom. Now people will
    get bugged about invalid objects even when not running Gremlins. :-)

    By the way, some people have reported that they get told about
    invalid or offscreen form objects, and that they think these reports
    are in error, as the objects reported have ludicrous positions and
    sizes and have object IDs that don't exist in the resource
    description file. If you get such a message, *strongly* suspect
    that your object list has been trashed, possibly due to the changing
    of a label or title. Eventually, Poser will be able to detect when
    form object lists get trashed at the moment it occurs, and not later
    on down the stream (such as when the form gets drawn).

*    When dragging files onto the LCD screen, the rules are: you can
    drag any number and assortment of .prc, .pdb, and .pqa files; you
    can drag only one .rom file, you can drag only one .psf file; you
    can't mix files from any of the three just-mentioned categories
    together. If you broke any of these rules, you'd get an error
    message. However, this error message was broken in different way on
    Mac and Windows (the Unix version doesn't support D&D yet -- any
    takers?). This was fixed by making the majority of the relevent
    code cross-platform, and then fixing *that*. [Roger Flores, Mike
    Turcotte]

*    (Mac) Fixed problem with Japanese text showing up as "Jinglish" on
    systems with the Japanese Language Kit installed. This was always
    supposed to be a feature in the New Gremlin dialog (so that you
    could see the application names correctly), but appears to have been
    broken for some time. That should be fixed, and the same facility
    has now been extended to the dialog containing error messages
    generated by the Palm OS. [Ken Krugler]

*    Fixed bug in NetLibGetHostByAddr where the address was assumed to be
    a NetSocketAddrType instead of anything that could be "len" bytes
    long. [Bill Pittore]

*    Turned on checking of writes to ROM. This was off for a reason
    (there's a comment next to the relevent part saying "HACK: We really
    want it to be true!"), but I don't remember what it is. I was able
    to run Gremlins on some apps for quite a while without error, so
    hopefully we can get away with turning this back on. [Scott Johnson]

*    Take another stab at fixing the problem with FntDefineFont, as first
    described in the release notes for 2.1d20. FntDefineFont can read
    off the end of a source buffer, causing Poser to report an invalid
    read access to memory manager data structures. This bug -- fixed in
    Palm OS 3.2 -- doesn't need to be alerted by Poser. [Scott Johnson]

*    If an application executes DbgBreak and no debugger is attached,
    display a friendly message in a dialog box that includes Continue
    and Reset buttons. [Ken Krugler]

*    Added Red Dutta's "Export Database" menu item. This command takes
    the place of the "posedbexport" Palm OS application, allowing you
    to export databases and applications from the emulator to your
    desktop computer as a .pdb or .prc file. [Red Dutta]

*    Patch SysReset so that Poser no longer complains about normally
    proscribed actions (like mucking with low-memory) during a reboot.

*    Bring us in-sync with the Simulator with regards to entering
    control-chars that map to Palm OS command-chars. The list is:

        control-A    menuChr
        control-B    lowBatteryChr
        control-C    commandChr
        control-D    confirmChr
        control-E    launchChr
        control-F    keyboardChr
        control-M    linefeedChr
        control-N    nextFieldChr
        control-S    autoOffChr
        control-T    hardContrastChr
        control-U    backlightChr

    B, S, T, and U are new in this release.


Changes for 2.1d29 (7/8/99)
---------------------------
*    Interim internal release dates:
        2.1d28.1    -    6/4/99
        2.1d28.2    -    6/4/99
        2.1d28.3    -    6/11/99
        2.1d28.4    -    6/11/99
        2.1d28.5    -    6/11/99
        2.1d28.6    -    6/15/99
        2.1d28.7    -    6/16/99
        2.1d28.8    -    6/23/99
        2.1d28.9    -    6/25/99
        2.1d28.10    -    6/30/99

*    For temporary backward compatibility, "bare" file names on the
    command lineare treated as paths to .psf files. [Andreas Linke, Mike
    McCollister]

*    (Mac) Fixed problem with receiving Error #18 when trying to download
    ROMs from actual devices over the serial port. [Mark Ordal]

*    In debugger/sockets communications, allow for recv() returning
    zero, indicating a closed connection.

*    Fixed multi-threading bugs. [Andrew Ball, Daniel McCarty, Scott
    Johnson, Keith Wolcott, Art Dahm, Mark Lussier, Alex Robinson,
    Patrick Porlan]

*    (Unix) Fixed some makefile problems. [many]

*    (Mac, Windows): new case graphics. Note that the edges of the new
    graphics are a little rough; we'll be working on that for the next
    release. Also, the Mac version doesn't set up a custom palette in
    8-bit mode, yet, so for best results, you might want to put your
    monitor in 16- or 24-bit modes. Finally, the Unix version will get
    the new graphics in the next release.

*    Print some logging information using signed instead of unsigned
    value. [Oliver King-Smith]

*    Don't go into infinite recursion when the emulated PC is set to an
    odd memory address. [John Marshall]

*    Fixed problem in the "NetLib redirected" version of
    NetLibSocketAccept where the address and address length were not
    returned. [Bill Pitore]

*    (Windows) As outlined in the notes for 2.1d28, Poser will look for
    the last-saved .psf file when starting up. It used to save the
    reference to this file as a full pathname. Now, if the attempt
    to locate the file with the full pathname fails, Poser will look
    for a .psf file with the same name in the same directory as
    Poser itself. This approach is similar to the way in which
    Poser looks for .ROM files that are no longer at their last known
    location. [Michael Hado]

*    Worked around VC++ 6.0 bug where tabbed dialogs in applications
    built by it won't show up on systems without IE 4.x installed.
    [Michael S. Davis]

*    (Unix) Fixed a minor bug in skin loading code which causes Emulator
    to crash when it tries to load a skin file from current directory.
    [Vlad]

*    Set the command key bit when Gremlins posts an up or down arrow.


Changes for 2.1d28 (5/21/99)
----------------------------
*    Interim internal release dates:
        2.1d27.1    -    5/17/99

*    (Windows) Added command line options:

    -psf <filename>: emulator loads the specified .psf file on startup.

    -rom <filename>: specifies the ROM image file to use.
    -ram <size>: specifies the amount of RAM to emulate (in K). Valid sizes
            are 128, 256, 512, 1024, 2048, 4096, and 8192.
    -device <type>: specifies the device to emulate. Valid types are pilot,
            palmpilot, palmiii, palmiiix, palmv, and palmvii.
    -silkscreen <type>: specifies the silkscreen to use. Valid types are
            english and japanese.

    Case is NOT significant.

    Examples:

        Emulator -psf C:\Data\Session.psf

        Emulator -ROM C:\ROMs\3.0\debug.rom -RAM 1024 -Device PalmIII
            -Silkscreen English

    Startup rules are now as follows:

    1 If the Caps Lock key is toggled in the ON position, always bring
      up the New/Open/... dialog.
    2 Scan the command line for startup parameters. If an error occurs
      trying to scan the command line, the error is reported and the user
      is presented with the New/Open/... dialog.
    3 Use the .psf file if one is specified. If an error occurs trying
      to load the file, the error is reported and the user is presented
      with the New/Open/... dialog.
    4 If any of -rom, -ram, -device, or -silkscreen are specified, try
      to start a new session based on those values. If all are specified,
      the new session is automatically created. If any of those four
      values are missing, the "New Configuration" dialog is displayed.
      If the user cancels the dialog, or if there is an error creating
      the new session, any error is reported and the user is presented
      with the New/Open/... dialog.
    5 If no command line options are specified, try re-opening the last
      saved .psf file (this step is skipped if the user last created a
      new session, but did NOT save that session to a file). If an error
      occurs trying to load the file, the error is reported and the user
      is presented with the New/Open/... dialog.
    6 Try creating a new session based on the settings the user last
      specified when creating a session. If there is an error creating
      the new session, the error is reported and the user is presented
      with the New/Open/... dialog.
    7 Finally, if all else fails, present the user with the New/Open/...
      dialog.

    Steps 1, 5, 6, and 7 describe the old startup rules. Steps 2, 3,
    and 4 are new. [Steve Haneman]

*    See that problem where I said that saving a session file could take
    up to 15 seconds or even longer?  And see where I said I didn't know
    why?  Well, it's because I'm an idiot. Under certain circumstances,
    Poser could get into a state where it would save the same session
    file several times in a row. Normally it would take a 1/2 second to
    save a file. But if it decided to save the same file 30 times in a
    row, that would take 15 seconds...

*    Fixed some more synchronization problems on SMP machines.

*    Force the Log####.txt files to be stored in the Poser directory
    instead of whatever directory happens to be the current one.


Changes for 2.1d27 (5/7/99)
---------------------------
*    Interim internal release dates:
        2.1d26.1    -    ???
        2.1d26.2    -    ???
        2.1d26.3    -    ???
        2.1d26.4    -    3/25/99
        2.1d26.5    -    4/1/99
        2.1d26.6    -    ???
        2.1d26.7    -    ???
        2.1d26.8    -    4/27/99
        2.1d26.9    -    4/29/99
        2.1d26.10    -    5/5/99

*    Added special messages for accessing NULL and for trying to access
    the A5 register (in order to access global variables or make inter-
    segment jumps) when A5 is not set up for the application trying to
    use it. [Roger Flores, Ken Krugler]

*    Better tracking of the current application. Now, if errors occur
    during an "action code" sequence, the application executing is
    reported, not the "main", "real" application that has the stack, A5,
    etc. [Roger Flores]

*    Fixed problem with installed Palm OS files and .psf files not
    getting added to the MRU lists if the list was empty. [Scott
    Johnson]

*    Fixed problem with MRU lists getting too long. [John Kinast]

*    In our SysUIAppSwitch patch, release any leftover command parameter
    blocks (if any). [Catherine White]

*    Fixed problem with the Escape key used to wake up a sleeping device
    also showing up in the event queue. [Daniel McCarty]

*    (Mac) Added support for Escape, F1-F4, page up/down.

*    Added special checks and error messages for SANE Math calling
    sequences. [Steve Lemke]

*    Rolled in Adam's fixes to conditional breakpoints. [Adam Dingle]

*    Fixed problems with NetLibSocketAddr. [Adam Dingle]

*    Converted to VC++ 6.0.

*    Fixed problem with getting names of library routines in the
    profiling system. [Adam Dingle]

*    Made Profile/Initialize implicit. [Adam Dingle]

*    Added explicit parent node information to profiling output text
    file. [Adam Dingle]

*    Fixed problem with recursive routines recursing too deep in the
    profile output functions.

*    The workaround for the old FindSaveFindStr bug didn't always work;
    it would sometimes still let the bug emerge. Made a small tweak to
    keep it suppressed.

*    Turned off stack overflow checking facility. It would occasionally
    cause the emulator to walk the dynamic heap when the heap was in an
    inconsistant state, resulting in spurious "...regular checkup..."
    error messages. [The IBM guys: Paul Silagi, Chunk Bazil, Mike Nagy]

*    Added support for the contrast button on Palm Vs. The button is on
    the right-hand side of the display graphic, opposite from the power
    button. You can't see it, but clicking on it brings up the contrast
    dialog. The dialog itself has no visual effect yet. [Ron Flax]

*    Breakpoints are now preserved across reboots. [Steve Lemke]

*    Totally rewrote the way (a) tailpatches and (b) soft breakpoints
    are handled. Previously, Poser would get control by handling
    special opcodes written into emulated memory: TRAP $D for
    tailpatches and TRAP $0 for soft breakpoints. However, there were
    problems with this approach. For one thing, what would happen if we
    needed to write a TRAP $D and TRAP $0 to the same memory location.
    You couldn't do it. There was no way, for example, to use
    PalmDebugger to set a breakpoint just after a call to a system
    function that Poser tailpatched. Also, Adam Dingle pointed out a
    problem with the actual routines involved: on very rare occasions,
    if an interrupt occured just after one of those TRAP $Ds or TRAP $0s
    were executed, then the next opcode executed would be the one that
    originally resided at that location, an NOT the first opcode in the
    interrupt routine. Anyway, this whole mess has been replaced with a
    new mechanism that doesn't alter the emulated RAM or ROM. [Steve
    Lemke, Adam Dingle and the IBM guys: Paul Silagi, Chunk Bazil, Mike
    Nagy]

*    Sped up performance by 20%. But since the previous change slowed
    things down by 13%, that's not as great as it first sounds.

*    Poser now ignores soft breakpoints (those installed by
    PalmDebugger) it encounters when it calls into the ROM for it own,
    nefarious, "between the cycles" purposes. [Steve Lemke]

*    Fixed problems trying to re-establish a sockets connection after
    it's been broken when the external debugger quit. [Ron Marianetti]

*    When asked for a ROM, .psf, .prc, etc., file, set the initial
    directory to the last directory used for that file type.

*    (Windows) Allow multiple .prc file selection in the Open File dialog.

*    ROMs with invalid header checksums now generate a non-fatal warning
    instead of a fatal error. [Paul Dugas]

*    Allow for zero-length objects appearing at the end of .pdb files.
    [Christopher Hunt]

*    Massive spring cleaning of sources. Now all major sub-systems have
    a consistant interface for creating new sessions, resetting
    sessions, saving sessions, loading sessions, and disposing of
    sessions. Also, a lot of naming and scoping inconsistancies were
    taken care of. Also, namespaces were forsaken in favor of classes
    with static member functions; the CodeWarrior and VC++ browsers
    didn't handle namespaces too well.

*    In line with the above, .psf files are now compressed, taking as
    little as 15K for one that was normally 1 Meg. Note: the time it
    takes to compress an image can vary wildly, sometimes being nearly
    instantaneous, and sometimes taking upwards of 15 seconds. I'm not
    sure why the difference occurs...

*    Also in line with the source code spring cleaning, the entire
    emulator state is now saved to the session files. When loaded from
    the session files, the entire emulator state is restored. This means
    that you are returned right back to where you were when you created
    the session file; you do not have to go through the process of
    rebooting the "device".

*    Only heed the "Continue Past Warnings" option if a Gremlin is
    actually running. [Scott Johnson]

*    Added support for the "dead battery" pin on Port D on EZ devices.
    This should cut down the number of low-battery warnings on those
    devices. [Jesse Donaldson]

*    Don't insert keyboard or mouse events if stopped in the debugger or
    if the main emulator window is not active.

*    Added support for 4-bit LCD mode on EZ's. Found and fixed a long-
    standing bug where the emulated copy of the LCD display might not
    get updated if the lcdPageWidth register was updated.

*    If the emulated serial port was open and a reset occured, close the
    host serial port.

*    On the Mac, session files contain references to their associated
    ROM files via aliases. On Windows, the references are stored as
    full pathnames. This is old news. What's new is that if either of
    these mechanisms fails to find the ROM file, Poser now looks first
    in the same directory as the session file for a ROM file with the
    same name, and if that fails looks in Poser's own directory for a
    ROM file with the same name.

*    When reporting a corrupted heap, display also the address of the
    chunk header. [Adam Dingle]

*    Implemented Part I of Gremlin Hordes: You can now periodically have
    the complete emulator state saved while a Gremlin is running. This
    state can be reloaded, and the Gremlin restarted.

    This feature is controlled via a new setting in the New Gremlin
    dialog. There is now an editable item called Snapshot Frequency.
    The number entered here is the number of events that should occur
    between the time snapshots are taken. The default value is 10,000
    events. By setting this value to zero, you turn off the snapshot
    feature.

    Snapshots are saved in a single directory in the Poser directory.
    The name of the directory is "GremlinStates_####", where #### is a
    value ensuring the directory name's uniqueness.

    Snapshots are saved in this directory with the name Event##########,
    where ########## is the event number at which the snapshot was
    taken. Snapshots are merely session files, so they can be reloaded
    at any time, just like regular session files.

    Since snapshots are taken while a Gremlin is running, the Gremlin is
    automatically turned off when the snapshot is reloaded. However,
    Poser displays the Gremlin Control window when the snapshot is
    loaded, allowing you to resume the Gremlin if you want. [Roger
    Flores]

*    Fixed a bug in the Gremlins facility that validates forms and form
    objects before trying to manipulate any of them. A call to
    FrmGetObjectType would occassionally return garbage, leading the
    rest of the function astray. [Scott Maxwell]

*    Added Unix release. [Ben Williamson, David Creemer]

*    Fixed serial port not working on EZ devices (a pin moved from port
    F to port D during hardware development, and I missed it). [Florent
    Pillet]

*    Don't validate form objects and their sizes on Palm OS 1.0 devices.
    The built-in applications don't appear to follow the rules. [Andrew
    Ball]

*    Rolled in Adam's conditional breakpoint enhancements. [Adam Dingle]

    "A register reference in a conditional breakpoint expression can
     use either a direct (i.e. "d5") or indirect (e.g. "12(a6)")
     addressing mode. Furthermore, a register expression can be
     suffixed with either ".b", ".w" or ".l" to indicate that the
     expression to be compared is either a byte, a word or a long. An
     an example, "12(a6).w == 1000" compares the two bytes at memory
     location (a6 + 12) to the value 1000. "d4.b == 100" compares the
     low byte of the d4 register to the value 100.

    "I have found this enhancement to be extremely useful for breaking
     into the debugger when a local variable (typically referenced off
     the a6 register in code built by CodeWarrior) has a certain value."

*    (Windows) Fixed problem with running on dual-processor boxes. [TBD]

*    There is now a "white paper" on Poser usage at the following URL:

        <http://www.palm.com/devzone/docs/pose.htm>


Changes for 2.1d26 (2/26/99)
----------------------------
*    On the Mac, we were saving the device type to the .psf file twice.
    On Windows, we weren't saving the country type at all. [Waddah
    Kudaimi]

*    System call logging now also gets the names of library functions,
    and attempts to make a better stab at "dispatch" system functions
    (where a single dispatch code sub-dispatches to a suite of other
    functions, as with the Floating Point Manager). [Ken Krugler]

*    Sped up debugger communications via sockets by 20x. [Ron Marianetti,
    Mark Corry]

*    (Windows) Added conditional breakpoints. This is bolted on for now,
    but should eventually be merged more seamlessly. Changes provided
    by Adam Dingle of AvantGo. From Adam's release notes:

    New debugging features

    I've posted several messages recently to the palm-dev-forum mailing
    list in which I've pointed out several limitations of the debugger
    in CodeWarrior for PalmOS: it doesn't support data breakpoints
    ("watchpoints") and its conditional breakpoint implementation is
    very slow. To work around these limitations, I've added both data
    breakpoints and fast conditional breakpoints to the 2.0b3 emulator.

    A data breakpoint allows you to monitor a range of memory addresses
    for writes. When any code tries to write to the addresses you've
    monitored, execution will immediately break into the debugger. This
    can be extremely useful when you know that a certain data structure
    is being corrupted somewhere in your program, but you don't know
    where it is happening.

    To set a data breakpoint in the modified emulator, first launch the
    emulator, then start the debugger in CodeWarrior. In the emulator,
    choose the "Breakpoints" menu item. A dialog window will pop up;
    select the "Enabled" check box, then enter a start address and a
    number of bytes to monitor. The address and number of bytes may be
    either hex ("0x89abc") or decimal ("12345"). As an example, if the
    start address is 0x10000 and the number of bytes is 8, then any
    write to addresses 0x10000 through 0x10007 will break into the
    CodeWarrior debugger.

    Conditional breakpoints allow you to break when execution reaches a
    given address, but only when a certain condition is true.
    CodeWarrior for PalmOS supports conditional breakpoints, but its
    implementation is so slow as to be practically useless. As I
    pointed out in an earlier post to the palm-dev-forum list, in the
    CodeWarrior debugger, each iteration past a conditional breakpoint
    takes about 0.3 second, so you have to wait for an eternity if you
    have to iterate hundreds or thousands of times until the break
    condition is true (as is often the case in debugging).

    The modified emulator allows you to set conditional breakpoints with
    virtually zero overhead. To set conditional breakpoints, first
    launch the emulator, then start the CodeWarrior debugger. When you
    are stopped in the debugger, choose the "Breakpoints" menu item in
    the emulator. A dialog window will pop up which will allow you to
    set up to 6 conditional breakpoints. To set a breakpoint, choose a
    breakpoint slot and press the "Edit" button. A window will pop up
    where you can enter an address and a condition. The address can be
    either hex ("0x2468ace2") or decimal. Typically, you will determine
    the address where you want to break by selecting the "mixed" view in
    CodeWarrior to see a mix of source and assembly code. The condition
    you specify must be of the form "<register> <condition> <constant>",
    where

        <register> is one of the 68000 registers D0...D7 or A0...A7
        <condition> is ==, !=, <, >, <=, or >=. (Important: for now,
            all comparisons are UNSIGNED. This means that you can't use
            a condition such as "D0 < 0", which will always be false).
        <constant> is a hex or decimal constant

    By choosing the "mixed" view in CodeWarrior, you can see which 68000
    registers represent which local variables in your program, and so
    you can construct an appropriate break condition involving a
    register.

*    Exposed Palm IIIx and Palm V support.

*    Fixed problem with ROM Transfer on EZ devices. [Yoshiyuki Kubo]

*    Relaxed form-object-validation rules to allow for zero-sized
    gadgets and tables. [Roger Flores]

*    Fixed bug in AutoRunAndQuit mechanism where Poser would quit when
    a sub-launched application would quit, not when the application-of-
    interest would quit. [Steve Haneman]

*    NetLibDmReceive should really work this time. [Doug Morrison]


Changes for 2.1d25 (2/16/99)
----------------------------
*    Different method for checking whether or not a library function is
    implemented. Poser still only checks known (i.e., Palm Computing)
    libraries, but now instead of using a hardcoded table, it takes
    advantage of the fact that the "open library" function appears right
    after the dispatch table. [Tim Wiegman, Adam Hampson]

*    (Mac) Added WDEF project to Poser project. Changed the way the
    outline region is calculated (should be more accurate now).

*    Added support for internal faster trap dispatching mechanism. The
    major effect of this is that Poser can now support non-debug Palm
    VII ROMs again. [Bob Ebert]

*    Added support for Autorun and AutorunAndQuit directories. These
    directories are like the Autoload directory. Applications in
    Autorun are loaded, and then one is chosen to be executed (by
    convention chosen to be the last file in the directory).
    Applications in AutorunAndQuit are loaded, and then one is chosen to
    be executed in the same way. When that application quits, Poser
    quits. If there are files in both Autorun and AutorunAndQuit, the
    last file in AutorunAndQuit is chosen for execution. The files in
    those directories can also contain launchable documents such as
    .PQAs. [Steve Haneman]

*    (Windows) Centered the LCD. [Scott Johnson]

*    Better logging facilities. Here's what we've got so far:

    - Error Messages: undefined and unimplemented.

    - Warning Messages: If a dialog comes up that can be dismissed by
        clicking on the Continue button, this option causes the message
        to be logged. - Misc Gremlin Info: undefined.

    - Assembly Opcodes: unimplemented. Will eventually log assembly-
        level trace information (registers, PC, opcodes, etc.)

    - Posted events: events entered into the system by calls to
        EvtAddEventToQueue, EvtAddUniqueEventToQueue,
        EvtEnqueuePenPoint, and EvtEnqueueKey.

    - Received events: events returned by EvtGetEvent, EvtGetPen, and
        EvtGetSysEvent,

    - System calls: calls to Palm OS functions.

    - Application calls: unimplemented. Calls to functions in your own
        application. In order for this to work, the name of the function
        needs to be stored in memory following the function itself.
        CodeWarrior supports this convention, gcc currently does not.

    - Serial Activity: serial port being opened and closed, changes in
        serial port settings.

    - Serial Data: data being transmitted and sent.

    - NetLib Activity: calls to NetLib functions, parameter values and
        return values.

    - NetLib Data: data being transmitted and sent.

    - ExgMgr Activity: unimplemented. You get the idea...

    - ExgMgr Data: You get the idea...

    - High-level Debugger Activity: messages received from an
        external debugger and the replies sent back.

    - High-level Debugger data: details of the messages sent back and
        forth. Not all packets currently display all their data when
        this mode is turned on.

    - Low-Level Debugger Activity: trace of the low-level mechanisms
        that receive raw data from external debuggers and the raw data
        being sent back.

    - Low-Level Debugger Data: dumps of the raw data being sent back and
        forth.

    These options are presented in a dialog box with two tabs: Normal
    and Gremlins. The first tab contains options which are active
    during normal emulator use. The second tab contains options which
    are active when using Gremlins.

    Along with this new dialog, the Debug Options and New Gremlins
    dialogs have also been changed. The logging checkboxes have been
    removed from Debug Options (along with some currently unused
    checkboxes). The logging checkboxes in the New Gremlins dialog have
    been replaced with a "Logging Options..." button that brings up the
    Logging Options dialog. A new checkbox ("Continue Past Warnings")
    has been added to the New Gremlins dialog that controls what happens
    when a non-fatal error dialog is displayed; if possible, these error
    dialogs will automatically be dismissed if this option is checked.

    For performance reasons, all logged information is stored in an
    internal memory buffer. Additionally, only the most recent
    information is kept. After a certain threshhold, old information is
    discarded. The amount of information kept is initially 1Meg, but
    that can be changed by calling HostSetLogFileSize.

    The in-memory buffer is flushed on the following occassions:

        - The application quits.
        - An error is displayed.
        - A new Gremlin is started.
        - An exception occurs.

    Logged data is written to a file named "Log####.txt", where #### is
    a number ensuring the file name's uniqueness.

    Contents of the log file are of the form:

        <Time>    <Gremlin Event>        <Text>

    <Time> is the zero-based time in seconds, where the base is the time
    of the first event logged.

    <Gremlin Event> is the event sequence number if Gremlins is running.
    This field is omitted if Gremlins is not running.

    <Text> is the log text. Output is generally one line long, with the
    major exception being hex dump information.

*    Allow more concurrent tailpatches. Previously, there was a fixed
    number of traps that could be tailpatched at once. Exceeding this
    limit would cause Poser to abort().

*    Allow one particular SecurityLib function to access low-memory.
    [Gavin Peacock]

*    Implemented HostSetLogFileSize. Increased default log size from
    128K to 1Meg. [Dan Rowley]

*    Support netSocketOptSockNonBlocking in NetLib redirection. Also
    support the UDP calls (NetLib[Send|Receive](PB) with non-NULL
    address specifications). With these changes, Windows users can now
    Network HotSync. [Tilo Christ]

    Here's now (this is for Palm III):

    - Get the Network HotSync files from:

        <http://www.palm.com/custsupp/downloads/netsync.html>

    - Run "nhssetup.exe" to extract "setupex.exe".

    - Run "setupex.exe" to install Network HotSync.

    - Fire up Poser.

    - Install "netsync.prc". There will be one in your ...\Palm
        Desktop\ <UserName>\Install\ directory or your ...\Palm
        Desktop\add-on\ directory.

    - Go to the HotSync application.

    - Select the "LANSync Prefs..." menu item.

    - Select the "LANSync" option. Click on "OK".

    - Select the "Primary PC Setup..." menu item.

    - Enter your Primay PC's IP address. You can get this from the
        HotSync Manager (click on the HotSync icon in the TaskBar,
        select the "Setup..." menu item, click on the "Network" tab, and
        click on the "TCP/IP Settings..." button). Click on "OK".

    - Click on the "Select Service" button under the "Modem Sync" icon.

    - Create a new service or modify an existing one. Either way, the
        result should be a service with the phone number set to "00".
        Exit that screen.

    - Click on the "Modem Sync" icon and sync as usual.

*    (Mac) Additional note on NetLib redirection: currently, support
    is implemented with the help of the GUSI library. This library is
    built on top of MacTCP, which doesn't support the settting or
    getting of socket options. Any call to NetLibSocketOptionGet or Set
    will result in an error on the Mac. Eventually, emulation will be
    performed with direct Open Transport calls, and this limitation will
    go away.

    In the meantime, this limitation affects the ability to Network
    HotSync from Poser runningon the Mac to HotSync running on a PC.

*    Fixed a long-standing 2.1dx bug where the screen would not always
    reflect the state of the LCD buffer. One way this could happen is
    if the screen were modified in a strictly high-memory to low-memory
    order (as happens when a document is scrolled down).

*    (Windows) Fixed byteswapping problem in the addresses returned by
    NetLibGetHostByName. [Pedro Alves]

*    The text file dumped by the profiler always used to be named
    "Profile Results.txt" and would be written to the current default
    directory. Now the file name is based on the file name used for the
    binary output and is stored in the same directory as it. This is
    mostly useful if the output file name is specified by
    HostProfileDump (). [Steve Haneman]

*    While running Gremlins, validate the sizes of various form objects.
    [Chris Antos] The rules are:

    *    frmBitmapObj, frmLineObj, frmFrameObj, frmRectangleObj,
        frmLabelObj, frmTitleObj, frmPopupObj: Don't care. We don't
        touch them or look at them.
    *    frmControlObj: if the style is popupTriggerCtl, allow a
        zero-width object.
    *    frmListObj: if the number of items is zero, allow a zero-height
        object.
    *    In all other cases, disallow zero-height or zero-width objects,
        as well as objects completely offscreen.

*    (Windows) Added mnemonic keys to the menus and dialogs. [Scott
    Johnson]

*    Adjusted the internal rectangle indicating the bounds of the hard
    buttons on the case graphics, now that the case graphics have
    changed. [Horace Ho]

*    Enable storage heap write access in our NetLib-redirected version
    of NetLibDmReceive. [Neil Rhodes]

*    Turned back on the checks to see if someone accesses the area from
    0x00018000 to 0x00020000. [Adam Dingle]

*    Added Japanese color silkscreens (no B&W yet).


Changes for 2.1d24 (1/25/99)
----------------------------
*    Yow...time to start updating those copyright statements...

*    Fixed performance problems with logging (this really affected
    Gremlins performance). [Andrew Ball]

*    Made small change to heap tracking routines; this change really
    speeds up overall performance. A lot of performance tuning still
    needs to be done here. [Scott Johnson]

*    NetLib redirection. [Phil Shoemaker, Ron Marianetti]

    Here's the current list of what doesn't yet work:

        Raw sockets (not on the Mac at all, works on Windows only if
            your version of WinSock supports it).

        PalmOS-specified options:
            netSocketOptSockNonBlocking
            netSocketOptSockRequireErrClear
            netSocketOptSockMultiPktAddr

        BSD sockets options that are not supported by "NetLib redirection"
        because the underlying WinSock library doesn't support them:
            netSocketOptTCPMaxSeg
            netSocketOptIPOptions
            netSocketOptSockSndLowWater
            netSocketOptSockRcvLowWater
            netSocketOptSockSndTimeout
            netSocketOptSockRcvTimeout

        Timeouts

        NetLibGetMailExchangeByName

        UDP stuff

        Turning NetLib redirection on and off on the fly. It might be
        safer to change the setting, then quit and restart Poser (or at
        least reset the ROM).

        MacTCP doesn't appear to timeout if it can't make a connection
        to a socket on a remote machine that isn't listening on that
        socket. There may be other cases where MacTCP doesn't timeout
        or return some other error. In those cases, the PalmOS application
        will hang. You should be able to press Cmd-. to get out of those
        situations, causing the patched NetLib to return a user-cancelled
        error number.

    Comments on how NetLib redirection works:  Poser has the ability
    to intercept process execution at the points where an OS function
    is called and where the OS function returns. Poser can then
    execute additional code at these points, and can even skip
    calling the ROM version at all if it wants. That's what happens
    with NetLib redirection. The following calls are intercepted
    and replaced with versions that make the corresponding host-
    specific Ethernet calls:

        NetLibClose                        NetLibConnectionRefresh
        NetLibDmReceive                    NetLibFinishCloseWait
        NetLibGetHostByAddr                NetLibGetHostByName
        NetLibGetMailExchangeByName        NetLibGetServByName
        NetLibHandlePowerOff            NetLibOpen
        NetLibOpenConfig                NetLibOpenCount
        NetLibOpenIfCloseWait            NetLibReceive
        NetLibReceivePB                    NetLibSelect
        NetLibSend                        NetLibSendPB
        NetLibSleep                        NetLibSocketAccept
        NetLibSocketAddr                NetLibSocketBind
        NetLibSocketClose                NetLibSocketConnect
        NetLibSocketListen                NetLibSocketOpen
        NetLibSocketOptionGet            NetLibSocketOptionSet
        NetLibSocketShutdown            NetLibWake

    Because NetLibOpen (and its cousins) is entirely patched out, the
    Palm OS version of NetLib never actually gets opened. This means
    that the remaining, unpatched NetLib calls are made in the context
    of a closed NetLib. I don't know if this differenc will affect
    anyone, but it might be good for you to know.

    Comments on Palm VII supprt:  Palm VII files are not yet externally
    available. But Poser is preparing for that day, so when they *are*
    available:

        Download the prerelease Palm VII rom and the required Debug
        wireless panel which will configure the emulated OS so that
        it knows you're not on a wireless device.

        To configure: open the ROM with the Palm OS Emulator, wait
        until it has completed the boot process, and install the debug
        wireless panel by dragging the prc onto the running emulator
        window. Now open the Prefs application and go to the "Wireless"
        panel.    You'll see a box there entitled "wireline ip".    Tap on
        it, clear the existing selection, and enter "207.240.80.136".
        Finally, open the preferences for the Palm OS Emulator itself
        (right-mouse click under Windows), and turn on the "Redirect
        Netlib calls to host TCP/IP" option.

        Now when you use a PQA, it will connect to the web site with
        your PC's internet connectivity instead of a wireless radio.
        (Warning: it is much faster that way!  Pages will load much
        more slowly when transmitted over the air.)     Also, there may
        be warnings where the Emulator notices possible bugs in the ROM.
        Usually these are benign (but indicate the prerelease status of
        the ROM) - pressing "continue" will usually get you by them.

        You can install PQAs by dragging them onto the Emulator window,
        or using the Emulator's menu commands.    Just make sure you're
        neither in the Launcher when installing an application, nor
        running the app you're upgrading. (Installing software while
        the device is booting is also a bad idea.)    The Calculator is
        a convenient, safe app to be in when installing software.

        Note that if your PC is behind a firewall, you might not be able
        to connect to our proxy server. Ask your system administrator to
        enable port 5002 for you.

        If you get an error message along the lines of "no interfaces
        found: net - 12", that's an indication that you didn't select
        "Redirect NetLib" in the preferences panel. Getting the error
        message "wrong socket type" is an indication that you didn't
        install the DgbWirelessPnl application.


*    More fixing of semaphore warnings: duh, changing the text of the
    warning message doesn't change the test itself! [many...]

*    Made .prc/.pdb file loading more robust: the contents are now
    validated before accessing them. This helps against problems
    like 0x0A's turning into 0x0D's because of network downloading.

*    (Windows) Fixed byteswapping problem that would occur when gdb
    tried to set a breakpoint [Ben Thomas].

*    (Mac) Fixed problem with getting 'Could not save the file ""
    because the file wasn't found' when trying to save a session.
    [???]

*    Palm OS 1.0 support: Gremlins and pen events weren't being
    inserted into the system because EvtGetSysEvent blocked on
    SysSemaphoreWait instead of SysEvGroupWait in that version
    of the OS (we were only patching the later to detect when
    to insert events into the system).    Now I patch both functions.
    [Andrew Ball, Ken Hancock]

*    Set HotSync user name at boot-up. Name specified in preferences.
    [Richard Ziegelmaier, et al]

*    Check for and report for out-of-range access to library functions.

*    The workaround for SysAppExit in d22 was hiding the workaround for
    FindSaveFindStr from d20. Adjusted the tests so that one no
    longer hides the other. [Matt Disher]

*    Exposed Palm VII and Web Clipping support in external release.

*    (Windows) Fixed a problem in the mechanism for exchanging debugger
    information with the Metrowerks debugger. This problem could
    potentially cause debugger packets to be ignored. However, there's
    no evidence that this is the cause of the "Failed to connect to
    remote" problem that some people are still reporting with R5.
    [Ben Sessoms]

*    When handling SysPktRtnNameCmdType packets, force the given address
    to be even. Otherwise, we'll generate an internal error as we try
    to walk the function starting at an odd address. [Ben Sessoms]

*    Fixed long-standing bug that would crash the New Gremlin dialog if
    the last item in the list had a ROM-based version and a RAM-based
    version.

*    Sort the items in the New Gremlin dialog case-insensitively.

*    Added support for launching applications with documents. [Gavin
    Peacock]

*    Updated bitmaps. [Ken Krugler, Stephen Best]

*    Fixed problem with getting a "yadda accessed memory manager data
    structure" while dumping the results of profiling. [Adam Dingle]

*    Removed most (hopefully all) influences of Poser's memory
    accessing and ROM function calling from the profiling output.

*    Fixed the profiler to handle TRAP $F / <dispatch number> sequences
    where <dispatch number> is not in the range 0xA000-0xAFFF (some
    development systems leave off the leading "A"). [Adam Dingle]

*    Bumped the default number of profiled functions from 20000 to 48K.
    [Steve Haneman]


Changes for 2.1d23 (12/16/98)
-----------------------------
*    (Windows) fixed some byteswapping problems with debugger packets.
    (Actually, put *back* some bugs that I fixed...)

*    Re-organized sources.

*    Added ability to patch libraries. Started patching out NetLib.
    Note that there is no actual NetLib redirection in this release.

*    Fixed some serial port emulation problems: mirrored UART settings
    into the host serial port just after the port has been opened, not
    after the first call to SerSetSettings (there may *not* be a call
    to SerSetSettings!).

*    Generating a profiling report now also produces a text file for
    use on platforms that can't read the Metrowerks Profile tool
    file format.

*    (Windows) Added profiling configuration.


Changes for 2.1d22 (12/7/98)
----------------------------
*    Only warn about MemSemaphore usage on ROMs that themselves hold the
    memory semaphore correctly. This time for sure!  Also only warn
    when the semaphore has been held for 1 minute (previously, the
    timeout was 10 milliseconds).

*    Completely overhauled serial port emulation. More of the UART is
    emulated more faithfully.

*    If no serial port is selected for emulation pass through and we're
    not in loopback mode, drop transmitted bytes on the floor.

*    If IrDA is using the serial port, drop transmitted bytes on the
    floor (for now).

*    Re: SysAppExit accessing a deleted stack: SysTaskDelete and a whole
    host of other functions also access the deleted stack until the
    kernel has a chance to switch to another stack. Much gnashing of
    teeth was performed until an appropriate alternate heuristic could
    be found.

*    Let PrvCompressedInnerBitBlt access bits outside the allocated
    buffer.

*    (Windows) Fixed problems with Cancelling a ROM download.

*    Another attempt at fixing -192 errors at startup.

*    (Mac) Fixed problem with opening two windows when starting a session
    by double-clicking on a ROM file.

*    (Mac) Goosed debugger messaging. Previously, an asynchronous
    callback function would get called when Poser received a packet from
    the debugger. The callback function would set a boolean and then
    call a Toolbox function to wake up the application. The next time
    the application's main-event-loop was entered, the packet would be
    handled. However, the overhead of waking up the app and waiting for
    the main-event-loop to be entered was way too high. I now handle
    the debugger packet immediately in the asynchronous callback
    function. This means that the debugger packet is handled in the
    context of the calling application (the external debugger). If
    Poser were to crash at this point, it would crash the debugger, not
    Poser.    Additionally, we have to be very careful not to allocate any
    memory from within the callback, as such memory would be allocated
    from the debugger's heap, not Poser's.

*    Changed some of the rules for interacting with an external
    debugger.

    In order for an external debugger (PalmDebugger, MWDebug, or gdb) to
    talk with a hardware device, the hardware device has to be
    listening. There are two things a debugger can talk to: the Debugger
    socket and the Console socket.    These two sockets correspond to the
    two windows in the PalmDebugger.

    To get the ROM to listen on the Console socket, the developer has to
    execute shortcut-.2. To get the ROM to listen on the Debugger
    socket, the device has to cause an exception somehow. Doing
    something wrong (e.g., access invalid memory, divide by zero, etc.)
    does this. Executing a DbgBreak also does this, which is what
    happens when the user executes shortcut-.1. After the exception
    occurs, the ROM goes into a loop where it listens for and responds
    to debugger commands.

    Because external debuggers talk directly with Poser and not the ROM
    when debugging with Poser, there's no need to execute shortcuts or
    cause exceptions in order to start debugging. Poser is perfectly
    capable of handling debugger requests, letting the emulated process
    continue on its merry way.

    However, when dealing with commands sent to the debugger socket,
    it's very important that the state of the emulated CPU not change.
    Often several commands are issued, ultimately ending with some
    command that restarts the emulated CPU. During these exchanges, the
    debuggers assume that the state of the CPU has not changed since
    "debug mode" was entered. For this reason, when commands are sent
    to the Debugger socket:

        - Poser needs to be in "debug mode". That means that it has
          halted any emulation and is just idling.    If commands for
          the Debugger socket come in and Poser is not in debug mode,
          the commands are ignored.
        - The exception to this rule is the "get state" Debugger
          command.    This command is issued when the user types "att"
          in the Debug window.    It indicates that the user wants to
          enter debug mode for the purpose of setting breakpoints,
          examine memory, etc.    As a courtesey, Poser will enter
          debug mode when it sees this command.

*    Install a 'pose' feature so that applications can tell if they are
    running under Poser or not.

*    Install a 'gdbS' feature when connected via a socket to an external
    debugger. The existance of this feature causes programs written
    with prc tools to enter the debugger when they are launched.

*    (Windows) Removed a minimum version check that was preventing us
    from using sockets on Windows '95.


Changes for 2.1d21 (11/12/98)
-----------------------------
*    Allow for NULL access bug in FldDelete.

*    (Windows) Allow for launching of Emulator by double clicking on
    .psf or .rom files.

*    Added back a whole slew of strings accidentally deleted in d20.


Changes for 2.1d20 (11/10/98)
-----------------------------
*    Support full debugger protocol for latest PalmDebugger (with
    integrated Poser support).

*    Fixed button mapping problems with EZ hardware.

*    Added Greg's Ymodem code for downloading ROMs.    This includes his
    fix for high-baud rate downloads.

*    Support for higher baud rates.

*    Re-establish serial port after Preferences/Properties dialog.

*    Allow for access to deleted stack in SysAppExit.

*    Allow for bug in NetPrvTaskMain that accesses low-memory.

*    Allow for bug in NetPrvSettingSet that reads from a random memory
    location.

*    Tighten up on some inappropriate memory checks (only let some
    functions get away with murder in older ROMs).

*    Removed check for bug in PrvGetBatteryPercent (it's been fixed).

*    Changed formatting of time displayed in Gremlins window. Was in
    seconds, now in hh:mm:ss.

*    Revoked Memory Manager privileges for MemCmp, MemMove, and MemSet;
    they no longer have free run of RAM like other MemMgr functions.

*    Allow for bug in FindSaveFindStr which reads off the end of the
    handle passed to it.

*    Similar comments for FntDefineFont. When copying the contents of
    an old, smaller buffer into a new, larger buffer, it copies the
    size of the new buffer, causing MemMove to walk off the end of the
    old buffer.

*    Let ScrInit access unallocate memory for when it copies the splash
    screen from its temporary location to the real LCD buffer.

*    Re-enabled code that reports a bug in BackspaceChar, but only for
    newer ROMs.

*    Internal change: Better encapsulation of what Palm OS bugs are
    fixed in what version.

*    Only warn about MemSemaphore usage on ROMs that themselves hold
    the memory semaphore correctly.


Changes for 2.1d19 (10/30/98)
-----------------------------
*    (Windows) Fixed problem with trying to auto-reload the previously
    used .psf file on startup.

*    (Windows) Allow the user to click on the close box of the Startup
    dialog (the one saying New/Open/Download/Exit).

*    Fixed two problems with heap management on 2.0 ROMs. First, I was
    failing to recognize the fact that zero in the size field of a heap
    header means that the heap is 64K bytes long. Second, I was
    attempting to access *all* fields of the terminating sentry chunk.
    This chunk only has a size field; attempts to access subsequent
    fields may fail if the sentry is at the end of RAM.


Changes for 2.1d18 (10/29/98)
-----------------------------
*    Disabled Profiling menu items and Transfer ROM menu item until
    the functionality behind those items can be implemented.

*    Added "Dragonball EZ" menu item to device list.

*    (Windows) Fixed problem with stale reference to the LCD window after
    its been closed.

*    Fixed problem in MC68EZ328Bank::GetLCDDepth where it would return 4
    for grayscale mode instead of 2.

*    (Mac) Fixed problem getting -192 error on startup with a fresh
    set of preferences.


Changes for 2.1d17 (10/28/98)
-----------------------------
*    Fixed bug in Gremlins when logging information; an internal counter
    would wrap to be negative after a large number of events have been
    posted.

*    Re-laid out Debug Options dialog and hooked up the checkboxes.
    Disabled the ones that aren't supported yet.

*    (Windows) Fixed another "Unhandled exception caught in main" bug:
    this one would occur when the Poser would try to open a .psf file
    that no longer existed (the previous cause of this bug occured
    when Poser would try loading a .rom file that no longer existed).

*    (Windows) Remove old registry entries when found.

*    When an error occurs accessing a file, consistantly set the
    %filename parameter for error messages.

*    Added "Log error messages" checkbox to New Gremlin dialog.

*    Save Gremlins settings to preferences.

*    Fixed bug in meta-memory tracking stuff that would trash the
    emulator's heap if it was emulating a device with 256K of RAM
    or less.

*    Clear the serial FIFO queues on reset.    Fixes a problem with their
    overflowing and serial port emulation is not turned on (that is,
    "None" is selected in the Preferences/Properties dialog).

*    Fixed problems with trying to use "&string_object[0]" on empty
    strings. Apparently, there is not necessarily any memory
    allocated for empty string objects, so attempting to take the
    address of the string fails.


Changes for 2.1d16 (10/22/98)
-----------------------------
*    Allow the opening of locked .rom, .psf, .prc, .pdb, and .pqa files.

*    Updated TextMgr calls to reflect new (long) parameters.

*    Added Autoload facility.

*    Fixed problem with trying to wakeup the device (by calling
    EvtWakeup) when the interrupt level mask was too high.

    (Backed this modification out temporarily.    Seems that the ROM can
     spend a large percentage of its time at this interrupt level if
     it's not doing anything else useful, preventing the user from
     clicking on the LCD screen.)

*    Conditionalized the code so that Palm VII, Palm V, and Palm IIIx
    could be included or excluded separately.

*    Conditionalized *this* file so that secret stuff can be stripped for
    external releases.

*    Fixed the "Save Screen" functions (I'd commented them out while
    adding support for lcdPageWidth and lcdPanningOffset registers and
    forgot to go back and clean up).


Changes for 2.1d15 (10/15/98)
----------------------------
*    Bug fix: my wrapper function for TxtGetCharBounds was accidentally
    calling TxtGetNextChar (copy & paste error).

*    Bug fix: the routine that maps a range of "host" memory into
    "emulated" memory would round the base address down to the nearest
    multiple of four, but the routine that would reverse the process
    didn't do the same rounding. This meant that the routine that
    unmapped would be unable to find the address recorded by the routine
    that did the mapping, and entries would never get unmapped.

*    Bug fix: if Poser reads in a preference setting for a Japanese
    silkscreen and that version of Poser doesn't support Japanese
    silkscreens, change the setting. Otherwise we'd get a -192
    (resource not found) error.

*    Relaxed MemSemaphoreReserve/Release testing. Now only track time
    when "write" parameter is true.

*    Bug fix: allow debuggers access to reserved parts of Dragonball
    memory.

*    (Mac) Show errors in a movable modal dialog, allowing the user
    to switch to the Finder, possibly for the purpose of launching a
    debugger.

*    Allow selection of multiple items in New Gremlins dialog. These
    items are also reselected when the dialog is re-opened.

*    Added HostGremlinNumber, HostGremlinCounter, HostGremlinLimit.

    "The reason I'm asking is that I'd like to log info to figure out
     what's going wrong with a specific gremlin test case, but since
     it's pretty far along (e.g. 100000 steps) I only want to start
     logging when I'm getting close to where I think the problem will
     occur. The problem occurs after a different number of steps
     depending on the gremlin seed value, thus I'd like to set my
     'start recording' value based on which gremlin is running."

*    Added HostLogFile.    This function returns a special HostFILE* that
    can be used by Palm OS applications that want to write to the same
    file that any Poser facility is writing to. This HostFILE* is
    special in that it is only open when Poser wants it to be, and is
    write-only; any HostFxxx functions that attempt to read from
    HostFILE's will fail when fed this HostFILE*.

    Output occurs only when Poser has turned on logging. If the
    Palm OS application calls something like HostFPrintF(hostLogFile,
    "...", ...) and logging is not turned on, the output is discarded.

*    Added HostSetLogFileSize. This function sets the maximum amount of
    text written to the file created by Poser's logging facilities.

*    If Gremlins is running and a message needs to be displayed that can
    be handled by clicking on a Continue button, write that message to
    a log file instead and implicitly click on the Continue button.

*    If Gremlins is running, turn off the facility that causes the user
    to be warned only once of certain kinds of errors (that is, the
    error will be logged to the file every time).

*    In the logging output file, add a comment that only the last xxxK
    of information is recorded.


Changes for 2.1d14 (10/1/98)
----------------------------
*    Added Japanese silkscreens, and support for selecting them.

*    (Mac) Added "None" menu item to serial port menu.

*    Added support for lcdPageWidth and lcdPanningOffset registers.

*    Fixed bug with power-off (I wasn't allowing HwrSleep to change some
    low-memory vectors, resulting in a warning dialog).

*    Check for and report heap corruptions.    Poser walks the heap
    regularly, usually to sync up its notion of where things are with
    the Palm OS Memory Manager's. These walks used to be vulnerable to
    corrupted heaps. Now they should be more robust.

*    (Mac) Updated to CodeWarrior Pro 4.

*    Instead of patching ErrDisplayFileLineMsg and showing its text in
    our own dialog, let the ROM version execute. It will eventually
    call SysFatalAlert, which we patch for the same reason (to show the
    text in our own dialog with our own buttons).

*    Added stack overflow checking.    On all opcodes changing the stack
    pointer (PEA, MOVE, MOVEM, JSR, and any opcode operating on a
    register with pre-decrement mode), I check the stack pointer to see
    if it's (a) close to the end of the stack or (b) past the end of the
    stack. The former gets a warning, the latter gets an error.

*    Initialize new RAM to 0xFF (make it look like new flash RAM).

*    (Windows) Fixed "Unhandled exception caught in main()" error message.
    This would show up when the previously-used ROM file could not be
    found. Now, the (slightly) less enigmatic message "Could not start
    a new session because the file could not be found" is displayed.
    More importantly, Poser no longer quits after displaying the message.

*    Added check for how long the Memory Manager semaphore was held.
    Show a warning if held for more than 10 milliseconds.

*    Fixed problem with multiple Gremlin Control windows appearing in
    the windows version. Doh!

*    Made the "Release" target an "External Release" target, and added
    an "Internal Release" target.

*    Added support for loading clippings.

*    Made a change that should take care of accesses to invalid memory
    from an external debugger.


Changes for 2.1d13 (9/15/98)
---------------------------
*    Fixed bug in heap validation functions; when I extract the *signed*
    hOffset value, I have to sign extend it from 24-bits to 32-bits.

*    Mark more parts of memory as memory manager data structures.

*    Support hardware sub-ID for EZ devices.

*    Support HwrIdentifyFeatures for 68328 devices.

*    Enable flash for Palm V and Palm IIIx.

*    Fixed long-standing Gremlins bug that would cause event pushing to
    stop after a while (the Gremlin mechanism would enqueue and event
    and then wait for it to be handled; however, in some cases, the
    event would fail to be enqueued, and the Gremlin mechanism would
    wait forever for that event to be handled).

*    Allow for a bug in SysBinarySearch in pre-3.0 systems that would
    cause it to pass an invalid pointer to the search callback.

*    Added Gremlins Control window (displays gremlin number, current
    event number, and elapsed time; also contains Stop, Resume, and Step
    buttons).

*    Tweaked Gremlins engine so that doing a Step doesn't reset the
    "until" value.

*    Performance enhancements:

    - Changing some functions to return "int" instead of "bool"
      resulted in better code generation, leading to a 7% performance
      boost in my canonical Gremlins test (Gremlin #0, 10000 events
      on the Memo Pad). (Actually, this was only a win in VC++;
      CodeWarrior worked better with "bool". Therefore, all of the
      sources were changed to use "Bool", which is typedeffed to
      "bool" or "int" depending on the compiler being used.)

    - Putting back an enhancement from Craig's Mac Copilot lead to a
      27% performance boost. The idea is to update only the scanlines
      that are changed since the last screen refresh. Craig did it by
      carefully comparing the current LCD buffer with a saved version
      of it. I now do it instead by noting when the LCD buffer is
      being written to (this is done fairly efficiently via the
      MetaMemory tracking routines) and remembering the first and last
      bytes in the LCD buffer that were written to.

*    Fixed trap number used in HostControl.h.

*    Updated to latest CharEncodingType values from TextMgr.h.


Changes for 2.1d12 (9/2/98)
---------------------------
*    EmuTraps -> HostControl.

*    Let PrvFindMemoryLeaks walk the heap.

*    Added HostGetEnv.


Changes for 2.1d11 (8/31/98)
----------------------------
*    Added support for AMD flash RAM.

*    (Windows) Fixed support for external debugging.

*    (Windows) Display dialog if an exception is caught in WinMain.

*    (Windows) Fixed the way the Dragonball(EZ) registers were
    initialized. After things got going, the registers were OK, but the
    initial values on little-endian machines was completely bogus.

*    (Mac) Updated creator resource text.

*    Totally revamped the way memory access checking is performed. In
    addition to detecting access to low-memory, system globals, the
    screen buffer and Dragonball(EZ) registers, I validate access to the
    dynamic heap. Now, only the Memory Manager can fiddle with chunk
    headers and trailers, free chunks, and unlocked chunks. For the time
    being, low-stack access is no longer detected, though this feature
    will return. Eventually, we should also be able to detect access to
    uninitialized memory.

*    Updated to latest sysFtrNumXXX values.

*    Fixed test for Intl Mgr in gremlins.

*    Test for NULL in stub Intl Mgr functions.

*    I now calibrate the pen before running a gremlin. This is
    necessary for consistant results between different runs across
    boots.

*    Fixed error reporting problem where previous messages reported in
    dialogs would inhibit other messages from appearing (that is, the
    attempt to display a second message would end up showing the first
    error message again).

*    Handle errant accesses to Dragonball space past the end of the last
    defined register.

*    Check for and report unimplemented traps. This check detects
    attempts to call functions that are off the end of the trap table as
    well as calls to SysUnimplemented.

*    Internal change: consolidated more error handling in the Errors
    class.    Cleaned up the Errors interface a little bit.

*    Added temp hack to let PrvGetBatteryPercent read off the end of an
    allocated chunk of memory.


Changes for 2.1d10 (8/18/98)
----------------------------
*    Display correct error message if a low-stack access is performed (I
    was previously re-using the "screen buffer access" message).

*    Low-memory access by the ROM was previously allowed up until
    UIInitialize was called. That's been changed to TimInit.

*    Consolidated some "device" menu items in the New Session dialog,
    and added a new device.

*    Per Steve and Roger, changed some function names in EmuTraps.h.

*    Added Control Panels to the list of applications in the New Gremlin
    dialog box.

*    Added support for DbgMessage.

*    Internal change: Roger Flores Memorial Constant Eradication. Got
    rid of a lot of constants in EmuTraps.cpp and TrapPatches.cpp. These
    constants were used to fetch parameters from the emulated stack
    pushed on there by emulated code. I now define structs that
    describe the stack contents and use those (with the offsetof macro)
    to get the values.    In addition to the obvious one, another
    advantage is that the macros can now determine the _size_ of the
    parameters and call get_long, get_word, or get_byte accordingly.

*    Rolled in more IK/Gremlin changes for Ken.

*    Double-clicking on an app in the New Gremlin dialog will implicitly
    click on the OK button.

*    (Mac) Support Cmd-D in the "Save this file?" dialog (you guys owe
    me!)

*    Mapped some internal error numbers to error strings.

*    Re-establish the interrupt state of any pressed buttons after a
    reset.    This allows you to press the Power key and Cmd/Ctrl-R,
    resulting in a Hard Reset.

*    Asking for the 'irda' ROM token will now always fail (at least,
    until we can get some IrDA support implemented). NOTE: the ROM
    calls HwrGetROMToken directly, which means that it will not see the
    effect of this patch.

*    Internal change: Platform::DisposeMemory now takes a reference to a
    pointer so that it can set that pointer to NULL.

*    Use Platform::AllocateMemory and Platform::DisposeMemory in the
    profiling code, so that it can use temp memory on the Mac.

*    EmuProfileDump now uses the name passed in to it. If NULL is
    passed in, a default name ("Profile Results") is used.

*    Started on checking memory accesses to the dynamic heap that are
    outside of allocated memory chunks. Not activated yet because (a)
    it's too slow and (b) the hueristics aren't quite right.

*    (Windows) Fixed bug with not reporting errors that occurred
    while initializing the emulator. Now you're notified if the serial
    port you've specified in the Properties dialog is busy.

*    (Windows) Made the example in the previous bullet a warning,
    not a fatal error.


Changes for 2.1d9 (8/4/98)
--------------------------
*    (Mac) Fixed problem where the emulator would attempt to save the RAM
    file with no name if "Save Always" were selected, the user chose
    Quit from the File menu, and the RAM file was new (i.e., had never
    been saved before).

*    Prevent the system from _reading_ low-memory if it's (a) not
    sleeping or dozing, (b) fully booted, and (c) not handling an
    interrupt.

*    Hooked up "Fill New Blocks" and "Fill Disposed Blocks".

*    Minor optimizations to DRAM_Bank accessors.

*    Rolled in Ron's changes for finding a function's name.

*    Detect and report accesses to the space below the stack pointer and
    above the low-end of the stack. Note that there is no checkbox in
    "Debug Options" for this, but performance timing shows that there is
    very low overhead for this check (less than .1%).

    IMPLEMENTATION NOTE: detecting accesses to the unused parts of the
    stack obviously entail knowing where the stack is.    This information
    is gleaned from the SysAppInfoType block when the application is
    launched. However, there's no really good way to hook into the
    launch process. There are advantages and disadvantages to every
    approach, but the one used here is to hook into SysAppStartup and
    SysAppExit. This works for applications that use the standard
    runtime startup code, but applications are not guaranteed to use
    that standard code. Hopefully, the percentage of such applications
    is low.

    Additionally, SysAppStartup and SysAppExit are called under
    circumstances other than starting up the application. For instance,
    they are called when action codes are sent to each application. To
    differentiate the two, we ignore all calls to SysAppStartup that
    don't contain sysAppLaunchCmdNormalLaunch in the cmd field of the
    SysAppInfoType block.

*    Hooked up "Clean Stack" check box.    This option was originally
    defined to overwrite the parts of the stack that were freed up by an
    UNLK or RTS opcode in an attempt to detect accesses to those ranges
    of memory. Since that's now handled by the previous "*", "Clean
    Stack" now means that we initialize the part of the stack reserved
    by LINK opcodes, thus initializing local variables to well-defined
    bad values.

*    Fixed bug that occured when setting a breakpoint on an opcode just
    after a call to a system function that was tailpatched by the
    emulator.


Changes for 2.1d8 (7/29/98)
---------------------------
*    EmuGrmIsRunning (and SysGremlins) FINALLY returns a correct value
    instead of just returning "true" all the time.

*    Implemented EmuDmLoadFile and EmuDmSaveFile.

*    EmuDmSaveFile now takes a card number as the second parameter.

*    Added parameter checking to all EmuXXX functions. Functions that
    return EmuErrs will return emuErrInvalidParameter.    StdIO functions
    return error values according to the ANSI/ISO standard, and will set
    errno to emuErrInvalidParameter (note that functions that alter the
    return values of ferror and/or feof will NOT affect those value,
    which is DIFFERENT from the standard).

*    Fixed bug when setting the current time on startup. This bug would
    cause a crash on new Palm V ROMs.

*    Fixed bug with new Palm V ROMs when they tried to identify how much
    memory was on the "device".

*    (Windows) Fixed problem where clicking OK in the Properties dialog
    would crash the emulator under Windows 95 (and could possibly crash
    Windows 95, too!).

*    (Mac) Fixed problem where the serial port name would get lost,
    prohibiting the emulator from being able to open the serial port.


Changes for 2.1d7 (7/27/98)
---------------------------
*    (Mac) Don't feed the ROM mouse/pen events if we've turned off
    emulator execution at idle time. This prevents problems if a dialog
    comes up while tracking the mouse/pen.

*    Fixed long-standing bug from Mac Copilot 1.3 days.    The functions
    GetRegisters and SetRegisters can be used to save and restore the
    CPU registers.    The emulator used those functions for just that
    purpose when it would call ROM functions as subroutines. However,
    the status register was not faithfully saved, which meant that it
    was not always put back the way it was found. Since status bits
    would then occassionally change spontaneously out from under the
    emulated process, things would occassionally go "haywire" (that's a
    technical term).

*    Value returned by EmuHostGetVersion is now based on the
    application's version resource instead of being hard-coded.

*    (Mac) Save window position.

*    Implemented EmuHostIsSelectorImplemented.

*    Holding down the Ctrl key while booting/resetting will stop the ROM
    at the start of the Big ROM by using a temporary breakpoint.


Changes for 2.1d6 (7/23/98)
---------------------------
*    Changed way emulated apps called host functions. Those functions
    used to be called by invoking TRAP $E. Now we latch onto the
    SysGremlins trap number.

*    Extended set of host functions.

*    Zap some card header fields to say that we don't emulate flash RAM.

*    Fixed problem with profiling.

*    (Mac) Fixed an execution path where an untyped ROM file wasn't
    getting its type set.

*    Fixed problem with stopping on A-Trap breaks.


Changes for 2.1d5 (7/21/98)
---------------------------
*    (Mac) added version that does profiling.

*    Fixed loading of "whole" ROMs (those with Big ROMs and Small ROMs).

*    Fixed loading of .pdb files that have an AppInfo block but no
    records.


Changes for 2.1d4 (7/20/98)
---------------------------
*    (Mac) turned Open and Install menu items into hierarchical menus.

*    Fixed bugs with saving and loading session files.

*    (Mac) turned Appearance Manager classes back on when running on a
    sufficient machine (I'd accidentally left the application using the
    GA classes even when AM was available).

*    (Mac) create the Debug Options window invisibly; reduces flickering
    when opened.

*    Converted over to using the cross-platform classes FileReference
    (reference to and unopened file) and FileHandle (reference to an
    opened file). These classes allow more platform-independent code
    (such as the preferences code, which used to deal with Aliases on
    the Mac and file paths on Windows).

*    Overhauled error/exception handling/reporting.    I now use a
    consistant approach to detecting, handling, and reporting errors.

*    Rewrote the ROM image loading function. ROM images consist of two
    parts: a Big ROM and a Small ROM. The ROM image file always
    contains a Big ROM part, but not always a Small ROM part. However,
    the Small ROM part is optional and can be dummied up if necessary.
    The original version of the ROM image loader needed to load in the
    ROM image file once, determine if the Small ROM was included, and
    start over if it wasn't. The new ROM image loader is smarter,
    reading in just the parts it needs to determine if the Small ROM is
    there or not. Only after that does read in the full ROM file.

*    Put more strings in resources.    Most, but not all, are now done.

*    Added StdIO wrappers (see EmuTraps.h).


Changes for 2.1d3 (7/7/98)
--------------------------
*    Added progress bars when loading .prc and .pdb files.

*    Added support for drag and drop.

*    Now close previous session before opening the next.

*    (Mac) Added StandardGetFile file filter to allow files that start
    with "rom." or end with ".rom" or " rom", but that don't have the
    right file type.

*    Updated to latest EZ ROM header files (as of 7/6/98).

*    Update the low-memory checksum if low-memory is altered by an
    external debugger.

*    Fixed long-standing bug that would corrupt ROMs when switching from
    one to another. Since you couldn't do this on Windows in 2.0, you'd
    only see it in the Mac 2.0 version.

*    When creating the preferences for the first time, set the amount of
    RAM displayed in the New Configuration dialog to 1Meg.

*    (Mac) Allocate large chunks of memory from temp mem.


Changes for 2.1d2 (7/2/98)
--------------------------
*    Revamped startup sequence. From the ERS:

    If the application is started by double-clicking on its icon, it
    will attempt to perform some appropriate startup action.

    * If the user had saved a session document the previous time the
      emulator had been run, the emulator will attempt to reload that
      document.

    * If the user created a new session document during the previous
      session but didn't save that document, the emulator will create
      a document with the same settings.

    * If either of those attempts fails (for instance, the previously
      used session document or ROM file cannot be found), or this is the
      first time the user has used the emulator, or the user holds down
      the Caps Lock key while starting the emulator, the user is
      presented with a "New Configuration" dialog box. The user uses
      this dialog box to specify a new hardware/RAM/ROM configuration
      from which a new document is created.

    If the user double-clicks on an emulator document icon, the emulator
    attempts to restore the session recorded in that file. This attempt
    may fail if the ROM image for the session can no longer be found. In
    that case, the session document can no longer be used; session
    documents can only be used in conjunction with the exact same ROM
    image that created them.

    If the user double-clicks on a ROM image icon, the emulator
    presents the user with the "New Configuration" dialog box with the
    selected ROM file entered into the ROM File text slot.

*    Brought Mac and Windows versions in sync with each other.

*    IMPORTANT NOTE: "RAM" files created by previous versions of the
    emulator are no longer supported. There are two reasons for this.
    First, they didn't contain enough information that today's session
    files need (such as the associated ROM file name, or the device type
    to emulate). Second, ".ram" is the extension for Real Audio Media
    files. Therefore, the emulator now reads and writes ".psf" (Palm
    Session File) files.

*    Added Registry information so that Emulator files have the right
    description next to their names.


Changes for 2.1d1 (6/22/98)
---------------------------
*    Updated to latest UAE.

*    (Mac) Updated to CodeWarrior Pro 3. (Requires "MW C/C++ PPC"
    from May 19th or later.)

*    Removed 68K and FAT versions.

*    New menu organization.

*    More emphasis on documents, rather than ROM image files.

*    Lots of internal changes:

    - RAM_ROM broken out into many smaller files.

    - Lots of small performance tweaks.

    - More strings moved to resources.

    - Names of system functions are fetched directly from the ROM
      instead of being hardcoded in StringData.cpp.

*    Added DragonballEZ support.

*    (Mac) Added white screen.

*    New Preferences/Properties... and Debug Options... dialogs.

*    New startup dialog for Windows (for when it needs help starting
    up).

*    Fixed the bug where you couldn't enter typed characters if there
    wasn't a blinking insertion point.


Changes for 2.0b3 (6/2/98)
--------------------------
*    Windows only: Added Greg's nifty Hyperlink controls to the About box.

*    Added Keith's Palm OS Emulator article from HHSJ.

*    Windows only: Added Greg's fixes to the serial emulation routines.

*    Fixed Headpatch::SysGremlins bug where it would read garbage from
    the stack when trying to get the selector parameter.

*    Got A-Trap breaks working ("atb" from the PalmDebugger).

*    Windows only: bumped number of recently loaded application from
    five to nine.

*    Fixed 1.0 ROM compatibility problem: the emulator was trying to call
    a function introduced in Palm OS 2.0 (PrefSetPreference).


Changes for 2.0b2 (5/25/98)
---------------------------
*    Windows only: Fixed bug that would zap the version number in the
    system preferences when attempting to set the timeout value to be
    infinite.

*    Fixed bug that would cause the emulator try to access invalid memory
    if the emulated app tries to access invalid memory.

*    If an exception occurs that an external debugger should handle and
    there is no external debugger, instead show a dialog that reports
    the problem and forces the user to reset.

*    When getting application version string, look for ID's 1 and 1000.

*    Windows only: Fixed bug that would leave a debugger-related messaging
    event object signalled if there were no external debugger.


Changes for 2.0b1 (5/??/98)
---------------------------
*    Prevent the device from ever going to sleep. Letting it go to sleep
    is only annoying, and doesn't accomplish anything. You can make the
    device go to sleep by setting the timeout value in the General
    Preference panel, but that setting will get reset back to "never" on
    any reset.

*    Mac only: Fixed problem that would cause the emulator to hang if
    you tried to open a console window from within MWDebug.

*    Got alarms working.

*    Mac only: Wasn't doing a "weak import" of AppearanceLib in non-debug
    version.

*    With regards to a change in 2.0d8, it's OK to load an application that
    already exists in ROM. The RAM-based version will override the one
    that's in ROM.


Changes for 2.0d9 (5/13/98)
---------------------------
*    The handling of DbgBreak was broken.

*    Allow application to resume after ErrDisplayFileLineMessage. This
    is handy in cases where you're using ErrNonFatalDisplayIf to display
    warning messages and it's OK for the application to continue
    execution.

*    Accidentally turned off checksumming of post-2.0 ROMs; turned it
    back on.

*    Windows only: there was supposed to be a feature whereby holding
    down the caps-lock key when starting the emulator would cause it to
    bring up a Download/Browse/Quit dialog box. However, the dialog box
    would actually come up if the caps-lock key had ever been toggled
    since the last time the emulator had been executed. Fixed.

*    Selecting "Gremlin/New..." while a gremlin is currently running
    will now switch to the new gremlin.

*    Fixed error message to say RAM instead of ROM when saving a RAM
    image file, and added an exception handler to deal with the ensuing
    thrown exception.


Changes for 2.0d8 (5/27/98)
--------------------------
*    Added Greg's Copilot article from HHSJ.

*    Added updates to Guide.html from Roger.

*    Removed support for the 2.0 Debug ROM that offset the low-memory
    globals byte four bytes. There were problems with that ROM that
    would cause it to crash.

*    Gremlins now latches into KeyCurrentState in order to randomly set
    the bits so that applications can respond to them.

*    The patched versions of SysFatalAlert and ErrDisplayFileLineMsg now
    display dialogs containing "Debug" and "Reset" buttons. Clicking on
    "Debug" will simulate a bus error, which an external debugger will
    catch. Clicking on "Reset" is the same as selecting the Reset menu
    item.

    NOTE: currently, clicking on the "Debug" button when there's no external
    debugger running will result in the ROM attempting to handle the
    simulated bus error. If the ROM is a Debug ROM, it will attempt to
    contact a debugger over the seial port. Since there is probably no
    such debugger listening on the serial port, the emulator will appear
    to hang with a blank screen. Selecting the Reset menu item will
    restart the ROM.

*    Failures reported via SysFatalAlert and ErrDisplayFileLineMsg
    during application execution have more info for users to report.
    The dialog has text along the lines of:

        SubHunt 2.0 has failed while running Gremlin 2 at 20,000 events.
        SubHunt complains "SubHunt.c: Line 100: Score too good." If
        this is the latest version of SubHunt, please report this to the
        application author.

    or:

        SubHunt 2.0 has failed. PalmOS complains "Window.c: Line 100:
        Invalid Window."  If this is the latest version of SubHunt,
        please report this to the application author.

    NOTE: currently no effort is made to determine if the application or
    the PalmOS is generating the message. This shortcoming will be
    addressed in a future version.

*    Use application name from 'tAIN' resource when generating the list
    of applications for the New Gremlin dialog box.

*    Reduced the 10% random pen tap chance in Gremlins to 2% (per
    Roger's recommendation).

*    Internal change: added "TrapIndex" inline function for converting a
    trap number (0xA000) to an index that can be used with the trap
    table. Standardized on its use all over. This function converts by
    merely masking off the upper byte (which mirrors what the ROM does)
    instead of subtracting "sysTrapBase" from trap numbers (which fails
    because some applications appear to use trap numbers that don't have
    "A" in their upper nybble).

*    Internal change: we patch ROM system functions. We also call ROM
    system functions. Now, whenever we call a ROM system function, any
    trap patches made to that function are called (they didn't used to
    be).

*    Fixed emulator/debugger communications bug that occured when one
    would quit and then restart (needed to refresh -- not just flush --
    the handle to the other guy's window). (Windows only)

*    After the UI has been initialized, prevent even the ROM from
    writing to low-memory (except from HwrSleep, which patches out the
    exception vectors).

*    When trying to load an application that's already in ROM, warn with
    a specific error message instead of a generic one.


Changes for 2.0d7 (4/27/98)
---------------------------
*    Gremlins now works better with "ill-formed" forms (forms with no
    "live" objects in them). Gremlins will also now generate random
    taps 10% of the time instead of always trying to target form
    objects.

*    Turn off Gremlins on a reset or fatal system error.

*    Only the last 128K of Gremlins logging text will now be written to
    the logging file.

*    Added "Debug" button to the "Application so-and-so just made an
    invalid memory access" dialog. Clicking on this button emulates a
    bus error. If an external debugger is running, it will be notified
    of this bus error. If not, the emulated ROM will take over,
    displaying a Fatal Error dialog.

*    Fixed a bug where the emulator itself would cause a "The current
    application just accessed Palm OS Globals ..." to appear after
    downloading an application via MWDebug.


Changes for 2.0d6 (4/22/98)
---------------------------
*    Added ROM Transfer project (for Mac).

*    Added Docs folder.

*    Added Plugins folder containing MWDebug plugins.

*    Tweaks to PCH file generation for CW Pro 3.

*    Added HotSync to Mac side.

    IMPORTANT: For both versions of the Emulator, HotSyncing has only
    been tested between two machines (in both cases, between a Mac and a
    PC). I don't know what performance is like when both the emulator
    and HotSync are running on the same machine.

*    Now enable and disable Gremlins menu items in WinEmulator.

*    Fixed bug in ROM Transfer.prc which would cause it to crash on Palm
    III's.

*    Fixed some bugs in RAM file management in Mac version.


Changes for 2.0d5 (4/??/98)
--------------------------
*    Checksum the ROMs when they're loaded. This checksum is used to (a)
    validate the ROM contents (for pre-3.0 ROMs, the checksum is checked
    against a hard-coded list of ROM checksums, for 3.0 ROMs and later,
    the checksum is checked against the checksum in the ROM header) and
    (b) determine if we're running the 2.0 debug ROM, which needs
    special hacks activated in order to run.

*    Allow *.* patterns when loading ROM and application files.

*    Support serial port emulation in WinEmulator (not in Mac, yet).

*    Added method for applications to call the emulator (see
    EmuTraps.h). Note that there currently is no method to see at
    runtime if an application is running under the emulator or not.

*    Tweaks to get Mac version to build under CWP R2 with CW4POS R4
    installed in it.

*    Backed out checksumming of pre-3.0 ROMs until I get get an accurate
    set of checksum numbers.


Changes for 2.0d4 (4/9/98)
--------------------------
*    Reset auto-off timer when processing debugger packets.

*    Changed way tailpatches are implemented. I used to set the return
    address of a system function I was tailpatching to point to a TRAP
    $E opcode (saving the original return address in an array in the
    emulator). This approach has the advantage of not modifying any
    opcodes in memory, and handles recursive system functions very well.
    Sadly, the critical flaw in this approach is that some operations
    need the real return address to be on the stack. For instance, the
    TRAP $F handler (TrapDispatcher) in the ROM looks at the return
    address so that it can look up the $Axxx function selector.
    Therefore, the approach I now take is to write a TRAP $E into memory
    after the TRAP $F / $Axxx sequence, saving the overwritten opcode in
    a table. Recursive functions are handled by incrementing and
    decrementing a counter associated with that memory location. When
    the counter gets back down to zero, the opcode is restored.

*    Added preliminary profiling stuff. Still disabled at this point.

*    Don't let the user enter pen and keyboard events if the CPU is
    stopped (for example, when it's in the debugger).

*    Don't run RPC commands from the debugger in TRACE mode (that is,
    run the commands, but turn trace mode off, first).

*    Respond better to a SysReset message from the debugger.

*    Remove tailpatch opcode when the CPU is reset.

*    In WinEmulator, restart the CPU after a screenshot only if it was
    running before we took the screenshot.

*    Fixed problems with synchronization between the UI thread and the
    CPU thread. One way to trip up over the bug was to run Gremlins and
    at the same time click like crazy on the screen or hardware buttons.
    After a while, you'd get the dreaded "ATrap::DoCall returned and
    exception" dialog box.

*    Fixed Gremlins bug that would put it into an infinite loop if the
    top form had only frmBitmapObj, frmLineObj, frmFrameObj,
    frmRectangleObj, frmLabelObj, frmTitleObj, and/or frmPopupObj
    objects in it. Games like SubHunt, HardBall, and Reptoids are
    examples of programs with such forms.

*    Changed the low-level mechanism for handling exceptions and
    entering the debugger. With the old way, an exception would cause
    the main CPU loop to exit before the debugger was notified about the
    exception. But if there were no external debugger to notify, there
    was no way to cleanly restart the CPU loop in order to have the
    emulated ROM handle the exception (either by trying to connect to a
    serial debugger, or by showing a fatal error dialog). Now, the
    attempt to contact an external debugger is made before exiting the
    CPU loop. If the attempt succeeds, the CPU loop is exited. If the
    attempt fails, the normal exception handling process takes over
    (i.e., the appropriate vector from the low- memory exception table
    is loaded into the PC and execution continues).

*    Added support for 2.0.2 Debug ROMs (added minimal support for 68681
    serial access and account for different low-memory variable
    locations).

*    Removed support for ROM files with the pattern "ROM *". This
    pattern also matches the "ROM Transfer.prc" application, causing the
    emulator to try loading it if there are no other candicates.

*    For MacEmulator, save RAM files with the type '(Pi)RAM' and ROM
    files with the type '(Pi)ROM'. Also, if untyped files are accessed
    as RAM or ROM files, they are changed to have the appropriate type.

*    Restarting and stepping of Gremlins in WinEmulator more reliable.

*    Better warning messages for applications that:

    -    Access low-memory directly
    -    Access system global variables directly
    -    Access the screen directly
    -    Access Dragonball registers directly


Changes for 2.0d3
-----------------
*    Lots of internal changes:

    -    Names changed from Foo_Bar format to Foo::Bar format (where Bar
        is a static member function of the Foo class). This change
        makes it easier to look things up in a class browser.

    -    Handling of memory access errors codified. All accesses are
        checked; any errors are reported through bottleneck functions;
        all accesses can be turned on/off with compile-time and/or
        runtime switches; added CEnableFullAccess class that allows
        internal functions full access to memory (i.e., turns off memory
        access checking) so that we can do things like set breakpoints
        in ROM.

    -    Added Palm copyrights to all files.

    -    Standardized on PowerPlant-style function header comments.

    -    Removed all mention of Pilot and Pilot-derivatives.

    -    Untemplatized some functions in UAE_Utils that didn't need to
        be templates.

    -    Sped up uae_memset by using long-word access.

    -    Changed way tail patches are implemented. Previously, I'd write
        a TRAP $E opcode into memory after the TRAP $F/Axxx instruction
        sequence that would invoke the trap. The problems with this
        approach is that (a) this wouldn't break on all returns from the
        function if it called itself recursively, and (b) it modified
        memory in a way that would need to be reversed on selected
        occassions (e.g., when we entered the debugger, or when RAM was
        saved to disk. Instead, when a trap function is called that I
        wish to tail patch, I modify the return address from the
        function in such a way that I get control when the function
        exits (I save the old return address and replace it with a
        pointer to a TRAP $E opcode, which leads to a CPU exception
        which I catch in Software::ProcessException).

    -    Patched calls to SysDoze and SysSleep so we better know the
        state of the ROM/CPU. We don't actually use this information,
        yet.

    -    Centralized all compile-time pre-processor symbols into
        Switches.h.

    -    Centralized all byteswapping into Byteswapping.h/.cpp. Used a
        templatized Canonical() function on the Mac so that Canonical is
        zero-overhead no matter what type it's called with.

    -    Added internal diagnostic logging functions.

    -    Lots of timing tests to optimize memory access and core CPU
        emulation loop.

    -    Committed to "word-swapped" memory layout in x86-compatible
        systems. Removed experimental alternative memory layout code.
        Look for WORDSWAP_MEMORY symbol for areas that are byte-order
        sensitive.

*    Fixed some bugs with external debugger hooks.

    -    Now support the same notion of "sockets" as the SerialLinkMgr.

    -    On some entries into the debugger, the PC would incorrectly be
        adjusted to point 2 bytes before the instruction causing the
        debugger entry instead of pointing right at the instruction.

    -    Finally added debugger hooks to WinEmulator (missing
        functionality's considered a "bug", right?).

*    Fixed bug where UAE would allocate a significant chunk of memory and
    then try using it without first checking that the allocation attempt
    succeeded. This would result in the Mac emulator freezing on startup
    when emulator a 4Meg device.

*    Fixed ROM Transfer.prc to download Palm III ROMs. Updated Mac
    "Download ROM" command to work with larger ROMs. Added "Download
    ROM" to Windows emulator.

*    On startup, if WinEmulator can't find a ROM, it will ask you to
    find one or download one.

*    On startup, if the capslock key is down, WinEmulator will _always_
    bring up the dialog asking you to find a ROM file or download one.

*    Removed "Backlight" and "Enter debugger on startup" properties from
    WinEmulator.

*    When looking for a ROM file, Emulator use the following search
    order:

    - The file you've explicitly specified in Properties (WinEmulator only).
    - The first file in the Emulator directory ending with ".ROM"
    - The first file in the Emulator directory ending with " ROM"
    - The first file in the Emulator directory starting with "ROM."
    - The first file in the Emulator directory starting with "ROM "

*    Fixed problem where mouse clicking wasn't working with 1.0 ROMs. I'm
    still not sure what the root of the problem is, but I've added a
    check for 1.0 ROMs and implemented a workaround.

*    In WinEmulator, RAM image is automatically saved to and read from a
    file based on the name of the ROM being used and the RAM size
    requested. The name of the RAM file is "<ROM file>.<RAM size>.RAM".
    In other words, the old Windows Copilot functionality is back (it
    was temporarily removed during the Grand Code Unification).

    The format of this RAM image file is a little different, however.
    There are chunks of data in this file, each tagged with an ID and a
    chunk size. For now, the only defined chunk is 'ram '. In the
    future, there will be chunks for register state, associated ROM
    version, etc.

    In MacEmulator, the file saving operation remains the same as in Mac
    Copilot, but the file format is the same as in WinEmulator.

*    Switched to new case graphics.


Changes for 2.0d2
-----------------
*    Fixed MacEmulator bug when switching between 1x and 2x screen
    sizes.

*    Fixed bug where loading an application would smash 4 random bytes
    of memory.

*    Sped up emulation on the Mac by 20% by streamlining some
    Mac-specific operations.

*    Sped up core emulation loop on both platforms by moving some
    not-oft-used functionality into some subroutines, thus simplifying
    the core loop.

*    Added hooks for external debugger.

*    Made Gremlins deterministic.

*    Ensured Gremlins did the same thing on both Mac and Windows.


Changes for 2.0d1
-----------------
*    Support for Palm OS 3.0:

    -    memChunk headers changed, so reliance on their format when
        loading applications has been removed.

    -    More complete support for Dragonball registers.

    -    Allow for 96K and 128K dynamic heaps.

*    Common source code base.

*    Refitted with latest UAE sources.

*    Can load .prc files larger than 64K.

*    Windows version is much smaller.

*    Mac version is much bigger (it's a FAT binary).

*    Removed built-in debugger.

*    Support grayscale mode.

*    Now emulates hardware method for turning on backlighting (holding
    down the power key). Properties option is no longer needed.

*    Added screen shot command.

*    Added Gremlins. (Note that user input (typing, mousing) is disabled
    while Gremlins is running due to the incompatible ways in which they
    enter their events.)

*    Set the Palm OS date on startup.

*    Trap dispatcher is now run as native code instead of being
    emulated. This results in about a 10 - 12% speedup.

*    Auto-calibrate the pen on startup; no more digitizer screen.

*    InternetConfig-savvy About box on the Mac. Panel-savvy About box on
    Windows.

*    Gremlins can now confine itself to a single application.

*    Emulate MemSet, MemMove, and RctPtInRectangle with native code for
    15% performance increase.

*    While Gremlins is running, can optionally log to a text file:

    -    posted events
    -    received events
    -    system functions called
    -    time of execution.