Node:System, Next:TFDD, Previous:StringUtils, Up:GPC Units
The following listing contains the interface of the System unit.
This unit contains only BP's more exotic routines which are not recommended to be used in new programs. Most of their functionality can be achieved by more standard means already.
Note: MemAvail and MaxAvail, provided in this
unit, cannot easily be achieved by other means. However, it is not
recommended to use them on any multi-tasking system at all, where
memory is a shared resource. The notes in the unit give some hints
about how to avoid using them.
On special request, i.e., by defining the conditionals
__BP_TYPE_SIZES__, __BP_RANDOM__ and/or
__BP_PARAMSTR_0__, the unit also provides BP compatible
integer type sizes, a 100% BP compatible pseudo random number
generator and/or BP compatible ParamStr (0) behaviour (the
latter, however, only on some systems).
{ BP and partly Delphi compatible System unit for GPC
This unit is released as part of the GNU Pascal project. It
implements some rather exotic BP and Delphi compatibility
features. Even many BP and Delphi programs don't need them, but
they're here for maximum compatibility. Most of BP's and Delphi's
System units' features are built into the compiler or the RTS.
Note: The things in this unit are really exotic. If you haven't
used BP or Delphi before, you don't want to look at this unit. :-)
This unit depends on the conditional defines __BP_TYPE_SIZES__,
__BP_RANDOM__ and __BP_PARAMSTR_0__.
If __BP_TYPE_SIZES__ is defined (with the -D__BP_TYPE_SIZES__
option), the integer data types will be redefined to the sizes
they have in BP or Delphi. Note that this might cause problems,
e.g. when passing var parameters of integer types between units
that do and don't use System. However, of the BP compatibility
units, only Dos and WinDos use such parameters, and they have been
taken care of so they work.
If __BP_RANDOM__ is defined (-D__BP_RANDOM__), this unit will
provide an exactly BP compatible pseudo random number generator.
In particular, the range for integer randoms will be truncated to
16 bits like in BP. The RandSeed variable is provided, and if it's
set to the same value as BP's RandSeed, it produces exactly the
same sequence of pseudo random numbers that BP's pseudo random
number generator does (whoever might need this ... ;-). Even the
Randomize function will behave exactly like in BP. However, this
will not be noted unless one explicitly tests for it.
If __BP_PARAMSTR_0__ is defined (-D__BP_PARAMSTR_0__), this
unit will change the value of ParamStr (0) to that of
ExecutablePath, overwriting the value actually passed by the
caller, to imitate BP's/Dos's behaviour. However *note*: On most
systems, ExecutablePath is *not* guaranteed to return the full
path, so defining this symbol doesn't change anything. In general,
you *cannot* expect to find the full executable path, so better
don't even try it, or your program will (at best) run on some
systems. For most cases where BP programs access their own
executable, there are cleaner alternatives available.
Copyright (C) 1998-2003 Free Software Foundation, Inc.
Authors: Peter Gerwinski <peter@gerwinski.de>
Prof. Abimbola A. Olowofoyeku <African_Chief@bigfoot.com>
Frank Heckenbach <frank@pascal.gnu.de>
Dominik Freche <dominik.freche@gmx.net>
This file is part of GNU Pascal.
GNU Pascal is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published
by the Free Software Foundation; either version 2, or (at your
option) any later version.
GNU Pascal is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with GNU Pascal; see the file COPYING. If not, write to the
Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
02111-1307, USA.
As a special exception, if you link this file with files compiled
with a GNU compiler to produce an executable, this does not cause
the resulting executable to be covered by the GNU General Public
License. This exception does not however invalidate any other
reasons why the executable file might be covered by the GNU
General Public License. }
{$gnu-pascal,I-}
{$if __GPC_RELEASE__ < 20030303}
{$error This unit requires GPC release 20030303 or newer.}
{$endif}
module System;
export System = all (FileMode {$ifdef __BP_TYPE_SIZES__},
SystemInteger => Integer {$endif});
import GPC (MaxLongInt => GPC_MaxLongInt);
var
{ Chain of procedures to be executed at the end of the program }
ExitProc: ^procedure = nil;
{ Contains all the command line arguments passed to the program,
concatenated, with spaces between them }
CmdLine: CString;
{$ifdef __BP_RANDOM__}
{ Random seed, initialized by Randomize, but can also be set
explicitly }
RandSeed: Integer attribute (Size = 32) = 0;
{$endif}
type
OrigInt = Integer;
OrigWord = Word;
{ Delphi }
SmallInt = Integer attribute (Size = 16);
DWord = Cardinal attribute (Size = 32);
{ Short BP compatible type sizes if wanted }
{$ifdef __BP_TYPE_SIZES__}
ByteBool = Boolean attribute (Size = 8);
WordBool = Boolean attribute (Size = 16);
LongBool = Boolean attribute (Size = 32);
ShortInt = Integer attribute (Size = 8);
SystemInteger = Integer attribute (Size = 16);
LongInt = Integer attribute (Size = 32);
Comp = Integer attribute (Size = 64);
Byte = Cardinal attribute (Size = 8);
Word = Cardinal attribute (Size = 16);
LongWord = Cardinal attribute (Size = 32); { Delphi }
{$else}
SystemInteger = Integer;
{$endif}
{$if False} { @@ doesn't work well (dialec3.pas) -- when GPC gets
short
strings, it will be unnecessary }
{$ifopt borland-pascal}
String = String [255];
{$endif}
{$endif}
const
MaxInt = High (SystemInteger);
MaxLongInt = High (LongInt);
{ Return the lowest-order byte of x }
function Lo (x: LongestInt): Byte; attribute (name = '_p_Lo');
{ Return the second-lowest-order byte of x }
function Hi (x: LongestInt): Byte; attribute (name = '_p_Hi');
{ Swap the lowest-order and second-lowest-order bytes, mask out the
higher-order ones }
function Swap (x: LongestInt): Word; attribute (name = '_p_Swap');
{ Store the current directory name (on the given drive number if
drive <> 0) in s }
procedure GetDir (Drive: Byte; var s: String); attribute (name
= '_p_GetDir');
{ Dummy routine for compatibility. @@Use two overloaded versions
rather than varargs when possible. }
procedure SetTextBuf (var f: Text; var Buf; ...); attribute (name
= '_p_SetTextBuf');
{ Mostly useless BP compatible variables }
var
SelectorInc: Word = $1000;
Seg0040: Word = $40;
SegA000: Word = $a000;
SegB000: Word = $b000;
SegB800: Word = $b800;
Test8086: Byte = 2;
Test8087: Byte = 3; { floating-point arithmetic is emulated
transparently by the OS if not present
in hardware }
OvrCodeList: Word = 0;
OvrHeapSize: Word = 0;
OvrDebugPtr: Pointer = nil;
OvrHeapOrg: Word = 0;
OvrHeapPtr: Word = 0;
OvrHeapEnd: Word = 0;
OvrLoadList: Word = 0;
OvrDosHandle: Word = 0;
OvrEmsHandle: Word = $ffff;
HeapOrg: Pointer absolute HeapLow;
HeapPtr: Pointer absolute HeapHigh;
HeapEnd: Pointer = Pointer (High (PtrCard));
FreeList: Pointer = nil;
FreeZero: Pointer = nil;
StackLimit: Word = 0;
HeapList: Word = 0;
HeapLimit: Word = 1024;
HeapBlock: Word = 8192;
HeapAllocFlags: Word = 2;
CmdShow: SystemInteger = 0;
SaveInt00: Pointer = nil;
SaveInt02: Pointer = nil;
SaveInt0C: Pointer = nil;
SaveInt0D: Pointer = nil;
SaveInt1B: Pointer = nil;
SaveInt21: Pointer = nil;
SaveInt23: Pointer = nil;
SaveInt24: Pointer = nil;
SaveInt34: Pointer = nil;
SaveInt35: Pointer = nil;
SaveInt36: Pointer = nil;
SaveInt37: Pointer = nil;
SaveInt38: Pointer = nil;
SaveInt39: Pointer = nil;
SaveInt3A: Pointer = nil;
SaveInt3B: Pointer = nil;
SaveInt3C: Pointer = nil;
SaveInt3D: Pointer = nil;
SaveInt3E: Pointer = nil;
SaveInt3F: Pointer = nil;
SaveInt75: Pointer = nil;
RealModeRegs: array [0 .. 49] of Byte =
(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0);
{ Mostly useless BP compatible pointer functions }
function Ofs (const x): PtrWord; attribute (name = '_p_Ofs');
function Seg (const x): PtrWord; attribute (name = '_p_Seg');
function Ptr (Seg, Ofs: PtrWord): Pointer; attribute (name
= '_p_Ptr');
function CSeg: PtrWord; attribute (name = '_p_CSeg');
function DSeg: PtrWord; attribute (name = '_p_DSeg');
function SSeg: PtrWord; attribute (name = '_p_SSeg');
function SPtr: PtrWord; attribute (name = '_p_SPtr');
{ Routines to handle BP's 6 byte Real type which is formatted like
this:
47 0
-|------- -------- -------- -------- --------|--------
| |
+----------+ +------------+
47 Sign Bit | 8..46 Mantissa | 0..7 Biased Exponent
This format does not support infinities, NaNs and denormalized
numbers. The first digit after the binary point is not stored and
assumed to be 1. (This is called the normalized representation of
a binary floating point number.)
In GPC, this type is represented by the type BPReal which is
binary compatible to BP's type, and can therefore be used in
connection with binary files used by BP programs.
The functions RealToBPReal and BPRealToReal convert between
this type and GPC's Real type. Apart from that, BPReal should
be treated as opaque.
The variables BPRealIgnoreOverflow and BPRealIgnoreUnderflow
determine what to do in the case of overflows and underflows. The
default values are BP compatible. }
var
{ Ignore overflows, and use the highest possible value instead. }
BPRealIgnoreOverflow: Boolean = False;
{ Ignore underflows, and use 0 instead. This is BP's behaviour,
but has the disadvantage of diminishing computation precision. }
BPRealIgnoreUnderflow: Boolean = True;
type
BPRealInteral = Cardinal attribute (Size = 8);
BPReal = record
Format: array [1 .. 6] of BPRealInteral
end;
function RealToBPReal (R: Real): BPReal; attribute (name
= '_p_RealToBPReal');
function BPRealToReal (const BR: BPReal): Real; attribute (name
= '_p_BPRealToReal');
{ Heap management stuff }
const
{ Possible results for HeapError }
HeapErrorRunError = 0;
HeapErrorNil = 1;
HeapErrorRetry = 2;
var
{ If assigned to a function, it will be called when memory
allocations do not find enough free memory. Its result
determines if a run time error should be raised (the default),
or nil should be returned, or the allocation should be retried
(causing the routine to be called again if the allocation still
doesn't succeed).
Notes:
- Returning nil can cause some routines of the RTS and units
(shipped with GPC or third-party) to crash when they don't
expect nil, so better don't use this mechanism, but rather
CGetMem where needed.
- Letting the allocation be retried, of course, only makes sense
if the routine freed some memory before -- otherwise it will
cause an infinite loop! So, a meaningful HeapError routine
should dispose of some temporary objects, if available, and
return HeapErrorRetry, and return HeapErrorRunError when no
(more) of them are available. }
HeapError: ^function (Size: Word): SystemInteger = nil;
{ Just returns HeapErrorNil. When this function is assigned to
HeapError, GetMem and New will return a nil pointer instead of
causing a runtime error when the allocation fails. See the comment
for HeapError above. }
function HeapErrorNilReturn (Size: Word): SystemInteger; attribute
(name = '_p_HeapErrorNilReturn');
{ Return the total free memory/biggest free memory block. Except
under Win32 and DJGPP, these are expensive routines -- try to
avoid them. Under Win32, MaxAvail returns the same as MemAvail, so
don't rely on being able to allocate a block of memory as big as
MaxAvail indicates. Generally it's preferable to not use these
functions at all in order to do a safe allocation, but just try to
allocate the memory needed using CGetMem, and check for a nil
result. What makes these routines unrealiable is, e.g., that on
multi-tasking systems, another process may allocate memory after
you've called MemAvail/MaxAvail and before you get to do the next
allocation. Also, please note that some systems over-commit
virtual memory which may cause MemAvail to return a value larger
than the actual (physical plus swap) memory available. Therefore,
if you want to be "sure" (modulo the above restrictions) that the
memory is actually available, use MaxAvail. }
function MemAvail: Cardinal; attribute (name = '_p_MemAvail');
function MaxAvail: Cardinal; attribute (name = '_p_MaxAvail');
{ Delphi compatibility }
function CompToDouble (x: Comp): Double; attribute (name
= '_p_CompToDouble');
function DoubleToComp (x: Double): Comp; attribute (name
= '_p_DoubleToComp');
function AllocMemCount: SystemInteger; attribute (name
= '_p_AllocMemCount');
function AllocMemSize: SizeType; attribute (name
= '_p_AllocMemSize');
procedure Assert (Condition: Boolean); attribute (name
= '_p_System_Assert');
procedure DefaultAssertErrorProc (const Message, FileName: String;
LineNumber: SystemInteger; ErrorAddr: Pointer); attribute (name
= '_p_DefaultAssertErrorProc');
var
AssertErrorProc: ^procedure (const Message, FileName: String;
LineNumber: SystemInteger; ErrorAddr: Pointer) =
@DefaultAssertErrorProc;
NoErrMsg: Boolean = False;