CIM: a portable SIMULA system based on C User's Guide by Sverre Johansen, Stein Krogdahl and Terje Mjs Department of informatics University of Oslo January 1991 0. FOREWORD ( J. Vaucher, 1995 ) ================================= This text is an abridged and reformatted plain ascii version of the "User and INSTALLATION guide" given in the "doc" directory of the GNU CIM distribution ( directories: /usr/local/cim/ or /usr/local/cim/doc). The original document called "intro" exists in ".DVI", ".tex" and ".ps" formats but not in plain text. The section on installation has been dropped. For full and exact information, the reader should refer to the originals. 1. INTRODUCTION ================ Cim is a compiler for the programming language Simula (except unspecified parameters to formal or virtual procedures). It offers a class concept, separate compilation with full type checking, interface to external C-routines, an application package for process simulation and a coroutine concept. Cim is a Simula compiler whose portability is based on the C programming language. The compiler and the run-time system is written in C, and the compiler produces C-code, that is passed to a C-compiler for further processing towards machine code. Cim is copyrighted by Sverre Hvammen Johansen, Stein Krogdahl and Terje Mjs, Department of Informatics, University of Oslo. Cim 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; version 2. 2. A SIMPLE EXAMPLE ==================== We show in this section a simple example of a SIMULA program and how to compile and run it. Create a SIMULA program with a text editor, and give it a name with extension ``sim''. In this example we name it "doesit.sim": begin Outtext("Simula does it in C"); Outimage; end You can now compile the program with the CIM command: %cim doesit Compiling: doesit.sim cc -w -c doesit.c cc -o doesit doesit.o /usr/local/lib/libcim.a -lm % The compiler will produce C code that is further processed towards machine code with a standard C compiler. All produced files have the same name as the input file, but with different extensions. The compiled and linked program can be invoked by entering: %./doesit Simula does it in C % 3. COMPILING ============= Cim is a Simula compiler that first compiles the source code into C. The C code will then be compiled with cc, and linked with other modules. The Cim command will accept one Simula program and other non-Simula modules. The specified Simula program will be compiled and linked with the modules. If a main Simula program is compiled, it will automatic be linked with the necessary Simula modules. If a separate Class or Procedure is compiled, then the linking will be suppressed. The diagnostics produced by the Simula compiler are intended to be self-explanatory. The following options are accepted by the CIM command: -a Try to produce an atr-file even if an error occurs. Compare the produced atr-file with the atr-file produced from previous compilation and if they differ return an error status code. Note: With use of this option it is possible to have external modules with circular dependencies. You will then need to compile all the modules with this option until no error status codes are returned. Then you should do a final compilation with option -R or option -d. To get this to work it is important that the topmost external head does not contain any external declaration that is part of the circular dependency. Such external declarations must be placed in an external head that comes after the first class- or procedure decraration. -c Suppress linking of the complete program. -C Only link the specified files. -cc The following argument is the name of the C-compiler. -d Compare the produced c-code with the code produced from previous compilation and if they are equal then touch the object-file instead of compiling the c-code. -Dname Define a symbol name. -E Run only the preprocessor and output the result to standard output. -g Make the C compiler produce debugging information. This option is useful for debugging the generated code. -gcc Invoke the Gnu Project C compiler instead of the standard C compiler. This option can be used if the standard C compiler don't generate correct code. -I dir Use the Simula include file located in directory dir instead of the standard directory. This information may also be given by setting the environment variable CIMINCLUDEDIR. -l Omit line number information in the compiled program. This will make the program smaller and faster. -llibrary Link with object library library. This option is parsed to the link-command. -Ldir Use the Simula library located in directory dir instead of the standard directory. This information may also be given by setting the environment variable CIMLIBDIR. -m The memory pool size may be set at runtime by an option -mn. -mn Set the initial memory pool size to n mega bytes. -Mn Set the maximal memory pool size to n mega bytes. -o The following argument is the name of the output executable file. -oc The following argument will be parsed to the CC-command. -ol The following argument will be parsed to the link-command. -q Run the compiler in quiet mode. -r Run the program after compilation. -R Recompile the module using the same timestamp. -s Only C-compile and link the specified files. -S Run the source file through Simula-compiler, only. -t Do not remove temporary files. If a main program is compiled with option -r, then the executable file will be removed unless this option or option -T is specified. -T Do not remove the executable file. -Uname Remove any initial definition of the symbol name (Inverse of the -D option). -v Run the compiler in verbose mode. -w Do not print warnings. 3.1 Arguments -------------- The following arguments are accepted by the Cim command: file.a Library of object files and attribute files. Include this simula library when compiling and linking. The simula library is created with ar(1V) and ranlib(1). file.o Object file of other non-Simula modules. file.sim Simula source file. A file name without an extension are assumed to be shorthand notation for the corresponding Simula file. 4. USE OF EXTERNAL PROCEDURES ============================== C is the only language supported for non-Simula external procedures. External C procedures must be specified in the following way: External C procedure is procedure ; ; [ Note: the is case-sensitive ] The rules for external C procedures are: - Avoid global symbols prefixed with "__", it may lead to conflicts with system names in Cim. - The procedure may have any type, except "ref". If the type is "text", then the null terminated string returned from C is converted to a Simula text object. - Parameters may not be a Simula-procedure, switch or label. - Parameters transmitted by value are always copied. Text or arrays are allocated by malloc, and are not deallocated by Cim. It's the C-programs responsibility to dealloc the space. - Parameters transmitted by reference or name are transmitted to C as pointers. Array or text are transmitted to C by the location of the first element. - External C procedures with variable number of parameters can be specified by use of ``...'' in the end of the parameter list. For example: PRINTF and SCANF can be specified as follows: EXTERNAL C PROCEDURE printf IS INTEGER PROCEDURE printf(t,...);TEXT t;; EXTERNAL C PROCEDURE scanf IS INTEGER PROCEDURE scanf(t,...);NAME ...;TEXT t;; 4.1 Extension to the environment --------------------------------- The following procedures have been added to the Simula environment and may be called directly from Simula: PROCEDURE Gbc;...; The garbage collector is called when the dynamic storage exceeds an implementation dependent limit. The garbage collector traverse and moves all the accessible objects, and leaves the free space as one area initialized to zero. The garbage collector may be called explicitly through the procedure Gbc. INTEGER PROCEDURE Argc;...; Returns the number of command-line arguments that the program was invoked with. INTEGER PROCEDURE Argv;...; Returns a pointer to an array of character strings (in C fashion) that contains the arguments. PROCEDURE Dump(t);TEXT t;...; Dump the state of the Simula-program to file. Before a call on Dump all files except SYSIN, SYSOUT and SYSERR should be closed. PROCEDURE UnDump(t);TEXT t;...; Read a previously stored state from file and start the program in that state. To get these procedures to work, they should be compiled into the same program. The program may not be re-compiled between a call on Dump and UnDump. REF(PrintFile) PROCEDURE SysErr;...; Returns the file object associated with standard error. 5. THE PREPROCESSOR ==================== CIM includes a pre-processor whose most used feature (in JV's opinion) is the %include directive: %include filename Will cause the compiler to include the indicated file in place of the INCLUDE directive line. This directive may be nested, but only to a level of 10. Here are the other directives supported: %whitespace ... A directive line starting with a whitespace is treated as a comment. %nocomment ... The rest of the line is treated as ordinary source text. Some other simula implementations will ignore this line, and give a warning message. But this can be useful as the following example shows. In this implementation formal procedures must be specified, but that should not be done in standard simula. This will work both on LSH (Simula implementation from Lund Software House AB, Sweden) and Cim: PROCEDURE P(i1,P2);INTEGER i1; %nocomment PROCEDURE P2 IS INTEGER PROCEDURE P2 %nocomment (i,j);INTEGER i,j; ; %comment Will cause the compiler to strip all lines until the corresponding %endcomment is reached. This directive may be nested. %eof Will cause the compiler to react as if the end of the source file was reached. Include files that is placed in a archive must be preceded with this directive line. %casesensitive ON/OFF The case sensitivity of identifiers and keywords is turned ON or OFF. Default value is OFF. %define name Define a name. Names such as alpha, amiga, apollo, atari, dec, rs6000, sgi, solaris, sparc, sunos, unix, vax, vms, are predefined dependent of the system. The name cim is defined for all implementations. %error ... Will cause the compiler to believe that it has found an error in the source text. The message that is preceded on the line is printed as an error message. %ifdef name If name is not defined then the compiler will strip all lines until the corresponding %else or %endif is reached. If name is not defined then the compiler will strip all lines between the optional %else and %endif. %include filename Will cause the compiler to include the indicated file in place of the INCLUDE directive line. This directive may be nested, but only to a level of 10. %line %nameasvar ON/OFF If it is turned ON, then transmission mode for name is implemented as reference. This will produce more efficient code. Default value is OFF. %staticblock ON/OFF If it is turned on, then data objects will be allocated static instead of dynamic, and the compiler may generate more efficient code. This option should be used with care and should not be used for blocks which may have more than one active data object at a given time. The option may not be used for classes that are given as prefix or virtual procedures or procedures that are parameter to other procedures. It may not be used for external classes or procedures. %stripsideeffects ON/OFF If it is turned ON, then the compiler can generate more efficient code, but not necessary correct code due to evaluation order for expressions. Default value is OFF. %undefine name Undefine a name. If the name is not defined the directive line has no effect. 6. IMPLEMENTATION ASPECTS ========================== 6.1 Language restrictions ------------------------- A formal or virtual procedure must be fully specified with respect to its type, and type, kind and transmission mode of its parameters. The Simula Standard and other implementations also allow short forms. To write portable code, see the %nocomment directive above. Allowed implementation restrictions: - The types short integer and long real are implemented as integer and real. - The standard access mode SHARED for files is not implemented. - The only and default byte size of access mode BYTESIZE is 8. 6.2 Implementation dependent characteristics --------------------------------------------- o Trailing blanks of image are not transferred to the external file on outfile.outimage excepts it's a direct file. o A parameter to printfile.spacing with value zero gives the standard effect of overprint. o The procedures lock and unlock are not implemented. o All open external files are closed when a program is terminated. 6.3 Implementation defined characteristics ------------------------------------------- - The internal characters are the same as the standard character set. - SYSIN, SYSOUT and SYSERR are connected to standard input, standard output and standard error. If they are closed and reopened they are connected to /dev/tty under UNIX, AIX, LINUX and MINIX and sysinput and sysoutput under VMS. - Inlength and outlength are equal to 80. - The relative value ranges of REAL are as DOUBLE in C and ranges of INTEGER are as LONG. - Conversion from an integer type to a real type is exact except for implementations where integer have better precision than real (which is the case for the alpha and cray implementation). - The effect is not defined if the range of a numeric item in a de-editing procedure exceeds the value range of the procedure result. - The exponent from ``putreal'' has 5 characters except for the cray implementation where it may be 6 characters. - A text frame has a maximum length of about 64K characters. - The return values of ``char'' and ``rank'' are as given by the standard character set. - The exact definitions of the standard mathematical functions are system specific. - The association between a file object and an external file is through standard procedures based on C's FILE. The object is connected to the external file when open is called. - Several file objects may represent the same external file, but the effect is not defined if some of them are opened for writing. - A minimum of checks are performed at ``locate''. - The default value to LINESPERPAGE is MAXINT. - The ``basic random drawing'' is implemented as suggested in the standard. - Two decimals are used for the field for seconds of the function ``datetime''. - Evaluation of arithmetic expressions is based on C, but a Simula expression is by default divided up in several expressions, to guarantee correct evaluation order. 6.4 Capacity limitations ------------------------- The compiler has the following logical limitations: - The maximal number of Simula-libraries that the compiler can search is 100 libraries. - Length of a token in the input stream is restricted to about 1000 characters. - The nesting of compiler directives is limited to 100. - The level of nesting of include files must not exceed 10. - The parser is written i YACC and the parser stack have a size equal 1500 elements. - Block nesting level is limited to 100. - The compiler builds an expression tree for each Simula-expression, and one tree is limited to 1000 nodes. - The code generator have a stack of labels with 1000 elements. - Temporary expressions may not consist of more than 100 value-type, 100 text, or 100 ref-type elements. - The nesting of temporary expressions may not be deeper than 100 levels. - The maximum number of dimensions for arrays is 100. - Text objects may not contain more than about 64K characters. - Some other limitations that are based on the underlying hardware, the operating system or the C Compiler, which are not checked by the compiler. 7. BUG REPORTS =============== Mail bug reports for Cim to "cim-bug@ifi.uio.no". Please include the Cim version number and the system type, which you can get by running ``cim -v''. 8. AUTHORS =========== - Terje Ms, Hydro Data, Oslo. - Sverre Hvammen Johansen, Department of Informatics, University of Oslo. - Stein Krogdahl, Department of Informatics, University of Oslo.