Piper's Pit BBS/FTP: ibm 0020

home *** CD-ROM | disk | FTP | other *** search

/ Piper's Pit BBS/FTP: ibm 0020 - 0029 / ibm0020-0029 / ibm0028.tar / ibm0028 / ADA-1.ZIP / ADA-1.TD0 / READ.ME < prev

Wrap

Text File | 1989-12-04 | 17.8 KB | 410 lines

10/25/89 Here's a summary of changes to version 2.1.3 of the 80x86 Ms-Dos Janus/Ada Compiler and Extended Functions Kit, along with some last minute notes. --------------------------------------------------------------------------- Information for New Janus/Ada Users --------------------------------------------------------------------------- New users should read this file thoroughly. Veteran users can skip to the section on installation and checkout (line number 115). You will start receiving the newsletter when you sign and send in your license card. We often get calls from people who want to know when an update is coming out. Here's a familiar conversation: "It's in the newsletter; didn't you see it?" "I didn't get one." "Did you send in your license card?" "No" ... The newsletter is published quarterly, so it may take as long as 3 months for your first issue to appear. --------------------------------------------------------------------------- Introduction to Memory Models for New Janus/Ada Users --------------------------------------------------------------------------- This compiler is distributed with two runtimes. One is "model 0", the other "model 1". These are used to produce programs of the appropriate memory model. All compilation units in a given program must be compiled with the same model; the compiler and the linker enforce this rule. The /O (the letter O, not the number 0) option is used on the compiler and linker to select the memory model. Model 0 is assumed if the option is not used. Model Max Code Max Data Max Constants ----- -------- -------- ------------- 0 64K 64K 64K 1 "Unlimited" 64K 64K (Use BigArray/Memory [see section 15.10] to expand your data area). Model 1 programs are about 15% bigger and slower than the equivalent Model 0 programs. All Model 1 programs are .EXE files; Model 0 programs can be either .COM or .EXE. Using memory models in the assembler is tricky. The important things to remember are: Use the Pragma Memory_Model to select a memory model. We usually declare a constant (EQU) "Memory_Model", and conditionally assemble both models. All external calls or jumps must be Near in Model 0; and Far (Call Far xxx) in Model 1. This includes references to local entry points. Returns should match the Calls; i.e. Near Returns (just RET) for Model 0; and Far Returns (RET FAR) for model 1. When accessing parameters, remember that there are two more bytes on the stack for the return address when using model 1 than model 0. The SEGMENT directive allows references to the segment addresses of any compilation unit under model 1. --------------------------------------------------------------------------- Potential Surprises for New Janus/Ada Users --------------------------------------------------------------------------- If a user has a directory or Read_Only file named "CONT.$$$", the compiler will refuse to run. In general, it is bad to have files with an extension of ".$$$", as this is used for temporary files. It is worse to have such files write-protected. The compiler and linker do obsoleteness checking; compiling things in the wrong order will result in a program that will not link. Use CORDER to get the compilation order right. Operators, just like identifiers, follow the Ada rules about scope and visibility. Don't assume that common operators are always available. For example: Package A Is Type B Is New Integer; End A; With A; Procedure C Is D : A.B := 2; Begin If D = 4 Then -- Oops, = for type B is not visible. -- .... End If; If A."="(D,4) Then -- OK -- .... End If; Moral: If you want infix operations, you need a Use clause. When we find bugs, they are printed in each newsletter; the current up-to-date bug list will be found on the Janus/Ada BBS system (608-244-1528, 24 hours a day, 300/1200/2400 baud). Look for Bulletin number four. R.R Software, Inc. produces compilers for serious programmers. We hope that you appreciate our professional stance on reporting bugs. If you find a bug, please send us the source code for a short program which illustrates the problem. You can submit these programs on a floppy disk or upload them to our BBS or the janus.ada conference on BIX, the Byte Information Exchange network. Please don't send printouts. Good Luck! --------------------------------------------------------------------------- Janus/Ada Checkout and Installation --------------------------------------------------------------------------- NOTE: Even if you used an earlier version of Janus/Ada, you should read the checkout and installation guide carefully. Improper set-up (including set-up methods that may have worked with earlier versions of Janus/Ada) may cause the compiler or the linker to report errors like "Time stamp for .SYM file does not match". We do not recommend using the compiler on a system without a hard disk (although it can be done). Use the Install batch file to install the compiler; or do the steps yourself (some people prefer that). If you don't want the compiler to go into a directory called \JANUS213, you must supply a directory name on the command line when running the install batch file. You might use \JANUS or \JANUSADA as the name of your directory. The calling convention for the install batch file is provided in your compiler Installation and Checkout Sheets. Run Chkdsk to determine the amount of free memory needed. You will need 570,000 "bytes free" (as reported by Chkdsk) for the compiler to run. Many commonly used device drivers will use up too much memory. For example, most mouse drivers will take about 40K of RAM, thus reducing the amount of free RAM below 570K. The compiler will use additional memory if it is available (to a maximum of about 80K) giving you a larger symbol table. You would need a 704K system (or a memory-remapping program like 386Max) to use all of the extra symbol table space. The compiler also can place some of the symbol table on disk. The Janus/Ada Compiler manual may have references to tools supplied with the Extended Functions Kit or the Compiler Tool Kit. These products include the following useful extensions to the basic Janus/Ada system: JASM86 - assembler; DISASM86 - disassembler; JWINDOWS - window management package; SYNTAX - syntax checker with pretty printer; JBIND - replacement linker which produces Microsoft OBJ files; this allows linking with MASM subprograms; PROFILER - program execution profiler; EFIND - regular-expression pattern matcher; (text searching utility) and other tools and interfaces to third party software. Contact your sales representative if you would like information about these products. R.R. Software also offers a source level debugger, integrated editors, and embedded systems/ICE support. --------------------------------------------------------------------------- Changes to 2.1.3 from 2.1.2 --------------------------------------------------------------------------- Version 2.1.3 has a number of new features and corrects several bugs in the 2.1.2 compiler. A brief summary follows. Programs will NOT need to be recompiled when switching to 2.1.3 from 2.1.2. Fixing certain bugs will require relinking existing programs. See the list below to determine if your software will need relinking. The 2.1 Product line consists of the following versions: Product Version Release ------- ------- ------- Compiler 2.1 . 3 Linker 2.1 . 3 Corder 2.1 . 3 Jmanager 2.1 . 3 Efind 2.1 . 3 J.E.T. 2.0 - Assembler 2.1 . 3 Syntax Checker 2.1 . 3 Disassembler 2.1 . 3 Jbind 2.1 . 3 Jscope 2.1 . 3 NAG libraries 2.1 . 2/3 Jwindows 2.1 . 3 Profiler 2.1 . 2/3 Xinotech Interface - . - Brief Interface - . - A.W.E. - . - CASE - . - Programs will need to be recompiled when switching to 2.1.3 from anything older than 2.1.2. In this case you will need to delete all old .SYM, .SRL, and .JRL files. If you do not do so, the results will be unpredictable. Here is a summary of changes and additions in 2.1.3: 1) Size clauses and Enumeration representation clauses produce better code. 2) Slices of arrays with address clauses produce better code. 3) An enhanced optimizer improves performance and eliminates previous code generation problems. 4) There are three new compiler options: /BM - (Brief Messages) - Use brief compiler error messages, but continue to produce compiler statistics and the long form of other messages. /BS - (Brief Statistics) - Use brief compiler messages and do not print compiler statistics, but continue to use the long form of error messages. The regular /B option continues to do both. /? - Print the values of all options used (including default values). Compilations submitted with bug reports should be run with this option. 5) There are three new linker options: /A - Set Maximum Stack Size. The A option sets the maximum stack size for a program. The stack size includes the heap on many systems, including Ms-Dos. This can be used to change the maximum amount of memory a program can use. For 80x86 Ms-Dos systems, the largest maximum is 64K (10000h). The minimum size is the data size plus 1K. /B - Brief output option. Produces minimal output. /? - Print the values of all options used (including default values). Linkages submitted with bug reports should be run with this option. 6) The linker output shows memory statistics. At the end of its run, the linker prints out the amount of memory the program will use. The report looks like this: Program memory usage - Minimum = 14162 bytes (14K) - Maximum = 77854 bytes (78K) The minimum memory usage number assumes that the program has no explicit allocators (which are indicated by the keyword "new"). If this is not true, the actual minimum might be much greater. For Ms-Dos, the maximum assumes that BigArray and/or Memory are not used. If either of these allocators are present, the actual memory use will be greater than what is reported. 7) There are two new linker error messages: Specified Maximum Stack Size too large The maximum stack size you specified is larger than the maximum allowed. Specified Maximum Stack Size too small The maximum stack size you specified is smaller than the data area for your program; this is not allowed for most systems. 8) JASM86 has a correction for shift and rotate instructions with the /1 and /2 option. 9) There are three new assembler options: /BM - (Brief Messages) - Use brief assembler error messages, but continue to produce assembler statistics and the long form of other messages. /BS - (Brief Statistics) - Use brief assembler messages and do not print assembler statistics, but continue to use the long form of error messages. The regular /B option continues to do both. /? - Print the values of all options used (including default values). Assemblies submitted with bug reports should be run with this option. 10) The runtime libraries CHAINLIB and MATH872 have some corrections. 11) CORDER has some corrections. Lastly, programs which used condcomp, ASM statements, or other /E option features will not compile with the 2.1.3 release compiler. The /C and /E options are NOT supported in the release compilers. Information on using these "compatibility features" can be found on the Janus/Ada BBS system. --------------------------------------------------------------------------- Information for users of older Janus/Ada versions --------------------------------------------------------------------------- The following is provided for users who upgrade from a version older than 2.1.2: Programs which used LONGOPS and LONGIO still can be compiled. However, we recommend updating the programs to use the built-in Long_Integer type since it is much more efficient. Also, Use clauses on LONGOPS can cause problems, as the predefined Long_Integer will take precedence. Thus all uses of the type LONGOPS.Long_Integer must be written with full dot notation. **** Changes to 2.1.2 from 2.0.2 Version 2.1.2 has significant new features, and corrects many bugs to the older compilers. For a complete list of bugs and their corrections, see your newsletter. All programs will need to be recompiled when switching to 2.1.2. Delete all old .SYM, .SRL, and .JRL files. If you do not do so, the results will be unpredictable. 1) Type Long_Integer is now a built-in type. You can use them as literals and array indices, in For Loops and Case statements, and so forth. The type has the normal 32-bit range. The values of System.Min_Int and System.Max_Int have changed appropriately. 2) Fixed point 'Small Length clauses are implemented. The 'Small value specified must be in the range of 2**(-99) .. 2**(99) and be less than or equal to the delta value. The type specified must be a first named subtype. 3) Enumeration representation clauses are implemented. The values specified must be in the range of Long_Integer. The default size of the resulting type is Byte if all of the values are in the range 0..255, Integer if they are in the range of Integer, and Long_Integer otherwise. 4) Size length clauses for Discrete and Fixed types are implemented. Any size between 1 bit and 1000 bits can be specified. Types are allocated to the nearest byte. This can be used to require an unsigned representation for a type. For example, the declarations: Type Byte Is Range 0 .. 255; For Byte'Size Use 8; has the intended effect. Notice that the base type of such a declaration is still 16-bit Integer; this means that math operations and the like are done in 16 bits. Bit packing (as opposed to byte packing) is not done. 5) Task 'Storage_Size clauses are implemented. The default Task Storage_Size has been increased to 2048 bytes. 6) Type System.Address has been changed to the following record: Type Word Is -- Unsigned Integer, sort of. Type Offset_Type Is New Integer; Type Address Is Record Offset : Offset_Type; Segment : Word; End Record; Aggregates of the record type are allowed. Addition and subtraction of Address types are allowed with Offset_Type items. This is defined on the Offset portion ONLY (we don't do segment math - it doesn't work anyway on OS/2 or other protected mode systems.). 'Address now returns the correct values for Labels and Subprograms. 7) System.Byte has been redefined as: Type Byte Is Range 0 .. 255; For Byte'Size Use 8; This mainly means that Integer literals are now compatible with type Byte; it should not require the changing of existing programs. The Type Word is also now a proper unsigned type. 8) The types of parameters to some of the library routines are different. Direct_IO.Count is now a Long_Integer type, allowing access to an entire file of type Character. Util.Memavail, Util.MaxAvail, and Util.StackAvail now return Long_Integer values. You no longer need special processing to find out how much memory is left. Bit.Peek and Bit.Poke now take parameters of type System.Address. This allows access to all memory in a "portable" fashion. Lastly, most registers in DOSCalls are now of type System.Word. This makes it easier to place addresses in registers, although you may need Unchecked_Conversion to put signed Integers into registers. A direct type conversion from Integer to Word will raise Constraint_Error for any negative values. Your source code may need other changes to accommodate the new types. --------------------------------------------------------------------------- Information for Brief editor users --------------------------------------------------------------------------- Janus/Ada is supported by Underware, Inc's Brief 3.0 editor. Those wishing to purchase Brief or update an older copy should contact : Solution Systems 541 Main St., Suite 410 South Weymouth MA 02190 (800) 821-2492 (617) 337-6963