.\" $Id$ .TH PC_PRLIB 7 "$Revision$" .ad .SH NAME pc_prlib \- library of Pascal runtime routines .SH SYNOPSIS .ta 11n 22n 33n 44n 55n .nf type alpha=packed array[1..8] of char; pstring= ^packed array[] of char; function _abi(i:integer):integer; function _abl(i:long):long; function _mdi(j,i:integer):integer; function _mdl(j,i:long):long; function _dvi(j,i:integer):integer; function _dvl(j,i:long):long; function _abr(r:real):real; function _sin(r:real):real; function _cos(r:real):real; function _atn(r:real):real; function _exp(r:real):real; function _log(r:real):real; function _sqt(r:real):real; function _rnd(r:real):real; type compared=-1..1; gotoinfo=record pcoffset:^procedure; { procedure id. without static link } nlocals: integer; end; type arrdescr=record lowbnd: integer; diffbnds:integer; elsize: integer; end; arr1=array[] of ?; arr2=packed array[] of ?; function _bcp(sz:integer; s2,s1:pstring):compared; function _bts(size,high,low:integer; base:^set 0..(8*size-1)) :set of 0..(8*size-1); procedure _gto(lb:^integer; p:^gotoinfo); procedure _rcka(a: arrdescr; index : integer); procedure _nfa(bool:integer); procedure _new(size:integer; var p:^integer); procedure _dis(size:integer; var p:^integer); procedure _sav(var p:^integer); procedure _rst(var p:^integer); procedure _pac(var ad,zd:arrdescr; var zp:arr2; i:integer; var ap:arr1); procedure _unp(var ad,zd:arrdescr; i:integer; var ap:arr1; var zp:arr2;nosgnext:boolean); function _asz(var dp:arrdescr):integer; procedure _ass(line:integer; b:boolean); procedure procentry(name:pstring); procedure procexit(name:pstring); const lowbyte=[0..7]; MAGIC =[1,3,5,7]; WINDOW =[11]; ELNBIT =[12]; EOFBIT =[13]; TXTBIT =[14]; WRBIT =[15]; PC_BUFLEN =1024; type file=record ptr: ^char; flags: set of [0..15]; fname: pstring; ufd: 0..15; size: integer; count: 0..buflen; buflen: max(PC_BUFLEN,size) div size * size; bufadr: packed array[1..max(PC_BUFLEN,size)] of char; end; filep=^file; const NFILES=15; _extfl:^array[] of filep; procedure _ini(args:pstring; var c:integer; var p:array[] of filep; mainlb:pstring); procedure _hlt(status:0..255); procedure _opn(size:integer; f:filep); procedure _cre(size:integer; f:filep); procedure _cls(f:filep); procedure _get(f:filep); procedure _put(f:filep); function _wdw(f:filep):^char; function _efl(f:filep):boolean; function _eln(f:filep):boolean; function _rdc(f:filep):char; function _rdi(f:filep):integer; function _rdl(f:filep):long; function _rdr(f:filep):real; procedure _rln(f:filep); procedure _wrc(c:char; f:filep); procedure _wsc(w:integer; c:char; f:filep); procedure _wri(i:integer; f:filep); procedure _wsi(w:integer; i:integer; f:filep); procedure _wrl(l:long; f:filep); procedure _wsl(w:integer; l:long; f:filep); procedure _wrr(r:real; f:filep); procedure _wsr(w:integer; r:real; f:filep); procedure _wrf(ndigit:integer; w:integer; r:real; f:filep); procedure _wrs(l:integer; s:pstring; f:filep); procedure _wss(w:integer; l:integer; s:pstring; f:filep); procedure _wrb(b:boolean; f:filep); procedure _wsb(w:integer; b:boolean; f:filep); procedure _wrz(s:string; f:filep); procedure _wsz(w:integer; s:string; f:filep); procedure _wln(f:filep); procedure _pag(f:filep); .fi .SH DESCRIPTION This library is used by the Pascal to EM compiler and contains all the runtime routines for standard Pascal programs. Most routines are written in C, a few in EM assembly language. These routines can be divided into several categories. A description of each category with its routines follows. .PP Arithmetic routines: .RS .IP _abi Compute the absolute value of an integer. .PD 0 .IP _abl Compute the absolute value of a long. .IP _mdi Perform the Pascal modulo operation on integers. .IP _mdl Perform the Pascal modulo operation on longs. .IP _dvi Perform the Pascal divide operation on integers. .IP _dvl Perform the Pascal divide operation on longs. .IP _abr Compute the absolute value of a real. .IP _sin Compute the sine of a real. .IP _cos Compute the cosine of a real. .IP _atn Compute the arc tangent of a real. .IP _exp Compute the e-power of a real. .IP _log Compute the natural logarithm of a real. .IP _sqt Compute the square root of a real. .IP _rnd Return a real that when truncated will result in the nearest integer (-3.5->-4). .PD .PP .RE Miscellaneous routines: .RS .IP _bcp Compare two strings. Use dictionary ordering with the ASCII character set. The EM instruction CMU can not be used, because it needs an even number of bytes. .PD 0 .IP _bts Include a range of elements from low to high in a set of size bytes at address base.(size can be divided by the wordsize) .IP _gto Execute a non-local goto. Lb points to the local base of the target procedure. A lb of zero indicates a jump to the program body, the lb of the main program is found in _m_lb, which is set by _ini. The new EM stack pointer is calculated by adding the number of locals to the new local base (jumping into statements is not allowed; there are no local generators in Pascal!). .IP _rcka Check if an array reference isn't out of bounds. This is should be done by the backends, but some don't. .IP _nfa Check whether a function is assigned or not. This could have been done by the compiler, but that would make the interface between the compiler and the run-time library messier. .PD .PP .RE Heap management: .RS .PP There is one way to allocate new heap space (_new), but two different incompatible ways to deallocate it. .PP The most general one is by using dispose (_dis). A circular list of free blocks, ordered from low to high addresses, is maintained. Merging free blocks is done when a new block enters the free list. When a new block is requested (_new), the free list is searched using a first fit algorithm. Two global variables are needed: .IP _highp 10 Points to the free block with the highest address. .PD 0 .IP _lastp Points to the most recently entered free block or to a block in the neighborhood of the most recently allocated block. .PD The free list is empty, when one of these pointers (but then at the same time both) is zero. .PP The second way to deallocate heap space is by using mark (_sav) and release (_rst). Mark saves the current value of the heap pointer HP in the program variable passed as a parameter. By calling release with this old HP value as its argument, the old HP value is restored, effectively deallocating all blocks requested between the calls to mark and release. The heap is used as second stack in this case. .PP It will be clear that these two ways of deallocating heap space can not be used together. To be able to maintain the free list, all blocks must be a multiple of n bytes long, with a minimum of n bytes, where n is the sum of the size of a word and a pointer in the EM implementation used. .PP In summary: .IP _new Allocate heap space. .PD 0 .IP _dis Deallocate heap space. .IP _sav Save the current value of HP. .IP _rst Restore an old value of HP. .PD .PP .RE Array operations: .RS .PP The only useful form of packing implemented, is packing bytes into words. All other forms of packing and unpacking result in a plain copy. .IP _pac Pack an unpacked array \fIa\fP into a packed array \fIz\fP. \fIap\fP and \fIzp\fP are pointers to \fIa\fP and \fIz\fP. \fIad\fP and \fIzd\fP are pointers to the descriptors of \fIa\fP and \fIz\fP. \fIi\fP is the index in \fIa\fP of the first element to be packed. Pack until \fIz\fP is full. .PD 0 .IP _unp Unpack \fIz\fP into \fIa\fP. \fIap\fP, \fIzp\fP, \fIad\fP and \fIzd\fP are as for _pac. \fIi\fP is the index in \fIa\fP where the first element of \fIz\fP is copied into. Unpack all elements of \fIz\fP. The boolean flag \fInosgnext\fP indicates whether the values should or should not be sign-extended. .IP _asz Compute array size. Used for copying conformant arrays. .PD .PP .RE Debugging facilities: .RS The compiler allows for the verification of assertions. It generates a call to the routine _ass to check the assertion at runtime. Another feature of the compiler is that it enables the user to trace the procedure calling sequence. If the correct option is turned on, then a call to the procedure \fIprocentry\fP is generated at the start of each compiled procedure or function. Likewise, the routine \fIprocexit\fP is called just before a procedure or function exits. Default procedure \fIprocentry\fP and \fIprocexit\fP are available in this library. .IP _ass 10 If \fIb\fP is zero, then change eb[0] to \fIline\fP (to give an error message with source line number) and call the error routine. .PD 0 .IP procentry Print the name of the called procedure on standard output. Output must be declared in the program heading. .IP procexit Print the name of the procedure that is about to exit. Output must be declared in the program heading. .PD .PP .RE Files: .RS .PP Most of the runtime routines are needed for file handling. For each file in the Pascal program a record of type file, as described above, is allocated, static if this file is declared in the outermost block, dynamic if it is declared in inner blocks. The fields in the file record are used for: .IP bufadr 10 IO is buffered except for standard input and output if terminals are involved. The size of the buffer is the maximum of PC_BUFLEN and the file element size. .PD 0 .IP buflen The effective buffer length is the maximum number of file elements fitting in the buffer, multiplied by the element size. .IP size The file element size (1 or even). .IP flags Some flag bits are stored in the high byte and a magic pattern in the low byte provides detection of destroyed file information. .IP ptr Points to the file window inside the buffer. .IP count The number of bytes (the window inclusive) left in the buffer to be read or the number of free bytes (the window inclusive) for output files. .IP ufd The UNIX file descriptor for the file. .IP fname Points to the name of the file (INPUT for standard input, OUTPUT for standard output and LOCAL for local files). This field is used for generating error messages. .PD .PP The constants used by the file handling routines are: .IP WINDOW 10 Bit in flags set if the window of an input file is initialized. Used to resolve the famous interactive input problem. .PD 0 .IP EOFBIT Bit in flags set if end of file seen .IP ELNBIT Bit in flags set if linefeed seen .IP TXTBIT Bit in flags set for text files. Process linefeeds. .IP WRBIT Bit in flags set for output files .IP MAGIC Pattern for the low byte of flags .IP NFILES The maximum number of open files in UNIX .PD .PP .RE Prelude and postlude: .RS .PP These routines are called once for each Pascal program: .IP _ini When a file mentioned in the program heading is opened by reset or rewrite, its file pointer must be mapped onto one of the program arguments. The compiler knows how to map and therefore builds a table with a pointer to the file structure for each program argument. One of the first actions of the Pascal program is to call this procedure with this table as an argument. The global variable _extfl is used to save the address of this table. Another task of _ini is to initialize the standard input and output files. For standard output it must decide whether to buffer or not. If standard output is a terminal, then buffering is off by setting buflen to 1. Two other task of _ini are the copying of two variables from the argument list to global memory, mainlb to _m_lb and c to _extflc. The first contains the local base of the program body, the second contains the number of paremeters the program is called with. A last task of _ini is to set the global variables _pargc, _pargv and _penvp from args for possible reference later on. Args points to the argument count placed on the stack by the EM runtime system, see chapter 8 in [1]. .PD 0 .IP _hlt If the program is about to finish, the buffered files must be flushed. That is done by this procedure. .PD .PP .RE Opening and closing: .RS .PP Files in Pascal are opened for reading by reset and opened for writing by rewrite. Files to be rewritten may or may not exist already. Files not mentioned in the program heading are considered local files. The next steps must be done for reset and rewrite: .IP 1. If size is zero, then a text file must be opened with elements of size 1. .PD 0 .IP 2. Find out if this file is mentioned in the program heading (scan table pointed to by _extfl). If not, then it is a local file and goto 7. .IP 3. If the file is standard input or output then return. .IP 4. If there are not enough arguments supplied, generate an error. .IP 5. If the file was already open, flush the buffer if necessary and close it. Note that reset may be used to force the buffer to be flushed. This is sometimes helpful against program or system crashes. .IP 6. If it is a reset, open the file, otherwise create it. In both cases goto 9. .IP 7. If the local file is to be written, then close it if it was open and create a new nameless file. First try to create it in /usr/tmp, then in /tmp and if both fail then try the current directory. See to it that the file is open for both reading and writing. .IP 8. If the local file is to be read and the file is opened already, then flush the buffer and seek to the beginning. Otherwise open a temporary file as described in 7. .IP 9. Initialize all the file record fields. .PD .PP The necessary procedures are: .IP _opn Reset a file .PD 0 .IP _cre Rewrite a file .IP _cls Close a file. Closing of files is done for local files when the procedure in which they are declared exits. The compiler only closes local files if they are not part of a structured type. Files allocated in the heap are not closed when they are deallocated. There is an external routine \fIpclose\fP in libpc(7), that may be called explicitly to do the closing in these cases. Closing may be necessary to flush buffers or to keep the number of simultaneously opened files below NFILES. Files declared in the outermost block are automatically closed when the program terminates. .PD .PP .RE General file IO: .RS .PP These routines are provided for general file IO: .IP _put Append the file element in the window to the file and advance the window. .IP _get Advance the file window so that it points to the next element of the file. For text files (TXTBIT on) the ELNBIT in flags is set if the new character in the window is a line feed (ASCII 10) and the character is then changed into a space. Otherwise the ELNBIT is cleared. .IP _wdw Return the current pointer to the file window. .IP _efl Test if end of file is reached. Is always true for output files. .PD .PP .RE Textfile routines: .RS .PP The rest of the routines all handle text files. .IP _eln Return true if the next character on an input file is an end-of-line marker. An error occurs if eof(f) is true. .PD 0 .IP _rdc Return the character currently in the window and advance the window. .IP _rdi Build an integer from the next couple of characters on the file, starting with the character in the window. The integer may be preceded by spaces (and line feeds), tabs and a sign. There must be at least one digit. The first non-digit signals the end of the integer. .IP _rdl Like _rdi, but for longs. .IP _rdr Like _rdi, but for reals. Syntax is as required for Pascal. .IP _rln Skips the current line and clears the WINDOW flag, so that the next routine requiring an initialized window knows that it has to fetch the next character first. .IP _wrc Write a character, not preceeded by spaces. .IP _wsc Write a character, left padded with spaces up to a field width of \fIw\fP. .IP _wri Write an integer, left padded with spaces up to a field width of 6 on 2-byte machines and a field width of 11 on 4-byte machines. .IP _wsi Write an integer, left padded with spaces up to a field width of \fIw\fP. .IP _wrl Write a long, left padded with spaces up to a field width of 11. .IP _wsl Write a long, left padded with spaces up to a field width of \fIw\fP. .IP _wrr Write a real in scientific format, left padded with spaces up to a field width of 13. .IP _wsr Write a real in scientific format, left padded with spaces up to a field width of \fIw\fP. .IP _wrf Write a real in fixed point format, with exactly \fIndigit\fP digits behind the decimal point, the last one rounded; it is left padded up to a field width of \fIw\fP. .IP _wrs Write a string of length \fIl\fP, without additional spaces. .IP _wss Write a string of length \fIl\fP, left padded up to a field width of \fIw\fP. .IP _wrb Write a boolean, represented by "true" or "false", left padded up to a field width of 5. .IP _wsb Write a boolean, represented by "true" or "false", left padded up to a field width of \fIw\fP. .IP _wrz Write a C-type string up to the zero-byte. .IP _wsz Write a C-type string, left padded up to a field width of w. .IP _wln Write a line feed (ASCII 10). .IP _pag Write a form feed (ASCII 12). .PD .PP .RE All the routines to which calls are generated by the compiler are described above. They use the following global defined routines to do some of the work: .IP _rf 10 Check input files for MAGIC and WRBIT. Initialize the window if WINDOW is cleared. .PD 0 .IP _wf Check output files for MAGIC and WRBIT. .IP _incpt Advance the file window and read a new buffer if necessary. .IP _outcpt Write out the current buffer if necessary and advance the window. .IP _flush Flush the buffer if it is an output file. Append an extra line marker if EOLBIT is off. .IP _wstrin All output routines make up a string in a local buffer. They call _wstrin to output this buffer and to do the left padding. .IP _skipsp Skip spaces (and line feeds) on input files. .IP _getsig Read '+' or '-' if present. .IP _fstdig See to it that the next character is a digit. Otherwise error. .IP _nxtdig Check if the next character is a digit. .IP _getint Do the work for _rdi. .IP _ecvt Convert real into string of digits for printout in scientific notation. .IP _fcvt Convert real into string of digits for fixed point printout .IP -fif Split real into integer and fraction part .IP _fef Split real into exponent and fraction part .PD .PP The following global variables are used: .IP _lastp 10 For heap management (see above). .PD 0 .IP _highp For heap management (see above). .IP _extfl Used to save the argument p of _ini for later reference. .IP _extflc Used to save the argument c of _ini for later reference. .IP _m_lb Used to store the local base of the main program. .IP _curfil Save the current file pointer, so that the error message can access the file name. .IP "_pargc, _pargv, _penvp" Used to access the arguments of the main program. .PD .SH FILES .IP ~em/lib/*/tail_pc 20 The library used by ack[5] to link programs. .PD .SH "SEE ALSO" .IP [1] A.S. Tanenbaum, Ed Keizer, Hans van Staveren & J.W. Stevenson "Description of a machine architecture for use of block structured languages" Informatica rapport IR-81. .PD 0 .IP [2] K.Jensen & N.Wirth "PASCAL, User Manual and Report" Springer-Verlag. .IP [3] Specification fo Computer programming language Pascal, BS6192: 1982 (ISO 7185) .IP [4] J.W. Stevenson, H. van Eck, "Amsterdam Compiler Kit-Pascal reference manual". .br (try \fItbl ~em/doc/pcref.doc | nroff\fP). .IP [5] ack(1), em_pc(6) .PD .SH DIAGNOSTICS All errors discovered by this runtime system cause an EM TRP instruction to be executed. This TRP instruction expects the error number on top of the stack. See [1] for a more extensive treatment of the subject. .PP EM allows the user to specify a trap handling routine, called whenever an EM machine trap or a language or user defined trap occurs. One of the first actions in _ini is to specify that the routine _fatal, available in this library, will handle traps. This routine is called with an error code (0..252) as argument. The following information is printed on file descriptor 2: .IP - The name of the Pascal program .PD 0 .IP - The name of the file pointed to by _curfil, if the error number is between 96 and 127 inclusive. .IP - The error message (or the error number if not found). .IP - The source line number if not equal to 0. .PD .PP The routine _fatal stops the program as soon as the message is printed. .PP The following error codes are used by the Pascal runtime system: .IP 64 more args expected .PD 0 .IP 65 error in exp .IP 66 error in ln .IP 67 error in sqrt .IP 68 assertion failed .IP 69 array bound error in pack .IP 70 array bound error in unpack .IP 71 only positive j in \fIi mod j\fP .IP 72 file not yet open .IP 73 dispose error .IP 74 function not assigned .IP 75 illegal field width .sp .IP 96 file xxx: not writable .IP 97 file xxx: not readable .IP 98 file xxx: end of file .IP 99 file xxx: truncated .IP 100 file xxx: reset error .IP 101 file xxx: rewrite error .IP 102 file xxx: close error .IP 103 file xxx: read error .IP 104 file xxx: write error .IP 105 file xxx: digit expected .IP 106 file xxx: non-ASCII char read .PD .PP .SH AUTHORS Johan Stevenson and Ard Verhoog, Vrije Universiteit.