Programming Languages Suite/ProgramD2.iso/Borland/Borland C++ V5.02/HELPME.PAK/HELPME.WRI

Text File | 1997-05-06 | 61KB | 1,493 lines

****************************************
BORLAND C++: ANSWERS TO COMMON QUESTIONS
****************************************

This file contains information about the following issues:

     * Getting Started
     * Exception Handling
     * Other Common C++ Questions
     * Common Windows and ObjectWindows Questions
     * Integrated Environment and Resource Workshop
     * Command-Line Compilers
     * General I/O
     * Graphics (DOS)
     * Math and Floating Point
     * Linker Errors
     * Other Questions

Getting Started
---------------------------------------------------------------
Q. How do I install Borland C++?
A. For help on a wide range of common installation issues, see
   the online document, INSTALL.TXT.

Q. What is a configuration file?
A. A configuration file tells Borland C++ what options to use as
   defaults and where to look for its library and header files.
   BCC.EXE looks for a file named TURBOC.CFG, and BCC32.EXE
   looks for a file named BCC32.CFG. The integrated environment,
   BCW.EXE, looks for the following default configuration files:
   BCCONFIG.BCW, BCWDEF.BCW and BCWDEF.DSW.

Q. How do I create a configuration file?
A. The INSTALL program creates TURBOC.CFG and BCC32.CFG for you.
   These file are ASCII files you can change with any text editor.
   They contain path information for the library and header files
   used by BCC.EXE and BCC32.EXE, respectively. In the case of
   the IDE (BCW.EXE), configuration files are created the first
   time you start the IDE. They are saved automatically each time
   you exit. You can turn this autosave feature off by selecting
   Options|Environment|Preferences. You can also explicitly save
   your options using Options|Save. The options listed on this
   menu and the files they correspond to are as follows:

    Menu Option      File in BIN directory
    ===========      =====================
    Environment      BCCONFIG.BCW
    Desktop          BCWDEF.DSW
    Project          BCWDEF.BCW

Q. What is project file or configuration file corruption?
A. Project file corruption occurs when an error occurs during
   program development and a "bad" portion of memory gets saved
   to disk. This can later cause problems in the IDE. You might
   suspect that your project files have become corrupt if you
   experience strange new behavior either when compiling or
   linking or if you are running very simple programs which
   don't behave as you would expect them to.

   If you suspect that your project files have become corrupt,
   exit the IDE and delete the default IDE configuration files
   that are listed above.

   You'll also want to check the files that the IDE uses for
   your specific project. This is especially true if the trouble
   you're having seems to happen in one project only.
   First, note what files are included in your project and what
   settings they use. Next, make a backup copy of your project
   file (.IDE) so that you can refer to it if needed (use the DOS
   COPY command). Give the backup an extension other than .IDE or
   .PRJ (.BAK is a good extension). Also look in your project
   directory for files with the same base file name as your
   project and with the extension of .BCW or .DSW . Delete or
   back up these files as well.

   Once you have deleted or renamed the default configuration
   files and the files associated with your particular project,
   rebuild your project and try running your program again. If
   you're still having problems contact Borland Technical Support.

Q. How can I prevent project file corruption?
A. One way to help minimize the chance of corruption is to
   turn off the autosave feature by selecting
   Options|Environment|Preferences and unchecking the autosave
   checkboxes. If you do this before you exit the IDE each
   time, you'll need to explicitly save your project and options
   using Options|Save. Don't save the files if you experience
   any problems in the course of development.


Exception Handling
------------------------------------------------------------
Q. How do I prevent the debugger from catching exceptions?
A. Go to the local option of Module View (press Alt+F10 or
   right-click in a Module View window). Press 'X' for
   Exceptions. Press 'N' to turn off C++ exceptions and 'o'
   to turn off C exceptions. Note that both types of
   exceptions must be disabled in order to stop the debugger
   from trapping exceptions.

Q. How do I step into the catch block of an exception?
A. You don't have to do anything special if you step over
   the throw() statement with either F7 or F8. You won't
   be able to step directly into the catch block if you
   step over a function which contains a throw statement.
   What you will have to do is place a breakpoint in the
   desired catch block. Note that this is what happens
   when attempting to step through new--the new default
   behavior of new is to throw an exception, xalloc,
  when an allocation fails.

Q. Why is the new operator behaving differently?
A. The default behavior of new in BC 5.0 is to throw
   an exception, xalloc, if the allocation fails. A NULL
   value is no longer returned. The current error
   checking for new should look something like this:
    char *temp;
    try
    {
         temp = new char[20];
    }
    catch(xalloc)
    {
   // Do error handling here; possibly assign NULL to temp
    }

Q. How do I disable exception handling when using operator new?
A. Refer to the following example:

   #include <iostream.h>
   #include <except.h>
   #include <new.h>

   int
   main() {
       set_new_handler(0);        // comment out to use exceptions
       try {
       char *p;
       while (1) {        // exhaustively allocate heap
           const unsigned k = 1024;
           const unsigned n = 100 * k;
           p = new char[n];
           if (!p)
           break;
       }
       cerr << "detected via pointer value" << endl;
       }
       catch(const xalloc&) {
       cerr << "detected via exception" << endl;
       }
       return 0;
   }

Q. How can I change the default behavior of new for global
   objects? For example, where a global instance of a
   class calls new within its constructor?
A. Global objects are created before execution gets to the
   main function. Set up a startup function which gets
   called before the global objects are created. This
   is done using the #pragma startup statement:
    #include <new.h>
    void set_new(void)
    {
      set_new_handler(0) ;
    }
    // global objects are created with a priority of 32,
    // so use 31 as the priority for the function so that
      // it gets called before global objects are created.
    #pragma startup set_new 31

Q. How do I catch exceptions that are not explicitly
   handled by a catch block so that abort isn't
   automatically called?
A. Add a catch block which has ellipses as its
   parameter to every try block:
    try
    {
      throw(1);
    }
    catch(...)
    {
      cout << "All exceptions are caught" << endl;
    }

   This ensures that there are no unhandled exceptions.

Q. What are some of the predefined exceptions that can
   be thrown?
A. There are a number of exception which are thrown from
   Borland functions (most of which are described in
   the Library Reference):
      xalloc     - thrown from new.
      Bad_cast   - thrown from dynamic_cast
      Bad_typeid - thrown when the operand of typeid
                   is a dereferenced 0 pointer.

Q. How can I display information when an exception
   goes unhandled?
A. The exception handling mechanism calls the
   terminate() function before calling abort(). This
   function does nothing by default, but a new function
   can be set using the set_terminate() function. The
   set_terminate() function takes as a parameter a
   function which returns void and does not take any
   parameters (void). You can display error messages
   within this function:
     #include <except.h>
     #include <new.h>
     #include <iostream.h>

     void term_func(void)
     {
     cout << "This exception went unhandled:" << endl;
     cout << "  Name: " << __throwExceptionName << endl;
       // The following information is only valid if the
       // -xp (Enable exception location Information)
       // compiler switch was set. Under Options|Project|
       // C++ Options in the IDE.
     cout << "  FileName: " << __throwFileName << endl;
     cout << "  LineNumber: " << __throwLineNumber << endl;
     exit(3);
     }

     int main(void)
       {
       set_terminate(term_func);
       return 0;
       }

  Note that you have to call exit() from the terminate
  function in order to prevent the abort() function from
  being called (which in turn displays the "Abnormal Program
  Termination" message).

Q. How can I get information about an exception?
A. The object you throw can contain information that
   can be accessed within the catch block. You can also
   get information about where an exception was thrown
   by the use of certain global variables:
   __throwExceptionName, __throwFileName, and
   __throwLineNumber. These variables are documented in
   the Library Reference. Note that you have
   to have 'Enable Exception Location Information' set
   in order to get the FileName and LineNumber (-xp from
   the command line, Options|Project|C++ Options in the
   IDE.) The following shows a simple method of getting
   additional information from an xalloc exception (which
   is thrown by new):
     #include <cstring.h>
     #include <except.h>

      int main(void)
     {
      char huge *temp;
      try
       {
       temp = new huge char[1000000Ul]; // This will fail in DOS
       delete temp;
       }
      catch(xalloc one)
      {
       cout << "Allocation failed in main" << endl;
       cout << one.why().c_str() << endl;
       cout << "  Name: " << __throwExceptionName << endl;
         // The following information is only valid if the
         // -xp (Enable exception location Information)
         // compiler switch was set. Under Options|Project|
         // C++ Options in the IDE.
       cout << "  FileName: " << __throwFileName << endl;
       cout << "  LineNumber: " << __throwLineNumber << endl;
       }
      catch(...)
      {
       cout << "An unhandled exception occurred" << endl;
      }
      return 0;
     }

   The xalloc class contains a member function, why()
   that returns a cstring object. The cstring class has a
   function, c_str(), which returns a char *.

Q. Why does my application abort and get an "Abnormal Program
   Termination" message?
A. When an exception is thrown within an application but
   isn't caught, first the terminate() and then the abort()
   functions are called. It's the abort() function that
   displays the "Abnormal Program Termination"  message.
   You can use the set_terminate() function to replace the
   existing terminate function (the terminate() function
   does nothing by default, but can be modified to display
   a message). Note that you need to call exit() from
   within the terminate() function if you don't want the
   "Abnormal Program Termination" message displayed. For
   more information, go to the section "How do I display
   information when an exception goes unhandled?"

   Note that the new operator aborts on failure if the
   xalloc exception isn't caught.

Q. Why is my program aborting with the "Abnormal Program
   Termination" message after I have turned off exception
   handling (-x-) for my modules?
A. The behavior of operator new failures isn't changed by
   turning off exceptions within your modules, because an
   exception (xalloc) is still being thrown.  Unless this exception
   is caught, abort() will ultimately be called (hence the "Abnormal
   Program Termination" error...).  To modify what happens when
   operator new fails, use the set_new_handler() function call.
   Calling set_new_handler(0) with zero as an argument disables an
   instance of xalloc from being thrown and 0 will be returned from
   the call to operator new.

Q. Why does my application terminate when throwing an
   exception from within a destructor?
A. What is most likely happening is that the exception
   is thrown while the compiler is cleaning up the stack
   from a previously thrown exception. The compiler
   automatically cleans up objects on the stack when an
   exception is thrown. These objects on the stack can
   have destructors, which are also automatically called
   by the exception mechanism. Throwing an exception at
   any time within this cleanup isn't allowed.

Q. How do I pass an exception up the chain if I have
   nest try...catch blocks?
A. Using throw without any parameters within a catch block
   throws the exception up the chain:
    try
    {
       try
       {
         throw(10);
       }
       catch(int i)
       {
        cout << "Caught exception: " << i << endl;
        throw;
       }
    }
    catch(int i)
    {
      cout << "Caught Exception: " << i << " again." << endl;
    }

Q. Where can I get more information regarding exceptions?
A. Exceptions are covered in the following books:
   "The Borland C++ 5.0 Programmers Guide"
   "The Annotated C++ Reference Manual" by Ellis and
    Stroustrup, from Addison-Wesley Publishing Company.
   "The C++ Programming Language, Second Edition" by
    Bjarne Stroustrup, from Addison-Wesley publishing Company.

Other Common C++ Questions
------------------------------------------------------------
Q. When linking C or Assembly language modules with C++
   modules I get undefined symbol errors at link time. It
   appears that none of the C or Assembly public symbols
   can be found.
A. Because of function and member function overloading prescribed
   by the C++ language, Borland C++ must internally attach
   information to function names and variables (name mangling)
   to maintain uniqueness.  Automatic name mangling generates
   names which do not conform to standard C names.  Name mangling
   can be disabled by using the 'extern "C"' modifier;  any name
   having name mangling disabled will result in a standard C name.
   This is the common manner in which C modules can be linked
   into C++ applications.  For example:

   in C module:
       int foo(char *p)  { /* ... */ }

   in C++ module:
       extern "C" int foo(char*);
       int main() {
           foo("Hello, world!");
           return 0;
       }

   In this example, the code for function foo() is compiled as a
   C module. By declaring the function prototype within the C++
   module as 'extern "C"', Borland C++ will not mangle the name in
   the call to the function.  Upon linking, the linker will be able
   to resolve both the C code and call found in function main().

   See related comments under Linker Errors. There is also
   more on extern "C" in the Programmer's Guide.

Q: How can I allocate a doubly-dimensioned array?
A: You can use either malloc() with C or C++, or the
   new operator with C++:
    malloc():  to create a 2 by 3 character array
        int i;
        char** p;
        p = (char **) malloc(2);
        for (i=0; i<2; i++) p[i] = (char *) malloc(3);
    new:
        int j;
        char** q;
        q = new char* [2];
          for (j=0; j<2; j++) q[j] = new char [3];

Q. Classes with static data members are getting linker
   errors ("undefined").
A. Given the example:

   class foo {
       static int i;
   };

   Remember that this is only a declaration and not a
   definition for integer i.  To allocate memory for i,
   define all static members at file scope with or without
   initialization.  For example:

   int foo::i = 0;
   int main() {
       foo f;
       // ...
       return 0;
   }

   For a definitive discussion, see section 9.4 of the ARM
   or 9.4.2 of the September WP.

   In the case of a template class, you need to similarly define
   static data outside the class definition, and also include the
   actual type information for any type you plan to instantiate the
   template class with. For example:

   template <class T>
   class bar {
       static int i;
   };

   bar<int>::i;      // provide definition for an integer template
                     //  class instantiation type
   bar<Myclass>::i;  // provide definition for a user-defined type
                     // template class instantiation

Q. What potential problems can arise from typecasting a
   base class pointer into a derived class pointer so that
   the derived class's member functions can be called?
A. Syntactically this is allowable. There is always the
   possibility of a base pointer actually pointing to a
   base class. If this is typecast to a derived type,
   the function being called might not exist in the base
   class. Therefore, you would be grabbing the address of
   a function that doesn't exist.

Q: What is the difference between the keywords "struct" and "class"?
A: The members of a "struct" are "public" by default, while
   in a C++ class, they are by default "private".  They are
   otherwise functionally equivalent.

Q: I declared a derived class from a base class, but
   I can't access any of the base class members with the
   derived class function.
A: Derived classes DO NOT get access to private members
   of a base class. In order to access members of a base
   class, the base class members must be declared as either
   public or protected. If they are public, then any
   portion of the program can access them. If they are
   protected, they are accessible by the class members,
   friends, and any derived classes.

Q: I have a class that is derived from three base classes.
   Can I ensure that one base class constructor will be
   called before all other constructors?
A: If you declare the base class as a virtual base class,
   its constructor will be called before any non-virtual
   base class constructors. Otherwise the constructors
   are called in left-to-right order on the declaration
   line for the class.

Q: Are the standard library I/O functions still available
   for use with the C++ iostreams library?
A: Yes, using  #include <stdio.h> functions such as
   printf() and scanf() are available. However, using
   them in conjunction with stream-oriented functions
   can lead to unpredictable behavior.

Q. In C++, given two variables of the same name, one local
   and one global, how do I access the global instance
   within the local scope?
A. Use the scope (::) operator. For example:
       int x = 10;
       for(int x=0; x < ::x; x++)
       {
          cout << "Loop # " << x << "\n";
          // This will loop 10 times
       }

Q. If I pass a character to a function which only
   accepts an int, what will the compiler do? Will
   it flag it as an error?
A. No. The compiler promotes the char to an int and uses
   the integer representation in the function instead of
   the character itself. The exception here is in the
   case of a template function where no implicit
   conversions or promotions take place.

Q. I was trying to allocate an array of function pointers
   using the new operator but I keep getting declaration
   syntax errors using the following syntax:
   new int(*[10])();   What's wrong?
A. The new operator is a unary operator and binds first
   to the int keyword producing the following:
    (new int) (*[10])();
   You need to put parentheses around the expression to
   produce the expected results:
    new (int (*[10]()); //array of function pointers

Q. What are inline functions? What are their advantages?
   How are they declared?
A. An inline function is a function that gets textually
   inserted by the compiler, much like a macro. The
   advantage is that execution time is shortened because
   linker overhead is minimized. They are declared by using
   the inline keyword when the function is declared:
  inline void func(void) {cout << "printing inline function \n";}
   or by including the function declaration and code body
   within a class:
     class test
     {
     public:
     void func(void) {cout << "inline function within a class.\n"}
     };

Q. If I don't specify either public or private sections
   in a class, what is the default?
A. In a class, all members are private by default if neither
   public nor private sections are declared.

Q. If I don't specify either the public or private keyword when
   inheriting from a base class, what is the default?
A. In Borland C++ 2.0, the default was public inheritance,
   but in versions 3.0 through 5.0, the default is private
   inheritance.

Q. What does the _seg modifier do?
A. Using _seg causes a pointer to become a storage place for a
   segment value, rather than an offset (or a segment/offset).
   For instance, if "int _seg *x" contains the value 0x40,
   then when you use "*x", the value pointed to will be at
   segment 0x40, offset 0. If you add a value to the pointer,
   the value is multiplied by the size of the pointer type. That
   new value is used as an offset and is combined with the
   segment value contained in the pointer. For example,
       int _seg *x;
       int value;

       x = (int _seg *)0x40;
       value = *(x + 20);

   value is assigned the value of the integer at 0x40:0x28
   (Remember, 20 * sizeof(int) = 40 = 0x28).
   You can find a more detailed description of _seg in the
   Borland C++ DOS Reference.

Q. Can I statically allocate more than 64K of data in
   a single module?
A. Yes. Far data items are now supported:
       ...
       char far array1[60000L];
       char far array2[60000L];
       ...
    For arrays larger than 64k use:
       char huge array3[100000L];

Q. What is a friend member function?
A. Declaring a friend gives non-members of a class access
   to the non-public members of a class.

Q. Why do I get a "Type name expected" error on my definition
   of a friend class in my new class?
A. You need to let the compiler know that the label you use
   for your friend class is another class. If you don't
   want to define your entire class, you can simply have
   "class xxx", where xxx is your label. For example:
     class Myclass1; //forward class declaration is required

    class Myclass2{
          friend class Myclass1;
          ...
    };

Q: How can I output hex values in uppercase using the
   iostream libraries?
A: You need to set the state of the stream using setf().
   For example,
      #include <iostream.h>
      int main()
      {
        cout << hex;
        cout << "\nNot upper-case : " << 255;
        cout.setf(ios::uppercase);
        cout << "\nUppercase     : " << 255;
        return 0;
      }

Q: The following program compiles/links successfully in
   C but not in C++. Why?
    #include <stdlib.h>

    int compare(const int *one, const int *two)
    {
       if (*one > *two)
          return  -1;
       else
          return 1;
    }
    int a[3] = { 50, 10, 20 };

    void main()
    {
         qsort(a, 3, sizeof(a[0]), compare);
    }
A: The fourth parameter to compare is the function
   pointer, and here's how it's declared in stdlib.h:
   void qsort (void *__base, size_t __nelem, size_t __width,
           int _Cdecl (*__fcmp)(const void *, const void *));

   However, the above program WILL NOT compile in C++,
   because of the strong typing features of the C++
   language. The compiler refuses to convert the void
   parameters in the declaration to __fcmp function to
   int parameters. However, because C++ permits casting
   of function pointers, you can fix the call to QSORT
   in C++ like this:
    qsort(a, 3, sizeof(a[0]),
          (int (*)(const void *,const void *))compare);

   By casting the COMPARE function to be of the same type
   as the declaration in stdlib.h, C++ will accept and
   compile it.

Q. What is the "this" pointer?
A. "this" is a local variable in the body of a non-static
   member function. It is a pointer to the object for
   which the function was started. It cannot be used
   outside of a class member function body.

Q. Why does a binary member function only accept a
   single argument?
A. The first argument is defined implicitly.

Q. Looking through the class libraries there are
   definitions in classes which look like:
       class test {
           int funct(void) const;
       };
    What is the const keyword doing here?
A. Using the keyword "const" after the parameter list
   designates that the member function is to be a constant
   member function.  Constant member functions should not
   modify class data, and the compiler should verify that no
   attempt is being made to change the value of a member
   within the scope of the member function.

Q. Is there any object similar to "cout" that will receive
   all printer output?
A. According to the C++ language, no.  However, you can create
   your own object that will be connected to the standard
   DOS printer handle 4 as follows:

   #include <fstream.h>
   #include <stdio.h>
   ofstream cprn(fileno(stdprn));
   int main() {
       cprn << "Hello, world!" << endl;
       return 0;
   }

Q: How can I use _new_handler and set_new_handler?
A: Borland C++ supports _new_handler and set_new_handler.
   You can find a discussion of them in the
   Borland C++ Programmer's Guide. The type of _new_handler
   is as follows.
      typedef void (*vfp)(void);
      vfp _new_handler;
      vfp set_new_handler(vfp);

Q: I would like to use C++ fstreams on a file opened
   in binary mode. How is this done?
A: Use ios::binary as the open mode for the file:
      #include <fstream.h>
      ifstream binfile;
      binfile.open("myfile.bin", ios::binary);

Q: How can I get at the DOS file handle associated
   with my iostream?
A: Using a combination of member functions fd() and
   rdbuf() you can get at the file handle.
      #include <fstream.h>
      #define fstrno(s)  (((s).rdbuf())->fd())
      ifstream test("test.txt");
      cout << "handle is " << fstrno(test) << '\n';

Q: How can I increase the number of FILES available to
   my program under Borland C++ 5.0?
A: Increasing the number of available files involves
   changing the following 3 files located in the Run-Time
   Library: _NFILE.H, FILES.C, and FILES2.C. For
   instructions on how to do this, download TI870 from
   the Borland fax line.

Q: When using the BIDS library, if I try to create an
   Array or Bag of integers, I get the error "multiple
   declaration for detach()...".
A: If you try to create a TI_ArrayAsVector<int>,
   TI_SArrayAsVector<int>, or TI_BagAsVector<unsigned>,
   the compiler yields an error message. This is because
   each of these templates provides a version of detach()
   that takes an argument whose type is the type for
   which the template is being instantiated, plus a
   version that takes an argument of type int or
   unsigned int, which lets you specify the index for the
   item to be detached. Therefore, when instantiated with
   an integer, these two versions have the same signature,
   which is what causes the error. Integer types aren't
   allowed here due to resulting ambiguity between the
   type and the index of the Bag or Array.

Q: How come bioscom() doesn't work on my computer?
A: The bioscom() function uses DOS interrupt 0x14
   directly, and thus bioscom()'s functionality is
   tied directly to the BIOS of your computer. MS-DOS
   support for the serial communications port might be
   inadequate in several respects for high-performance
   serial I/O applications. First, MS-DOS provides no
   portable way to test for the existence or status of
   a particular serial port in a system. If a program
   "opens" COM2 and writes data to it, and the physical
   COM2 adapter isn't present in the system, the
   program may simply hang. Similarly, if the serial port
   exists, but no character has been received and the
   program attempts to read a character, the program
   hangs until one is available. There is no traditional
   function call to check if a character is waiting.
   MS-DOS also provides no portable method to initialize
   the communication adapter to a particular baud rate,
   word length, and parity. An application must resort
   to ROM BIOS calls, manipulate the hardware directly,
   or rely on the user to configure the port properly with
   the MODE command before running the application that
   uses it. Because of all the problems mentioned above,
   we strongly recommend getting a third party
   communications package when attempting to do
   serial communications, or downloading the
   example program SERIAL.ZIP from our BBS at
   [phone].

Common Windows and ObjectWindows Questions
------------------------------------------------------------
Q: How do I convert my old OWL 1.0 to OWL 2.5?
A: You can use a utility called OWLCVT to help you do this.
   This utility is documented in the ObjectWindows for C++
   Programmer's Guide.

Q: How do I get OWL 1.0 to work with BC 5.0?
A: The on-line file, COMPAT.TXT, provides information on
   doing this.

Q. Why isn't my DLL working correctly?
A. One possibility is that you are not linking in the
   correct 'cw' library. If you are building a small
   model DLL, you should be using cwc, not cws. If you are
   building a medium model DLL, you should be using cwl,
   not cwm. Compact and large models should use cwc
   and cwl respectively.

Q. Why isn't my program working correctly?  I'm getting a
   message box from Windows saying "Unrecoverable
   Application Error".
A. One possible answer is that the program was not built
   correctly. For example, the linker didn't get the correct
   information to export functions. Diagnostic messages from
   the linker could indicate this. To check that you have
   built your program on DLL as expected, review the
   section in the Programmer's Guide that deals
   with exports. This section has a table with eight
   columns describing the possible combinations you might
   have used to build your program. If the setup of your
   program corresponds to one of the last three columns,
   chances are that your program was not built correctly
   (or, at least, as you intended). Column 5 corresponds
   to the 'classical' method of building Windows programs.
   That is, all exports are declared in the module
   definition file (the .def file).

   The columns that use -WE or -WDE will build 'better'
   code in the sense that the compiler won't make 'exportable'
   any functions that it doesn't actually export.
   However, it is here that many people run into
   problems. If you have any functions declared as exports
   in the .def file but the module is compiled with
   -WE or -WDE, then you probably have built the program
   incorrectly (the function will be exported only if it
   is preceded by _export in the source code).

Q. How do I use the _export key word?
A. Put the "_export" immediately before the function name
   in the function declaration to export that function.
   Here is a quick example:
      long FAR PASCAL _export func(void) {...}

Q. I run BCW.EXE and get the error message:
     Fatal: <filename>.def (<line #>): syntax error
A. Check your DATA statement on line number # in
   <filename>.DEF for the correct code (DATA PRELOAD).

Q. Why do I get a 'suspicious pointer conversion'
   warning or 'cannot convert' error (in C++ code)
   when I try to use the address of my window
   procedure or call back function?
A. Windows 3.1 has introduced a new type, WNDPROC, which
   takes the place of the FARPROC type in some cases,
   such as the data type of the lpfnWndProc member of
   the WNDCLASS structure. If you are getting a warning
   or error when setting up a WNDCLASS structure that
   is passed to RegisterClass(), use a WNDPROC cast to
   resolve the type mismatch. If you were previously
   using a FARPROC cast, simply change it to a WNDCLASS
   cast. For example,
    void FAR PASCAL f(void);
    WNDCLASS wcTemp;
    wcTemp.lpfnWndProc = f;
                // Warning in C or error in C++
    wcTemp.lpfnWndProc = (FARPROC)f;
                // Windows 3.0 style type cast
    wcTemp.lpfnWndProc = (WNDPROC)f;
                // Windows 3.1 style type cast

Q. How can I use the features of Windows 3.1 in my
   applications?
A. There are a number of examples that use the
   features of Windows 3.1, such as OLE, DDEML
   and Common Dialogs. There are also lengthy
   descriptions of programming techniques for the
   features of Windows 3.1 in the online Help files.

Q. Why don't some of the Windows 3.1 API functions, such as
   ChooseColor(), do anything when I call them?
A. The Windows 3.1 functions that take structures as
   parameters require that the size field of the structure
   be initialized to the size of the structure. This
   technique allows for backward compatibility in future
   versions of these functions. If the size field isn't set
   correctly the function won't do anything. For example,

   CHOOSECOLOR ccTemp;                // Data structure
   ccTemp.lStructSize=sizeof(ccTemp); // Set the size first!
   if(ChooseColor(&ccTemp)!=0) etc... // Then call the function

Q. Why is my DDEML application crashing?
A. DDEML can crash in seemingly random ways if the
   conversation handle or application instance identifier
   is incorrect or corrupted. Use TDW to watch the value
   of the conversation handle and the application instance
   identifier. If it changes, is corrupted, or you
   inadvertently pass the DDE Management Library an invalid
   value, that particular call might not fail but DDEML
   might become unstable and crash at some time in the
   future. Also note that before DdeInitialize() is called
   for the first time, the application instance identifier
   argument MUST be set to 0.

Integrated Environment and Resource Workshop
------------------------------------------------------------
Q:  How do I use Source Pools?
A:  A sample project called srcpool.ide in the
    \examples\ide\srcpool subdirectory shows you how.

Q:  How can I create multitarget programs?
A:  An example project called multitarg.ide in the
    \examples\ide\multitarg subdirectory show you how.

Q:  How can I take advantage of Style Sheets?
A:  A sample project called stylsht.ide in the
    \examples\ide\stylsht subdirectory shows you how.

Q. Why can't Borland C++ find any of my #include files?
A. The compiler searches for include files in the Include
   Directories path. You can specify this path by selecting
   Options|Project|Directories. The INSTALL program initially
   sets this path to the directory where it copied all the
   Borland C++ *.h files.

Q. How do I get Borland C++ to link in my own libraries
   or use multiple source files?
A. Borland C++'s Project Manager is designed to let you
   work with multiple files.

Q. Why does Borland C++ report "Unable to open include
   file 'stdarg.h'" when I try to #include <stdio.h>?
A. The most probable reason is that you need to check to
   ensure you have correctly specified the path to your
   include directories by selecting Options|Project|
   Directories and examining the Source
   Directories|Include section.

   One other possible reason is that you have exceeded the
   number of files that DOS can have open simultaneously.
   Add the line  FILES=40  to your DOS CONFIG.SYS file.
   This lets DOS open up to 40 files at the same time.
   CONFIG.SYS will only be effective after you have
   rebooted your computer. See the IBM DOS Reference
   Manual for details on the CONFIG.SYS file.

Q. When I Make, Run, or Trace a program, Borland C++
   sometimes goes through the compile and link process
   even when the object files are up-to-date.
A. Borland C++'s MAKE logic works solely on a file's date
   and time stamp. If one of your source files is marked
   with a date that's sometime in the future, the object
   files that are created from it will always be older
   than the source file, and Borland C++ will always try
   to rebuild the file. You can fix this by using
   TOUCH.COM to set the file to the current date
   and time. You should also make sure that your system's
   date and time are always properly set. TOUCH.COM is
   documented in the User's Guide chapter on MAKE.

Q. How can I convert my earlier project files to the new
   IDE format?
A. Using Project|Open, select "3.1 Project Files (*.prj)"
   into "List Files of Type" drop-down box. Select the 3.1
   project you want to run--the project is automatically
   converted into *.ide format. The original *.prj file
   remains unchanged. Be sure to save the new *.ide file.

Q. How can I find out where my "null pointer assignment"
   is occurring?
A. Set a watch on the following expressions:
            *(char *)0,4m
            (char *)4

   Step through the program. When the values change, the
   just-executed line is the one causing the problem.

Q. When I try to load a new file after editing a file,
   the first file remains on the screen. How do I close
   the first file?
A. Use Ctrl-F4 to close the current file.

Q. I'm doing a search and replace operation and the editor
   prompts me for each replacement. I've selected "Change
   All", but it still does it.
A. To disable the prompting, you must unselect the
   "Prompt on replace" option on the left side of the
   Search dialog box.

Q. When I try to use any of the pseudo registers, like
   _AX, I get the error message "Undefined symbol '_AX' in
   function..." when I compile. Why?
A. You are only allowed to use the pseudo registers in
   the Borland C++ and ANSI modes of the compiler. You can
   change this setting by selecting Options|Project|
   Compiler|Source.

Q: How do I stop all of the files I have ever edited
   from constantly being open when I start Borland C++?
A: By default, Borland C++ saves what is called the
   desktop configuration. This configuration is saved
   in a file with a .DSW extension. By deleting any files
   of this type (usually located in the current directory
   and/or the BC5\BIN directory), then entering
   Options|Environment|Preferences and unchecking the
   'auto save desktop' option, you'll begin with a
   clean desktop each time you start Borland C++.

Q: How do I view 32-bit registers in the debugger?
A: Add the appropriate 32-bit register pseudovariable
   to the Watch window. For example, if you were
   interested in the value of EAX, you'd add _EAX to
   the Watch window.

Q: Why does Resource Workshop say "Cannot open file" on
   a header file I included in my .RC or .DLG file?
A: Either the file doesn't exist, or Resource Workshop
   doesn't know what directory to look for it. See if
   the file exists. If it does, then add the directory
   name to the Include Path under File|Preferences.

Q: Why is the Include Path disabled when I look under
   File|Preferences in Resource Workshop?
A: The Include Path is disabled when a project is
   opened. Close the project and then change the path.

Q: In Resource Workshop, why can I select only bold fonts
   when I'm in the dialog editor?
A: This is a limitation of Windows, not Resource Workshop.
   Currently, Windows allows only bold fonts in dialog
   templates. You can send WM_SETFONT messages while
   your program is running if you want other fonts or styles.

Q: Why does my dialog paint funny when I select Test mode?
   The menu and/or the minimize/maximize buttons don't
   draw correctly.
A: Again, this is a limitation of Windows. There are
   certain styles for a dialog box that shouldn't be
   combined. Namely, the Modal Frame style can cause
   painting problems when combined with the Min/Max
   buttons or menus.

Q: I did everything the documents suggested, but
   Resource Workshop still doesn't work or behaves
   strangely. Why?
A: Look for any .RWS files and delete them. These files
   are maintained by Resource Workshop and can cause
   unusual behavior in Resource Workshop if they become
   corrupted or out-of-sync with the .RC or .DLG files.

Command-Line Compilers
------------------------------------------------------------
Q. Why can't Borland C++ find any of my #include files?
A. The compiler searches for include files in the Include
   Directories path. You specify this path with the -I
   option or with Options|Project|Directories in the IDE.
   The INSTALL program initially writes a configuration
   file (TURBOC.CFG for BCC.EXE, BCC32.CFG for BCC32.EXE)
   that sets this path to the directory where it copied all
   the Borland C++ *.h files. You can edit this file to
   change the default path, or create this file in your
   current working directory and place any relevant paths
   and other command line compiler options in it. A sample
   .cfg file might contain the following:
     -ml
     -IC:\BC5\INCLUDE;C:\BC5\CLASSLIB\INCLUDE
     -LC:\BC4\LIB;C:\BC5\CLASSLIB\LIB

Q. Why does the linker tell me that all the graphics
   library routines are undefined?
A. BCC won't search the graphics library unless you
   tell it to. You should specify the graphics library
   on the command line. For example, to compile BGIDEMO,
   type BCC BGIDEMO.C GRAPHICS.LIB and press Enter.

Q. I run BCC.EXE and get the error message:
   Fatal: <filename>.def (<line #>): syntax error
A. Check your DATA statement on line number # in
   <filename>.def for the correct code (DATA PRELOAD).


General I/O
------------------------------------------------------------
Q. The '\n' in cprintf() doesn't return the cursor to the
   beginning of the line--it only moves it down one line.
A. cprintf() interprets '\n' as a Line Feed. To force the
   cursor to the beginning of the line, manually insert a
   Carriage Return:
      cprintf("\n\r");

Q. How do I print from within a Borland C++ program?

A:  The answer depends upon two issues:
      1.  What kind of printer is being used?
      2.  What environment is targeted by the application?
   As for printers, the answers below assume the printer  is
   accepting simple ASCII sequences.  If you are wanting to print
   graphic images and/or using a specialized printer such as PostScript,
   please consult the printer documentation.

   For DOS-targeted applications, you can route the output through
   the standard DOS file handle 4, or mnemonically, 'stdprn'.  For a C
   program, the code would look as follows:

   #include <stdio.h>
   int main()
   {
       fprintf(stdprn, "Hello, world!\n");
       return 0;
   }

   For a C++ program, you would code it as follows:

   #include <fstream.h>
   #include <stdio.h>
   int main()
   {
       ofstream cprn(fileno(stdprn));
       cprn << "Hello, world!" << endl;
       return 0;
   }

   For EasyWin applications, please consult TI1717 located on Borland's
   ftp site, TechFAX, BBS, or CompuServe.  Please consult the "Borland
   Assist Support Reference" for more information on these locations.

   For Windows programs, please consult Windows programming
   documentation on the subject of device contexts.

   In addition, Borland C++ uses a FILE pointer (stdprn) defined
   in the STDIO.H file. You do not need to open stdprn before
   using it:
       #include <stdio.h>
       int main(void)
       {
           fprintf(stdprn, "Hello, printer!\n");
       }

   Note that if your printer is line-buffered, the output
   is flushed only after a '\n' is sent.

Q. I'm reading and writing binary files. My program is
   translating the Carriage Return (0x0D) and Line Feed
   (0x0A) characters. How do I prevent this from happening?
A. Files opened in text mode will translate these characters
   for DOS. To read a file in binary mode, open it in binary
   mode. For example,
     #include <stdio.h>
     int main(void)
     {
       FILE *binary_fp;
       char buffer[100];

       binary_fp = fopen("MYFILE.BIN", "rb");
       fread(buffer, sizeof(char), 100, binary_fp);
                    :
     }

   The default file mode is text.

Q. Why don't printf() and puts() print text in color
   in a DOS application?
A. Use the console I/O functions cprintf() and cputs()
   for color output, as shown here:
     #include <conio.h>
     int main()
     {
        textcolor(BLUE);
        cprintf("Hello, world!");
        return 0;
     }

Q. How do I print a long integer?
A. Use the "%ld" format:
      long int l = 70000L;
      printf("%ld", l);

Q. How do I print a long double?
A. Use the "%Lf" format.
      long double ldbl = 1E500;
      printf("%Lf", ldbl);


Graphics (DOS)
------------------------------------------------------------
Q. Why do I get the error message:
   BGI Error: graphics not initialized (use 'initgraph')
   when I use a graphics function? My program has already
   called initgraph().
A. For some reason initgraph() failed. To find out why,
   check the return value of graphresult(). For example:
      #include <graphics.h>
      int main(void)
      {
        int gerr;   /* graphics error */
        int gdriver = DETECT, gmode;

     // Initialize graphics using auto-detection and look
     // for the .BGI and .CHR files in the C:\BC5\BGI
     // directory.
        initgraph(&gdriver, &gmode, "C:\\BC5\\BGI");

        if ((gerr = graphresult()) != grOk)
        {
            printf("Error : %s\n", grapherrormsg(gerr));
            exit(1);
        }
               :
      }

Math and Floating Point
------------------------------------------------------------
Q. Why do I get incorrect results from all the math library
   functions like cos(), tan() and atof()?
A. You must #include <math.h> before you call any of the
   standard Borland C++ math functions. In general, Borland
   C++ assumes that a function that isn't declared returns
   an int. In the case of math functions, they usually
   return a double. For example
   /* WRONG */                       /* RIGHT */
                                     #include <math.h>
   int main(void)                    int main(void)
   {                                 {
    printf("%f", cos(0));             printf("%f", cos(0));
   }                                 }

Q. How do I "trap" a floating-point error?
A. See the signal() and matherr() functions in the Borland
   C++ Library Reference. The signal() function might be
   used to trap errors in the 80x87 or the 80x87 emulator.
   The matherr() function traps errors in the Math Library
   functions.

Linker Errors
-------------------------------------------------------------
Q. I am linking C functions in a C++ application. The linker
   reports that all of my C functions are undefined. Why?
A. To facilitate function overloading as a feature in C++,
   the compiler takes the name of the function along with the
   types of its parameters, and creates an internal mangled name
   to differentiate this function from another by the same name but
   with different arguments.

   This same mangled name is also used by the linker.
   Functions compiled as C code are not mangled in
   the same manner as functions compiled as C++ code (Use
   the TDUMP utility to determine the exact names involved.).
   The linker cannot resolve these differences;  it is up to the
   programmer to programmer to modify how name mangled occurs.

   For the case where C functions are required within a C++
   program, simply change to function prototypes with the
   modifier extern "C".  That is, assume that the following
   function prototypes denote the function signatures for two
   functions compiled as C code:

   char *StringCopy(char*, char*);
   void ClearScreen(void);

   To make these functions callable within a C++ application,
   instruct the C++ compiler to not mangle their names when they
   are prototyped in a C++ module.  For example:

   extern "C" char *StringCopy(char*, char*);
   extern "C" void ClearScreen(void);

   Likewise, both prototypes can be placed in a single block:

   extern "C" {
       char *StringCopy(char*, char*);
       void ClearScreen(void);
   }

   As for the other case where C code is calling C++ functions,
   the programmer may not have many options.  In many cases, the
   programmer can use the extern "C" modifier in the definition of
   the C++ functions.  This, however, will effect whether
   overloaded functions can still exist (Name uniqueness may be lost.).
   Another alternative is to write a simple assembly language module
   that will translate between the C function names and the C++
   mangled function names (which are not legal C names).

Q. Why do I get the following message when creating a DOS
   application:
   "Linker Error: Unable to open input file 'C0x.OBJ'"
A. The linker searches for Borland C++ start-up and library
   files in the Borland C++ Library Directories path. You can
   specify this path by selecting the Options|Directories.
   The INSTALL program initially sets this path to the
   directory where it copied the start-up and library
   files. Also be sure that you installed the memory model
   that the linker is looking for. The 'x' in the error
   message corresponds to the memory model, e.g. 's'
   for small, 'l' for large, etc.

Q. Why do I get the following message when creating a DOS
   application:
   Linker Error: Undefined symbol '_main' in module C0
A. Every C program must contain a function called main().
   This is the first function executed in your program.
   The function name must be all in lower case. If your
   program doesn't have one, create one. If you're using
   multiple source files, the file that contains the
   function main() must be one of the files in the Project.
   Note that an underscore character '_' is prefixed to
   all external Borland C++ symbols.

Q. Why does the linker tell me that all the graphics
   library routines are undefined in my DOS application?
A. See the "Integrated Environment" and "Command-line
   Compiler"  sections above.

Q. What is a 'Fixup overflow'?
A. This usually means you're attempting to link object
   files that weren't all compiled under the same memory
   model.  Publication TI1150, "Coping with Fixup
   Overflow Messages", provides more information. It is
   available as TI1150.ZIP on Borland's Download BBS
   [phone], and as document # 1150 on Borland
   Techfax [phone].

Q. I'm linking my own assembly language functions with
   Borland C++. The linker reports that all of my
   functions are undefined.
A. Make sure that you have put an underbar character '_'
   in front of all assembly language function names to
   be called by Borland C++. Your assembly language
   program should be assembled with Case Sensitivity.
   If compiling as C++ (rather than C), see the "Common
   C++ Questions" section above for a discussion of
   extern "C".

Q: I'm getting an error from the linker "segment group
   exceeds 64K :  _text" in my DOS application.
A: If you're using the BGIOBJ utility, the default segment
   into which the objects will be place is _text. You
   should try using BGIOBJ with the /f option to place
   the resultant objects into a separate segment.
   You'll then need to use the functions registerfarbgidriver
   and registerfarbgifont to register the objects for the
   graphics system. See UTILS.TXT for instructions on
   using these functions. In addition, publication TI703,
   "Resolving 'Segment or Group xxxx Exceeds 64K',
   provides more information on this issue. It is availble
   as TI703.ZIP on Borland's Download BBS [phone],
   and as document #703 on Borland Techfax [phone].

Q: Why am I getting the error "printf: floating point
   formats not linked"?
A: You probably have your libraries out of order on the
   TLINK command line. (See the User's Guide for TLINK
   syntax.) For example, if you are using the large
   memory model and BGI routines, the TLINK line might
   look like the following:
     tlink /v c0l myobj,,,mylib graphics emu mathl cl
   If the object file or library isn't in the current
   directory, the complete pathname must be supplied.
   Frequently this causes the command line to exceed
   128 characters (you'll need to use a response file--
   See the User's Guide).

Q. I'm porting an application that uses communal
   variables to C++. I've set up the compiler to
   recognize them, but I still get linker errors:
   Error: <name> defined in module <a> is duplicated
   in module <b>
A. C++ doesn't support explicit COMDEFs; you must use
   static variables or switch to C.

Other Questions
------------------------------------------------------------
Q. I get a "floating point formats not linked" message when
   I run my program. What can I do about it?
A. Floating point formats (for scanf() and related functions)
   aren't always linked, for savings in executable size. To
   force their inclusion, put the following somewhere in your
   source files:
      extern int _floatconvert;
      #pragma extref _floatconvert

Q. How do I change the stack size?
A. In a DOS program, the size of the stack of a Borland C++
   program is determined at run time by the global variable
   _stklen. For example, to change the size to 10,000 bytes,
   include the following line in your program:
      extern unsigned _stklen = 10000;
   This statement must not be inside any function definition.
   The default stack size is 4,096 bytes (4K), and you might
   increase the stack to 65519 (0xFFEF) or just under 64K
   in the compact, large, or huge memory models.

   In a Windows Program, the size of the stack is controlled
   by the STACKSIZE line of the module definition (.DEF) file.
   See the Programmer's Guide and the User's Guide for
   more information on module definition files.

Q. I'm getting a 'Stack Overflow!' message when I run my
   program. How can I work around this?
A. If you are using the compact, large, of huge memory models,
   you might increase the stack size by following the
   procedure above. In the smaller memory models, your only
   option is to decrease the amount of stack space or near
   heap space used in your program. Stack overflows are
   usually caused by a large amount of local data or
   recursive functions. You can decrease the amount of stack
   space used in several ways:
   1) By declaring your local variables static (see the
      Programmer's Guide for the effects of using the
      "static" keyword):
        int main(void)                int main(void)
        {                             {
         char x[5000];     -->          static char x[5000];
            :                                :
        }                             }
   2) By making your variables global rather than local:
        char x[5000];    //global allocation above main()
        int main(void)
        {
               :
        }
   3) By allocating your variables dynamically off the
      far heap:
      #include <alloc.h>
      int main(void)
      {
       char far* x;
       x = (char far*)farmalloc(5000);  //dynamic allocation
       // or in the case of C++ you can use the new operator
       // x = new char[5000];
           :
      }

Q. My program comes up with the message 'Null pointer
   assignment' after it terminates. What does this mean?
A. Before a small-data model Borland C++ program returns
   to DOS, it checks to see if the beginning of its data
   segment has been corrupted. This message warns you that
   you have used uninitialized pointers or that your
   program has corrupted memory in some other way.

Q. Why do I get "declaration syntax error" messages on dos.h?
A. You have set the "ANSI keywords only" option ON. Keep this
   option OFF when using any keywords specific to Borland C++.
   See the Programmer's Guide for a list of keywords.

Q. I get errors when compiling the windows.h header file. Why?
A. Be sure that you have "Borland Extensions" selected as your
   keywords option. This option can be toggled under
   Options|Project|Compiler|Source in the IDE. It is on by
   default in the command-line compilers (-AT or -A-).

Q. I have a working program that dynamically allocates memory
   using malloc() or calloc() in small data models (tiny,
   small, and medium). When I compile this program in large
   data models (compact, large, and huge), my program hangs.
A. Make sure that you have #include <alloc.h> in your program.

Q. I'm linking my own assembly language functions with Borland
   C++, but the linker reports that all of my functions are
   undefined. Why?
A. See answer above in the "Linker" section.

Q. My far pointers "wrap around" when they are incremented
   over 64K. How do I reference a data object that is
   greater than 64K?
A. Use huge pointers.

Q. How do I get Clipper to link with Borland C++?
A. If you have trouble, contact Nantucket Technical Support.

Q. I'm trying to build an app based on one of Borland's
   libraries (ObjectWindows, the container classes
   in the CLASSLIB directory, or the Run-time Library),
   and I get linker errors, or it won't run right.
   What's going wrong?
A. You might be using a switch that affects linkage in
   your files that wasn't used when the library itself
   was compiled, or you need to change the library in
   question. Here are some examples:
   - If you use far vtables (-Vf or Options|Project|
     Compiler|C++|Far virtual tables) to compile a file
     you developed which includes iostream.h, it won't
     build correctly until you rebuild the iostream
     library with the same option.
   - If you opt to use the templates implementation of
     the container class library to build ObjectWindows
     applications, you must rebuild the necessary
     ObjectWindows libraries from source using the
     templates implementation of the class library
    (the BIDxxxx.LIB files.)

Q. I got a "bad call to intrinsic function" message when
   compiling one of my source files. What does this mean?
A. This message appeared because you tried to use an
   intrinsic function in a small model DLL. Either avoid
   using intrinsic functions in your DLL, or turn off
   the -Oi and -O2 switches (Options|Project|
   Optimizations Speed|Inline Intrinsic Functions).

Q: I open up a file in append mode and append some data
   to the end of the file. When I look at the data in an
   ASCII editor, I can't see the appended data. Why?
A: The data is being appended after the End-of-File mark,
   and the ASCII editor isn't displaying the data after
   the EOF mark. To eliminate the EOF mark:
   1) Get the file length with the filelength() function:
       FILE *file_pointer = fopen("file.nam","a");
       long length = filelength(fileno(file_pointer));
   2) Use the chsize() function to change the file
      length to the current length-1:
       chsize(fileno(file_pointer), (length -1));
   3) Then write your appended data to the file.

Q: I run my DOS program, allocate some memory, and check the
   amount of memory available with coreleft(). Then I
   free some memory and call coreleft() again. It reports
   the same number. Why?
A: Coreleft does NOT return the amount of memory available.
   It returns the total memory available above the highest
   block allocated. It does NOT return any amount of memory
   available in "holes" below the highest allocated block.
   The code for a function that returns the amount of total
   free memory is provided in the document TI1723. This
   document is available on Borland's Download BBS,
   [phone], as document # 1723, and on Borland
   Techfax, [phone], as TI1723.ZIP.

-------------------END OF FILE HELPME.WRI-------------------