 
 History HDPMI 

  16.11.2009, version 3.17

   bugfix: when int 33h, ax=000Ch or ax=0014h was used by a client,
    HDPMI might have corrupted DOS memory on exit.
   Linker switched to jwlink (modified wlink).
  
  20.01.2009, version 3.16

   Assembler switched to JWasm.
  
  01.03.2008, version 3.15
  
   bugfix: int 31h, ax=507h didn't return number of processed pages
    in ECX if function failed.
   bugfix: an exception 11h might have occured inside hdpmi's cpu test
    if CR0 AM bit was set on init.
   bugfix: translation of int 21h, ax=7305h didn't set realmode-DS
    to TLB, making the call fail under certain conditions.
   int 31h, ax=507h now returns error code 8013h if functions fails 
    because of lacking free physical memory.
   XMS memory allocation strategy slightly changed to avoid running out
    of XMS handles if client allocates just small portions and number
    of XMS handles is limited to 32.
   kernel debugger support split into WDEB386 and 386SWAT. Int 22h (Win386)
    protected-mode debug "API" disabled.
   Exception 11h (alignment fault) is now supported.
   HiWords of general purpose registers preserved on initialization.
   CR0 MP and EM bits are now cleared for the first client only. This
    allows to implement an FP emulator DPMI TSR.
  
  11.01.2008, version 3.14

   IOPL0 version only: added support for string IO.
   group names changed to allow linking HDPMI object modules into other
    binaries.
   "stub" version of HDPMI (HDPMI32.BIN) added.
   bugfix: translation of int 15h, ax=C207h (set PS/2 mouse event proc),
    has caused a GPF if called with ES=0.
   bugfix: if client used Int 33h translations, set a mouse event proc
    and HDPMI wasn't installed residently, a memory corruption may have
    occured on host exit.
   bugfix: int 31h, ax=506h didn't work correctly for mapped pages (int
    31h, ax=508h/509h). It returned them as "committed".
   bugfix: a missing "assume" caused a destructive write to memory in
    HDPMI's conventional memory portion (XMS mode).

  15.07.2007, version 3.13

   bugfix: the EMS handle allocated in v3.12 might have been released 
    multiple times if HDPMI was installed with option -a or HDPMI=32 was
    set. This could have affected other applications using EMS.
   bugfix: the EMS handle allocated in v3.12 wasn't released if host
    ran out of memory on init.
   bugfix [HDPMI16 only]: translation for int 21h, ax=7160h used ES:EDI
    as target for translated name. Corrected to use ES:DI only.
   bugfix: commandline option -m didn't work in v3.07-v3.12.
   real-mode callbacks are disabled and mouse is reset before requesting
    key for client/server termination in the default exception handler.
   options -n, -k implemented.
   option -g implemented in source (currently deactivated) to prohibit
    usage of INVLPG opcode.
   an additional 4k page is allocated on startup, which avoids a possible
    GPF if free extended memory on init is small (<= 3 MB).

  15.03.2007, version 3.12

   bugfix (HDPMI16 mainly): if int 31h, ax=0100h succeeded to alloc DOS 
    memory, but no more LDT descriptors were available, the call returned with
    Carry set, but the DOS memory block wasn't released.
   bugfix (HDPMI16 only): if int 31h, ax=102h failed because of insufficient
    descriptors, it didn't reset the memory block's size to its previous
    value.
   bugfix (HDPMI16 only): int 21h, ah=48h, BX=-1 did return a valid memory
    block. Now it fails in any case, returning free memory (max. FFFEh) in
    register BX (similar to Win9x' DPMI host).
   bugfix: if the XMS hook chain did contain a hooker which doesn't
    start with 0EAh/0EBh, HDPMI has caused a crash.
   bugfix: error condition DOS < v4 caused a garbage display.
   allocated size of dynamic TLB was still 0FFF0h, though 0FE00h bytes were
    used only.
   on FATAL exceptions now value of CS:EIP is additionally displayed.
   before calling XMS "query free memory" register BL is set to zero.
    Makes HDPMI work better with some (old) XMS hosts.
   minimal DOS version required to run HDPMI is now v3.3.
   if a EMM/VCPI host is detected, an (empty) EMS handle is allocated on
    startup to ensure the EMM host keeps ON and installed.
   the host now uses the DOS SDA (swappable data area) if it needs to know
    the current PSP segment for internal reasons.

  14.12.2006, version 3.11

   bugfix: if a client allocated very large uncommitted memory blocks,
    HDPMI may have grabbed (almost) all XMS memory.
   bugfix: on init option -i may have caused a stack exception if HDPMI was 
    running in VCPI mode.
   bugfix: int 21h, ah=4Ah first increased memory block and then tried
    to alloc additional descriptors (16-bit only). If there were not enough
    descriptors, the function failed, but the memory block kept its new
    size, thus wasting memory.
   size of dynamic TLB reduced from 0FFF0h to 0FE00h, which increases
    speed of large file read/write operations in FreeDOS/DR-DOS.

  15.10.2006, version 3.10

   bugfix: the routine which checked whether an XMS memory handle could be 
    released (relevant only if HDPMI is installed with -r) may have caused
    a GPF.
   bugfix: translation for int 21h, ax=6300h checked carry flag if dos call
    was successful, but this call returns AL=00 in case of success instead.
   bugfix: if Ctrl-Alt-Del is detected, there is no longer an int 21h,
    ax=4CFFh issued in protected-mode if current PSP is not the one of the
    active client.
   bugfix: int 4Bh, ax=8102h (get VDS version) was not routed to real-mode,
    but since it uses no pointers it should be routed.
   the VDS functions which expect a pointer to a DDS in ES:E/DI (03-04, 
    07-0A) are now  translated and routed to v86-mode.
   VDS functions 05h and 06h are handled by HDPMI.
   int 21, ah=4Ah translation: if memory block to resize is in conventional
    memory, now the error code received from real-mode dos is returned.
    Previously it always returned ax=8 and available paragraphs in E/BX.
   when installed permanently with -r HDPMI now grabs less XMS memory.
   Int 31h, ax=0000: if the request has no chance to be successful (too
    many descriptors requested), size of LDT now remains unchanged.
   new value for environment variable HDPMI:
    32768 - prevent HDPMI from touching CR0 NE bit.

  14.09.2006, version 3.09
  
   bugfix: if DOS int 21h, ax=440Dh did not require translation, it may
    have returned with Carry set, but in fact did not call DOS at all.
   bugfix: translation for int 26h, cx=ffffh (FAT16 absolute disk write)
    may have caused data corruption on the drive, translation for int 21h,
    ax=7305h, si[0]=1 (FAT32 absolute disk write) failed in any case.
   when running under a VCPI host, bit 1 in host flags (int 31h, ax=0400h)
    returned in BX will now be cleared.

  15.08.2006, version 3.08

   mode switches now slightly faster when running as VCPI client.
   bugfix: exceptions during initialization while no client is running yet
    (inside a VCPI memory function for example) weren't handled properly,
    which made the system unstable and required a reboot.
   bugfix: option -a used on a system with VCPI but no XMS may have caused
    troubles on nested execution of clients.
   bugfix: the default exception handler displayed garbage as client's
    CS:EIP and SS:ESP on ring 0 exceptions if there was no client running
    (host idle or initializing).
   HDPMI v3.07 was not compatible with VMware, because VMWare seems to not
    properly emulate the 80386 paging TLB. Workaround implemented.
   cmdline options -v and -y added.
   in raw mode port 92h ("fast" A20 gate) is no longer touched if BIOS
    is telling that this method is not available.

  21.07.2006, version 3.07a

   bugfix: environment variable HDPMI=512 didn't work in v3.07.

  15.07.2006, version 3.07

   bugfix: if the host run more than one client, it may have modified
    the real-mode stack of previous client during client termination.
   optionally run each client in its own address context
   meaning of value 32 for environment variable HDPMI changed. The previous
    meaning is now achieved by setting HDPMI=16384 (which is rarely used).
   new cmdline options -i, -s and -v.
   the HDPMI binary now is linked so that the DOS loader will load the
    real-mode part only. The protected-mode part is loaded delayed, which
    allows to load HDPMI in much smaller UMBs.
   HDPMI will now use the TLB of any instance of the host found in
    memory, enabled or disabled.
   additional security check if a parent psp has to be translated back
    to a segment.

  14.06.2006, version 3.06

   new value for environment variable HDPMI:
    8192 - break into client after initial switch to protected mode.

  15.05.2006, version 3.05

   bugfix: int 31h, ax=0101h may have caused an exception if selector of  
    memory block to free was contained in DS.
   bugfix: in previous versions the real-mode DTA was set permanently to the
    hosts's PSP+80h. This apparently was a problem for some clients not using
    the host's DOS translation services (F-PROT.EXE).
   bugfix: if int 31h, ax=101h|102h did not fail it still returned with a
    modified AX register
   bugfix: int 21h, AH=4Ah returned with modified AX register.
   bugfix: int 31h, AX=050Bh returned 0 in "maximum locked bytes" field. Now
    the "free" bytes are returned.
   bugfix: int 31h, ax=0507h didn't work with ECX=0 in versions 3.02-3.04
    (32RTM.EXE does such a call and expects it to succeed)
   bugfix: cmdline option -u didn't work if environment variable HDPMI=1
    was set and HDPMI run as VCPI client.
   bugfix: raw mode switch to protected mode may have enabled interrupts
    one instruction too early, which caused a GPF if a IRQ occured.
   CR4 now displayed by default exception handler
   default exception handler no longer makes assumptions about value of SS
    if an exception occured in ring 0.
   some free entries in GDT added which is required by 386SWAT to debug
    VCPI clients.

  02.05.2006, version 3.04

   bugfix: in Int15/raw mode only part of physical memory was used
   bugfix: HDPMI installed with option -r and DOS memory included in
    page pool (HDPMI=2) didn't work reliably.
   bugfix: translation for Int 33h, ax=0016h (save driver state) didn't
    work.
   bugfix: freeing a real-mode callback which was already free didn't
    set the carry flag.
   bugfix: total physical memory reported by Int 31h, ax=0500h was too low
    after more than 1 XMS memory block was allocated.
   bugfix: free pages reported  by Int 31h, ax=0500h now are the true
    free physical pages. Previously it was just the same value as in field
    "max unlocked page allocation". 
   XMS memory handles (except the first one) now released if a resident
    installed HDPMI enters idle state (no client running).
   CR0 no longer saved/restored when a client starts/exits, just the
    FPU bits are maintained on the client level (to support Int 31h, 
    ax=0E0xh). CR0 MP bit is no longer reset on initialization. 
    Meaning of cmdLine option -t changed to "don't touch NE bit", 
    previously it was "do reset NE bit".
   in Int 31h, ax=0500h VCPI host was called if no free XMS pages exist
    anymore. This is a problem with at least one popular VCPI host and
    should therefore be avoided.
   now flags are saved/restored in Int 15h/2Fh real-mode hook procs
   new value for environment variable HDPMI:
    4096: make pages for GDT, IDT, LDT readonly

  21.04.2006, version 3.03

   bugfix: in V3.02 API translation Int 21h ah=29h didn't work for 16-bit
    clients.
   bugfix: in V3.02 routing interrupts 78h - FFh to real-mode didn't work
   bugfix: client initialization errors (out of memory or selectors) 
    may have left the system unstable.
   bugfix: low memory conditions on host initialization may have left the
    system unstable.
   in previous versions HDPMI needed some pages to be temporarily mapped
    in page table 0 on startup, which may cause a problem in VCPI mode
    if there are no free PTEs left. Now this temporary mapping is no longer 
    needed.
   command line option -b added to further reduce DOS memory usage.
   missing LFN translations (ax=71A7h, 71A8h and 71AAh) added.
   DPMI v1.0 functions to get/set multiple descriptors (int 31h, 
    ax=000Eh/000Fh) implemented
   DPMI v1.0 functions to get/set exception handlers for protected-mode
    (int 31h, ax=0210h/0212h) implemented
   DPMI v1.0 "exception restartability" supported.

  12.04.2006, version 3.02

   bugfix: internal function to alloc XMS memory was unable to detect
    a XMS host failure in V3.01.
   bugfix: change in V1.87 caused the critical error interrupt (INT 24h)
    to not work reliably.
   bugfix: resize linear memory block (Int 31h, ax=0505h) did always
    commit the added pages (ignored bit 0 of edx).
   GDT now moved to extended memory (unless switch HDPMI=512 is set).
   rest of missing DPMI v1.0 memory functions implemented:
    ax=505h: segment descriptor update now supported
    ax=508h: map device in memory block
    ax=509h: map conventional memory in memory block
    ax=50Ah: get memory block size and base
    ax=50Bh: get memory information
   API translation Int 33h, ax=0016h and ax=0017h wasn't documented.
   new value for environment variable HDPMI:
    2048: clear hiword of ESI/EDI on client's initial switch to protected
          mode.

  07.04.2006, version 3.01

   bugfix: int 21h, ah=4F (find next file) in v3.0 may have caused
    data corruption.
   some more data moved into extended memory to free conventional DOS
    memory.
   new value for environment variable HDPMI:
    1024: disable support for DPMI v1.0 memory functions

  04.04.2006, version 3.0

   HDPMI32 and HDPMI16 will share a TLB, which reduces conventional
    memory consumption by 8 kB if both hosts are loaded residently.
   host code running in protected mode moved to extended memory.
    This reduces HDPMI's conventional DOS memory usage from about
    32 kB to 14-15 kB (including a 8 kB TLB).
   host protected-mode code now runs in a 32-bit code segment.
   use VCPI memory allocation if no XMS host found.
   new values for environment variable HDPMI:
    128: disable LFN API translation
    256: enable HMA mapping in protected mode if HDPMI runs as VCPI client
    512: allocate IDT and LDT in client address space (SBPCI problem)

  21.03.2006, version 2.07

   bugfix: internal function to call a real-mode interrupt cleared
    the carry flag before calling the INT. This caused troubles with
    LFN functions on MS-DOS < 7 and DR-DOS.
   bugfix: trying to terminate a client if Ctrl-Alt-Del was pressed
    didn't work for HDPMI16.

  20.03.2006, version 2.06

   Exception 05 is no longer routed to real-mode. It is still routed
    to protected mode Int 05, though. To real-mode is will only be routed
    if it is a programmed INT 05, else the client will be terminated.
   bugfix: default exception handler didn't display a register dump.
   small code size reduction.

  18.03.2006, version 2.05

   bugfix: strings displayed with int 21h, ah=9 weren't checked if their
    size exceeds size of TLB.
   debug exceptions now detected by examining DR6, not by content
    of cs:eip or trace flag.
   added a test in raw mode if size of extended memory between 1 MB and 16 MB
    as returned by Int 15h, ax=e801, does not exceed 15360. Else this function
    is regarded as 'not implemented' (required for DosBox in raw mode).
   test if vector is 0000:0000 before calling Int 68h in real-mode.
    This makes HDPMI compatible with DosBox (tested with DosBox V0.63).
   bugfix: int 31h, ax=090xh didn't work in HDPMI16.
   int 31h entry in IDT now directly jumps into the host's int 31h
    handler, which improves this int's speed (especially useful for virtual
    interrupt functions, which are now 4 times faster than before). 

  06.02.2006, version 2.04

   HDPMI16 and HDPMI32 now both may be installed resident concurrently.
   ES, FS and GS selectors caches now updated after int 31h, ax=000Ch
    has modified a descriptor.

  22.01.2006, version 2.03

   check if VCPI host has remapped PICs. Exit with error 6 if this
    is true.
   supply some extra bytes of stack space to make HDPMI run
    with QEMM V8+V9 VCPI hosts.

  29.12.2005, version 2.02 

   bugfix: freeing memory (int 31h, ax=502h) or setting page
    attributes (int 31h, ax=506h) did not invalidate the TLB
    entries. This may have caused problems on some conditions.

  18.11.2005, version 2.01

   small size reduction

  30.10.2005, version 2.00

   default exception handler now sends an EOI to slave PIC if
    any interrupt is in service.

  15.10.2005, version 1.99

   default exception handler now sends an EOI to master PIC if
    any interrupt is in service.

  27.09.2005, version 1.98

   functions int 31h, ax=506h and ax=507h now accept ecx=0
    (number of pages) to make 32rtm.exe not complain about
    insufficient extended memory.
   bugfix: versions 1.89-1.97 of HDPMI32.EXE may have caused
    a stack exception on 80386 + 80486 cpus when client has
    just switched to protected mode. HDPMI32 used a 16-bit
    IRET to return to the client, but this may have trashed
    hiword of ESP.

  16.09.2005, version 1.97

   if HDPMI is resident real-mode vectors are now restored after last client
    has terminated. Previously it was done when HDPMI was uninstalled.
   new switch HDPMI=64 implemented.
   bugfix: Unsupp\HDPMI16.EXE had a bug in int 2Fh, ax=1684h
    making it unable to run windows 3.1.
   bugfix: if int 31h, ax=504h, ebx=0 failed it returned with modified EBX.
   bugfix: using DOS memory for the page pool (HDPMI=2) didn't
    take into account a bug in many DOS versions where a failed
    call to resize a memory block made this block as large as possible.
   bugfix: allocating linear memory with int 31h, ax=504h, ebx!=0
    may have failed, but Carry flag wasn't set. This occured if a
    memory block was allocated previously with a higher linear address.

  06.09.2005, version 1.96

   exc 09 checks disabled if cpu is 80486+.

  18.08.2005, version 1.95

   flag for RMS usage implemented. This allows in raw jump to
    protected mode to not touch current RMS if it is not in use.
    (the win9x dpmi host has problems with this constellation).

  13.08.2005, version 1.94

   bugfix: a mode switch may have changed values of the FPU emulation bits
    in CR0. Now all bits of CR0 except PE and PG are copied between modes.
   translation for int 21h, ah=69h added. Previously calling this function
    failed with an error message displayed.

  30.07.2005, version 1.93

   mov CRx,reg now emulated. This feature (mov CR0, reg) is required by
    some clients (MS C v7.0 [16-bit]).
   XMS handles are now released after HDPMI has finally switched to
    real-mode. The previous behavior seemed to cause problems with
    Bochs/Qemu running HDPMI.
   documentation of HDPMI error codes 8 and 9 was wrong
   new switch HDPMI=32 implemented
   bugfix: environment switch HDPMI=16 didn't work
   owner psp of dyn TLB memory block set to psp of current client. This
    may avoid memory leaks if Ctrl-C is pressed while reading the console
    with int 21h, ah=3Fh.
   previous versions of the 16-bit host HDPMI16 used 286-interrupt gates
    to switch from ring-3 to ring-0 protected mode, thus loosing HIWORD of
    the client's ESP. Now such transfers are done thru 386-interrupt gates
    and the client's ESP remains untouched. This finally makes IR41.DLL 
    (16-bit Intel Indeo video codec for Windows 3.1) work with HDPMI16.

  20.07.2005, version 1.92

   bugfix: HDPMI may have frozen the machine on initialization
    errors (low memory)
   bugfix: in raw mode switch 'real-mode to protected mode' now the
    real-mode ss:sp is stored as current real-mode stack, because
    the previous stack cannot be used anymore, it is "busy".
   bugfix: int 31h, ax=0401h didn't return the correct host
    minor version number.
   bugfix: translation of int 25h/26 (absolute disk read/write)
    may have overwritten 2 bytes in the host's code segment.
   bugfix: dpmi functions int 31h, ax=0301h/0302h didn't work
    in version 1.91 if stack parameters had to be copied.
   new environment flag HDPMI=16, which prevents HDPMI from
    using a 64 kB dynamic TLB.

  15.07.2005, version 1.91

   bugfix: when shutting down HDPMI marked PTEs as notpresent
    which were owned by VCPI host. This caused 386MAX and FreeDOS
    EMM386 to crash.
   bugfix: uninstalling a resident instance of HDPMI (option -u)
    may have overwritten dos memory due to the usage of a real-mode
    stack which was no longer valid.

  10.07.2005, version 1.90

   bugfix: now *all* files are closed before going resident.
    Previously files 0-2 were left open, which may have been a
    problem if they have been redirected.
   int 21h API translator didn't set Carry flag before calling
    real-mode dos int 21h (HDPMI32 with LFN support only). This
    caused problems with DR-DOS.

  04.07.2005, version 1.89

   code modified to use a static task state buffer, which
    simplified the raw mode switching code.
   bugfix: raw mode switch to real-mode didn't clear FS and GS
   display of open real-mode callbacks in default exception
    handler rearranged.
   bugfix: int 31h, ax=0300h, 0301h, 0302h didn't save/restore
    client's real-mode segments. So the values in the RMCS became
    the client's new real-mode segments.
   standard real-mode callbacks (used for IRQs, int 1Ch, 23h, 24h)
    did save real-mode segment registers on the real-mode stack, but
    didn't update the client state. So if the dpmi client called
    a real-mode proc from inside the callback, it didn't get
    the true client real-mode segments. Don't know if this could
    be a problem, but other hosts don't behave this way, so it
    has been changed.
   some small changes to reduce size of binary

  22.06.2005, version 1.88

   bugfix: initialization errors caused garbage to be displayed
   bugfix: making HDPMI stay resident using both options -R + -L
    was hazardous
   default exception handler code partly rewritten so it becomes
    readable again.

  12.06.2005, version 1.87

   CR0 NE bit now is set as default. Makes command line option -n
    superfluous. New command line option -t will reset CR0 NE bit
    instead.
   execute FNINIT in default exception 10h handler. 
   bugfix: an int 10h handler may have been called as if an
    exception 10h had occured, that is, onto the locked stack.
              
  17.05.2005, version 1.86

   "mov reg,crX" is fully emulated now (opcode 0F 20 xx), except reg == ESP.              
   clear TS bit in CR0 on startup.
   translation of int 21h, ah=38h now documented

  14.05.2005, version 1.85

   bugfix: default exception handler may have ignored key strokes
   emulation of "mov eax,cr0" and "mov eax,cr2" reactivated

  13.12.2004, version 1.84

   translation for int 21h, ax=71A0h added
   exception 07 default handler now routes exception to protected
    mode int 07. It is not routed to real-mode, though.
   command line switches -d and -e added
   command line switch -l (=enviroment switch HDPMI=8) added
   dos memory functions now return error codes 8011h and 8022h
   dos memory alloc for 16-bit: selector tiling modified to
    exactly match dpmi specs
   internal: task state restored before server terminates

  01.12.2004, version 1.83

   switched to Digital Mars C++ linker to reduce size of HDPMI
   DR7 now cleared on client termination as specified in DPMI V1.0
   bugfix: special registers DR7 and CR0 are no longer accessed in
    V86 mode when HDPMI terminates (VCPI mode only).
   bugfix: a GPF occured if client exited int 23h with RETF
    instead of RETF 4
   CR0 MP flag now saved and restored as are EM + NE
   command line switch -n added
   locked protected mode stack selector now in LDT. Required
    by WIN87EM.DLL of windows 3.1 which tries to create a code
    alias for LPMS selector (found in Netscape 4.08).
   bugfix: int 15h mouse event proc had wrong stack parameters
   bugfix: int 33h mouse event proc caused GPF on RETF

  24.11.2004, version 1.82

   bugfix: TF was lost when save/restore task state was called in real mode
   translation for int 21h, ax=6521h, 6522h, 65A1h, 65A2h added
   translation for int 21h, ax=5D0Ah added (set ext error info)
   bugfix: exc 0D with EIP > limit(CS) caused an exc 0D in ring 0

  07.11.2004, version 1.81

   bugfix: int 21h, ax=440Dh, minor codes 40h,41h,61h did use
    ds:dx for 32bit clients, not ds:edx

  10.10.2004, version 1.80

   bugfix: address space allocation may not have worked in all
    cases if HDPMI was installed with option -r (resident)
   bugfix: pressing Ctrl-C during console read may have caused
    a fatal dos memory error (asw.exe).

  02.10.2004, version 1.79

   bugfix: function 0503h may have modified ES register

  25.09.2004, version 1.78

   new command line switch -m to disable DPMI 1.0 memory functions
   function 401h, AX now has WRITE-PROTECT CLIENT bit set.
   bugfix: function 0401h modified E/DI.

  17.09.2004, version 1.77

   API translation for int 21h, ax=71A6 added (LFN version only)
   bugfix: there was still a problem with releasing XMS memory,
    though it worked better than in v1.75

  12.09.2004, version 1.76

   bugfix: HDPMI v1.75 crashed on exit if more than 1 XMS handle
    had to be released. 

  11.09.2004, version 1.75

   bugfix: int 23h/24h traps did work for 1. client only 
   command line switches -r and -u added, to install/uninstall
    HDPMI as a TSR.
   bugfix: int 31h, ax=507h may have caused memory leaks
   DOS API translation AX=6300 added (get DBCS table)

  29.08.2004, version 1.74

   unsupported version with LFN API translation released

  10.06.2004, version 1.73

   client CS:EIP stored on top of locked protected-mode stack
   HLT now executed, not simulated

  05.06.2004, version 1.72

   bugfix: int 21h, ah=49h: if memory to free was in conv. memory,
    ES may have been released even if dos function failed.
   function 0503h no longer fails if a block which contains
    uncommitted pages has to be moved
   bugfix: function 0503h (resize memory) has lost HIWORD
    of EBX in some cases.
   bugfix: exception 01 handler wasn't called if TF wasn't set,
    i.e. cleared by an IRET. INT 01 handler was called in any case.
   bugfix: redirection IRQs from real-mode to protected-mode
    didn't work properly
   bugfix: dos api translation AH=65h, AL>0 now works
   terminating client after an exception now uses LPMS
   waiting for a keystroke after exception dump now works better
   bugfix: failure of phys address mapping didn't set carry flag

  12.05.2004, version 1.71

   bugfix: HDPMI v1.70 didn't work with VCPI!
   bugfix: TLB cache is now cleared after system tables have
    been moved high. Because of this bug HDPMI versions 1.62-1.70 
    may not run in non-VCPI modes on 80486 and 80586 cpus.
   (re)set real-mode DTA when client calls int 21h, ah=1ah

  10.05.2004, version 1.70

   bugfix: reset DTA after int 21h, ax=4b00h
   bugfix: int 31h, ax=0002 sometimes didn't recognize
    an already returned selector for a real-mode segment
   bugfix: environment for 16-bit clients in int 21h, ax=4b00h
    exec parameter block now used if != 0
   count of raw switches deleted
   if trace flag set on raw mode switch to protected mode,
    now halt on first instruction, not the second one
   bugfix: when a client other than the first one terminated,
    int 21h, ah=4ch was called with host stack, which could cause
    problems in some circumstances.
   API translation support for int 33h, ax=0009h,ax=0016h,ax=0017h
   bugfix: restrict DOS AH=0Ch API translation to subfunction 0Ah 
   bugfix: tool DPMIRS16 didn't load HDPMI16.EXE 
   switch HDPMI=4 added (return V1.0 server version)

  10.04.2004, version 1.67

   now int 15h in real mode is modified in raw mode only
   function 0505h (resize linear memory block) added
   set CR4 OSXMM flag (200h) on server entry

  18.03.2004, version 1.66

   understands vendor "VIRTUAL SUPPORT" for 32RTM.EXE support
   emulate 'mov eax,crx' and 'mov crx, eax'
   bugfix: int 31, ah=507h was unable to commit a page
   bugfix: int 2f, ax=168ah, vendor MS-DOS, returned function
    address for 32bit didnt clear HIWORD(edi)
   bugfix: int 31, ax=506h didnt return read/write attribute
   bugfix: int 21, ah=00h wrongly expected current PSP in bx

  29.02.2004, version 1.65

   bugfix: int 31, ax=604h returned a wrong page size

  24.01.2004, version 1.64
 
   int 21, ah=71h (LFN) no longer routed to DOS if LFN translation
    service deactivated (which is default)
   bugfix: linear address returned by function 0800h had always bits
    0-11 zeroed
   functions 0600h/0601h now return C if memory region is invalid
    (not allocated) 

  05.01.2004, version 1.63

   in psp kill function (int 21, ah=0) the psb isn't touched
    any more after real mode dos call returned
   translation support for long filenames (AH=71h), optional
   bug fixed for translation of int 21h, AX=7305h

  29.12.2003, version 1.62

   now pagedir in sysarea 0, user addr space now 4088 MB
   bugfixes for AllocSpecAddrSpace + FreeMem
   page tables now mapped in last page table, thus
    increasing user address space to 4084 MB (really)
   DPMI 1.0 function 0801h (unmap phys region) impl.
    max mem block in func 0500 now limited to largest free
    addr space region.

  21.12.2003, version 1.61

   DPMI 1.0 functions 0401h (get capabilities), 0504h
    (uncommitted memory) and 0507h (set page attributes)
    implemented.
   HDPMI environment variable read before real mode initialization

  11.12.2003, version 1.56

   release part of int15 memory if an app is launched.
    this allows multiple instances of HDPMI in int15 mode.
   HDPMI properly terminates if initialization fails (low memory)
   As default DOS memory is no longer in memory pool
   no more FFFF pages limit for memory allocation
   check es, fs, gs if a selector is freed by int31, ax=1
   pressing ctrl-alt-del now executes int21, ax=4c00
   exceptions in ring0 will now stop in kernel debugger
   DPMI 1.0 function 0506h (get page attributes) impl.
   int 3 in ring 0 is ignored now
   bugfix: scan for HDPMI looped if another var beginning
    with H occured in environment
   now using current environment block for int 21h, 4b00h
   total/free address space now handled correctly.
    no crash anymore if server runs out of address space
   bugfix: map physical memory was limited to FFFF pages,
    but returned no error otherwise.

  30.11.2003, version 1.55

   use int 15h, ax=e801h to get extended memory > 64 MB

  27.11.2003, version 1.54

   new switch HDPMI=4 restricts reporting of available
    memory to max. block size for I31, ax=0500h
   bugfix: a 8 kB memory block wasn't released if a task
    other than the first one has terminated.
   bugfix: for exceptions in ring 0 pressing (c)lient didn't work

  24.11.2003, version 1.53

   kernel debugger is now called for ring 0 exceptions as well

  20.05.2003, version 1.52

   kernel debug support wasn't activated in release version

  17.05.2002, version 1.51

   bugfix: reallocate memory (I31, AX=502h) didn't work in all cases
   memory handle is equal to base address for memory allocations
   system area moved to ff800000h, leaving 4084 MB for client
   some enhancements for kernel debugger support
   environment variable HDPMI implemented
   bugfix: wrong query for XMS host 3.0
    bugfix: memory request for 0 bytes will fail now
   i31swt: reset trace flag (call rm int)
   bug fixed in a20 server (xms gate 20 status)

  28.08.1999, version 1.50

   support of FAT32 API translation (Int21,AX=73xx)

  08.01.1998, version 1.40

   support for kernel debugger
   server now is in .EXE fromat (formerly .OVL)
   INT 96h in real mode to prevent debuggers stepping
    into server code.
   guard page 0 for IVT modifications
   temporary 64kb TLB for int 21 3F/40
   save/restore RMS now handled correctly

  14.05.1996, version 1.30

   reset NT-Flag in raw mode switch
   reset NT-Flag in real mode callbacks
   fixed bug in stack manipulation
   real mode int21 is saved and restored on exit
   for exception exit send EOI to kbd if necessary
   restore PM register for real-mode callbacks
   recognize 586 cpu
   recognize 486 cpu

  06.11.1994, version 1.20

   IRQs occuring in real mode now routed to protected mode

  06.06.1994, version 1.10

   many changes, clients now run in ring 3

  01.10.1993, version 1.00

   clients running in ring 0


