Written by Brooke Anderson
A Free C Compiler for OS/2 2.0
(Note: This document was originally written by Brooke P. Anderson. It has been modified slightly to conform to the .INF format. The original may be obtained from the same source as EMX itself.)
[And now it is in HTML.. -- Ed]
This document describes a free GNU software-development system for OS/2 v2.0. It tells you how to acquire all of the necessary software, how to install it, how to start using it, and where to look for more information.
The GNU software-development system includes a C and C++ compiler, a debugger, an assembler, a make utility (for automating the compilation of programs made of many source files), and a hypertext reader (for reading the documentation). The compiler generates full 32-bit optimized code, and it supports all of the OS/2 API calls, so you software developers out there can use it for PM programming, manipulating semaphores, manipulating threads, using named pipes, etc. Optional packages include curses (a standard library for manipulating screen-fulls of text and for moving the cursor), a collection of sample programs, and full source code for the system.
GNU software is originally developed by the Free Software Foundation, an organization which produces a lot of free software for UNIX. After the Free Software Foundation releases the UNIX versions, people often port them to many other operation systems (such as OS/2). Despite the fact that the software is free, the UNIX community considers it a standard and often prefers it over other products because of its high quality. The compilers, for example, produce well-optimized code.
Sometimes, there is more than one port of a GNU program. For OS/2, there are two different ports of the GNU compiler (called "gcc"). This document discusses only one of them (the EMX port) since the EMX port provides faster floating point routines and works with a debugger. This document deals with version 0.8f of the EMX port, which is the latest version. People frequently produce new versions of and enhancements for the GNU software and the ports based on it.
IMPORTANT: I have not tested the software on FAT file systems. I think it will work, but I have not checked all of the file names to make sure they comply with the egregious 8.3 file-naming convention.
How to Get EMX
The full software-development system (and the various optional packages described later in this document) are available from a variety of sources. If you have access to Internet, you can get the files from anonymous ftp sites. In the USA, the main anonymous ftp site for OS/2 is ftp-os2.nmsu.edu. In Germany, the main anonymous ftp site for OS/2 is rusinfo.rus.uni-stuttgart.de. Also, see the end of this document for a list of people who are willing to distribute the whole system through regular mail.
You need to obtain the following files (the sizes of the files are listed next to the names): emxdev.zip (620k), gnudev.zip (873k), gppdev.zip (944k), gobjcdev.zip (426k), gnumk362.zip (255k), whlinf8f.zip (814k), and unz50x32.exe (114k). This file (emxst31.doc) is available in emxst31.zip (10k).
On ftp-os2.nmsu.edu, these files are available in pub/os2/2.0/programming/emx-0.8f -- except for emxst31.zip, whlinf8f.zip, and the correct version of gnumk362.zip, which are temporarily in pub/uploads. On rusinfo.rus.uni-stuttgart.de, check in pub/os2/emx-0.8f.
How to Install EMX
The following subsections describe how to install the various pieces of the GNU software-development system. Go through the procedures step by step as described - the order is important. Don't feel compelled to read through any of the readme files or other documentation spit out during the unarchiving process - I think you will have a much easier time if you go through this document beforehand.
You will need unzip v5.0 to unarchive the files. Earlier versions of unzip (including PKZip versions earlier than 1.9) will not work.
To unarchive unz50x32.exe, you simply move it to a convenient directory and type "unz50x32". Add the name of the directory unzip is in to your path. For example, I have unzip in c:\apps\unzip, so I appended "c:\apps\unzip;" to the "SET PATH" statement in my config.sys file. Then reboot. Now, you can unarchive any zip archive by typing "unzip filename", where "filename.zip" is the name of the archive.
Copy whlinf8f.zip to any convenient directory and unarchive it. It will disgorge a program called "info.exe" (the hypertext reader), several auxiliary files, and a bunch of documentation files for itself, the compiler, the debugger, and make.
Copy gnumk362.zip to any convenient directory and unarchive it. The archive will create the directory ".\make" and disgorge several files into it. In other words, if you unarchive gnumk362.zip in the directory "\apps", the process will create the directory "\apps\make". You can delete the info directory (the one in the make directory) as its contents are duplicated in whlinf8f.zip.
Add the name of the directory make is in to your path statement in config.sys. (I have make in c:\apps\make, so I appended "c:\apps\make;" to the "SET PATH" statement in my config.sys.)
GNUDEV.ZIP, EMXDEV.ZIP, GPPDEV.ZIP, and OBJCDEV.ZIP
These archives create the directory ".\emx" and a bunch of directories under that. Thus, unarchive the files from the directory in which you want this emx directory. (In other words, if you unarchive gnudev.zip in the directory "\apps", the process will create \apps\emx, \apps\emx\bin, \apps\emx\lib, and many other such directories.)
The Final Steps
Now, you need to modify your config.sys file. The examples below are from my config.sys file, and I have the emx directory installed under c:\apps. Thus, my system has the directories "c:\apps\emx\dll", "c:\apps\emx\lib", "c:\apps\emx\include", "c:\apps\emx\bin", and so on -- you will need to modify the following examples so that the directories are specified correctly.
First, you need to specify in your libpath the location of emx.dll and other DLL files. In my config.sys, I have
Second, add the name of the emx\bin directory to your path statement in config.sys. (I appended "c:\apps\emx\bin;" to the "SET PATH" statement in my config.sys.)
Third, you need to set a few environmental variables so that the compiler knows where to find and how to use various files:
IMPORTANT: you need to use forward slashes ("/") and not backward slashes ("\") when setting these environmental variables. However, do NOT use forward slashes in your libpath statement.
Now, reboot your machine so that these definitions take effect. Then, go into an OS/2 window and type "cd \apps\emx\lib" (or whatever you would type to get to emx\lib), and type "omflib". This completes the installation process.
Go into the directory in which info.exe resides. Type "info". You are now looking at a screen that has some information on the top half (information such as "Typing 'd' returns here, 'q' quits, '?' lists all info commands, 'h' gives a primer for first-timers . . .") and a list of subjects near the bottom (subjects such as "Info", "Make," "Gcc", "Gdb", etc.).
Go ahead and type "h" for a tutorial. The most basic functions to remember are: type "q" to quit, type "?" to get a list of commands, hit the space bar or PgDn key to go down a page, press the Del key or the PgUp key to go up a page, press "n" to go to the next node (think of a node as being a collection of one or more pages dealing with a single topic), press "p" to go to the previous node, use the up and down arrow keys to highlight choices of new nodes to jump to, and press enter to jump to the highlighted node.
Just play around with it a little while, and you will get the hang of it. Type "d" to get back to the main directory, use the down arrow to highlight "Gcc", press enter, press the down arrow to highlight "Contributors", press enter, scan through the pages of text by hitting the PgDn key a couple of times, type "?" to see a list of commands, type "p" a couple of times, etc.
Using the Compiler
To compile a C source file called "myprog.c", type "gcc -o myprog.exe myprog.c". The -o switch tells gcc that it should call the resulting executable "myprog.exe". To compile the C++ source file "myprog.cc", type "gcc -o myprog.exe myprog.cc -lgpp". C++ source files should have the extension ".cc". The -lgpp switch tells gcc to link the C++ libraries. You can also tell gcc to optimize your code by using the -O switch. There are two levels of optimization (-O and -O2, the highest being -O2). Thus, for the fastest-executing code (at the expense of time to compile and of size of the executable), you would type "gcc -O2 -o myprog.exe myprog.c" for myprog.c and "gcc -O2 -o myprog.exe myprog.cc -lgpp" for myprog.cc.
Note: Specifying "-o myprog.exe" is important. If you don't specify ".exe" as the suffix of the output file name, the compiler will generate a UNIX-style executable which will not run under OS/2 even if you subsequently rename the file so that it has a .exe extension.
Using the Debugger
To debug a program, you need to compile it with the -g switch: "gcc -g -o myprog.exe myprog.c" for myprog.c and likewise for myprog.cc. After compiling, you then type "gdb myprog.exe" to start the debugger. Type "h" at the debugger prompt to get a list of help topics. gdb supports all sorts of breakpoints (including conditional ones), and you can watch variables, set variables to different values, etc. For example, to get help on running programs from within the gdb, you can type "help running". That will give you a list of commands such as run, kill, step, etc. You can get help on individual commands by typing, for example, "help step".
The following is a sample compile and debug session. Go into the emx\test directory and type "gcc -g -o hello.exe hello.cc -lgpp" to compile hello.cc. Type "hello" to run it, to see if the compiler is functioning. The program will print "Hello, world!" on the screen. Now type "gdb hello.exe" to start gdb. At the prompt "(gdb)", type "list main" to list the function "main". Then type "break 5" to cause execution to stop at line 5 in the main function. Type "run" to start execution -- it will stop at line 5. Type "print argc" to see the value of the variable "argc". Type "step" to run line 5 then halt execution at the next line. Type "quit" to quit. To list a function (main(), say) that is longer than a screen-full, type "list main"; then type "list" again to list the next screenful of main.
Assume you have a program made of the following source files: "myprog1.c" and "myprog2.c". You might manually compile these files by typing "gcc -o myprog.exe myprog1.c myprog2.c". Of course, if you change only one of the files, typing such a command causes the recompiling of both source files. During development, this could be tiresome if the files take a long time to compile. A better way would be to type "gcc -c myprog1.c" which compiles myprog1.c into the object file "myprog1.o" (the -c switch tells gcc to make an object file), then to type "gcc -c myprog2.c" to generate myprog2.o, and finally to type "gcc -o myprog.exe myprog1.o myprog2.o". This way, if you change only myprog1.c, you can recompile it and relink it with myprog2.o to create myprog.exe (skipping the "gcc -c myprog2.c" step). This will be much faster if the source files take a long time to compile (or if you have a lot of source files).
Of course, doing all this typing is tiresome, too. Also, if myprog1.c happens to depend on myprog1.h, and you change myprog1.h, you must recompile myprog1.c. Thus, you have to keep track of all the file dependencies in order to know, after changing one header file, which other files need to be recompiled.
Fortunately, make takes care of all of this automatically. All you have to do is create one text file that describes the various dependencies and the various steps to compile the program. You name the text file "Makefile". From then on, whenever you type "make", make examines the Makefile, looks for files which have changed since the last compile, recompiles any files which depend on the changed files, and relinks everything into a new executable.
For example, suppose that myprog.exe is made from myprog1.c and myprog2.c, that myprog1.c contains the lines "#include "myprog1.c"" and "#include "mainhead.h"", and that myprog2.c includes myprog2.h and mainhead.h. The Makefile describing all of this is
myprog.exe: myprog1.o myprog2.o gcc -o myprog.exe myprog1.o myprog2.o myprog1.o: myprog1.c myprog1.h mainhead.h gcc -c myprog1.c myprog2.o: myprog2.c myprog2.h mainhead.h gcc -c myprog2.c
The first line shows that myprog.exe depends on myprog1.o and myprog2.o. If either of those has changed since the last time make was invoked, make will relink them to create myprog.exe by giving the command under the first line. The fourth line shows that myprog1.o depends on myprog1.c, myprog1.h, and mainhead.h. If any of these three files have changed since the last time make was run, make will recompile myprog1.o by issuing the command on line five. It will also realize that myprog.o has changed, that myprog.exe depends on myprog.o, and will relink myprog.exe. If mainhead.h is changed, make will recompile and relink everything since myprog1.o needs to be changed, myprog2.o needs to be changed, and thus myprog.exe needs to be changed.
The example above shows the general form of a Makefile. You give a target (like "myprog.exe" or "myprog1.o") followed by a colon, followed by a space, followed by a space-delimited list of files the target depends on. The next line specifies the action to be taken when any of the dependencies change: the first character MUST be a tab (not just a bunch of spaces used for indentation); then you type the command make should issue. A Makefile is just a list of such targets, dependencies, and actions.
Using the Assembler
The compiler provided with this system (gcc) can handle not only C and C++ code but assembly as well. It can also generate assembly language from C or C++ source. To assemble and link assembly-language programs, type "gcc -o myprog.exe myprog.s" where "myprog.exe" is the name of the executable and where "myprog.s" is the name of the source-code file. Assembly-language source-code files should have the extension ".s". To generate assembly from C code, type "gcc -S myprog.c" where "myprog.c" is the name of the C-source-code file -- the resulting assembly language will be put into the file "myprog.s".
Where to Get More Information
The GNU software-development system does not come with documentation like that you would get with, for example, Borland C++. However, emxdev.doc (in the emx\doc directory) does contain a complete list of library functions, including a list of headers you need to include, what the functions do, and what parameters they accept.
Also, since the C compiler is ANSI-C compliant (or at least close to it) and since the C++ compiler is close to AT&T-C++-2.0 compliant, you can use just about any reference manuals for ANSI C and AT&T C++ 2.0. I use the ones I got with an old version of Borland Turbo C++. If you don't have such manuals, you should be able to find something suitable in a bookstore. If you want a C reference manual, I recommend C: A REFERENCE MANUAL, by S. P. Harbison and G. L. Steele, Jr. (Prentice-Hall, 1991). If you are just learning C or C++, there is a large number of books to choose from, and you shouldn't have any trouble finding one that is suitable.
For those of you developing applications that use the PM or that use special OS/2 functions, the system DOES support all of the OS/2 API functions, including ones for semaphores, PM programming, named pipes, threads, etc., and it supports Kbd, Mou, and Vio functions. See emxdev.doc (in the emx\doc directory) for a list of the supported functions. The documentation does not contain a manual on how to use these API calls -- you need an OS/2 programming book for that. For information on programming the PM, take a look at:
The book by Petzold was written for OS/2 1.3, but the information in it is still valid. You can also get the IBM redbooks (which are quite economical and of which there is a large assortment of titles).
For using assembly language, the best choice would be a book about assembly-language programming in OS/2, supplemented perhaps with a book on programming the 80386 or 80486.
Also, way back when you were unarchiving, you might have been itching to examine the various readme files and other documentation. Now is the time to do that to your heart's content. Browse through the files in the emx/doc directory and the information available from the hypertext reader.
Additional sources of information include GEnie (the OS/2 category of the IBMPC bulletin board), Usenet news (the comp.os.os2 newsgroups), and CompuServe. These are places where you can exchange information with other people who use and program for OS/2. In particular, CompuServe is one of the official homes for the OS/2 developers' assistance program. If you are a member of the program, IBM will (for only $15) provide you with a CD that contains a beta version of the software development kit (including the C and C++ compiler and debugger and a full set of on-line documentation), a beta version of OS/2 (which contains enhancements such as Windows 3.1 compatibility), and many other goodies. For more information, type "go os2dap" on CompuServe or call 1-407-982-6408. The people at 1-800-3-ibm-os2 might also be able to provide more information.
I wrote this to help people get started with a free -- yet powerful -- 32-bit software-development system for OS/2. For (at most) the price of a few books, you have a full programming system for C++, C, and assembly language. For the additional price of an OS/2 programming book, you have a bargain-basement SDK.
If you find errors in this document, or if you have suggestions for its improvement, please let me know. My GEnie address is "BROOKE", and my Internet address is "email@example.com".
There are three optional packages you can get for this software-development system, packages which are not necessary but which can nevertheless be important. All of them are in the pub/os2/2.0/programming/emx-0.8f directory on ftp-os2.nmsu.edu. On rusinfo.rus.uni-stuttgart.de, they are probably available in a directory such as pub/os2/emx-0.8f.
The first package contains curses. Curses is a library of functions that allow you to move the cursor around on the screen, manipulate screen-fulls of text, and get input. You need the files "bsddev.zip" and "bsddoc.zip". The source code is available in "bsdsrc.zip".
The second package contains full source code to the software-development system. Most people do not need the source code to everything, but the source code to the libraries (i.e., to all the functions) is sometimes useful. The source code to the C libraries is in "emxlib.zip", and the source to the C++ libraries is in "gccsrc.zip". install.doc (in the emx\doc directory) gives the names of all the other relevant archives.
The third package is a collection of sample and test programs that you can use to test the compiler and to learn about various aspects of programming. These programs are in the file "emxtest.zip."
Sources of Distribution
This document already described some places from which you can get the necessary archives: ftp-os2.nmsu.edu and rusinfo.rus.uni-stuttgart.de. However, some people don't have access to these sites or don't have modems fast enough to download megabytes of data in a reasonable amount of time. For these people, I am including the following list of people who are willing to distribute the whole system through regular mail. Keep in mind that people might change their prices, cease distributing the software, move, etc., so contact them first to get current details. Make sure you ask them if they have the latest version of the EMX port (which is currently version 0.8f).
Brooke Anderson Juergen Egeling Wey J. Ho Doug Robison
Wey J. Ho
If you would like to get on this list and if you have access to Internet or an on-line service, just send me your name, a description of how people can contact you (including your e-mail address), how much money you want for the job (such as "$20", "disks + shipping + $30", or whatever you want to charge), and what you are offering for that price. It is helpful to have a list that includes people in various countries.
Distributing This Document
I give permission to use, to distribute, and to copy this document freely. If you want to upload it to any bulletin-board or on-line service, please do so. I do update this document occasionally (and put the latest version on GEnie and ftp-os2.nmsu.edu), so you might want to make sure you have the latest version before distributing it.
Brooke Anderson 1155 E. Del Mar #312 Pasadena, CA 91106 USA Phone: (818) 577-7555 GEnie: BROOKE Internet: firstname.lastname@example.org