Programming Languages Suite

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

/ Programming Languages Suite / ProgLangD.iso / C++-7 / DISK1 / ERRATA2.TX$ / errata2

Wrap

Text File | 1992-03-19 | 51.3 KB | 1,383 lines

ERRATA2.TXT File Documentation Errata, Part Two, for Microsoft(R) C/C++, Version 7.0 (C) Copyright Microsoft Corporation, 1992 This document contains information on known errors and omissions in the documentation provided with version 7.0 of Microsoft C/C++. The information in this document and in the Microsoft Advisor Help system is more up-to-date than that in the manuals. Microsoft updates its languages documentation at the time of reprinting, so some of the information in this file may already be included in your manuals. ================================< Contents >================================ This file has nine parts: Part Title ---- ----- 6 Notes on "Run-Time Library Reference" 7 Notes on "Class Libraries Reference" 8 Notes on "Class Libraries User's Guide" 9 Notes on "Comprehensive Index and Errors Reference" 10 Notes on "CLANG.HLP" (Help) 11 Notes on "CL.HLP" (Help) 12 Notes on "LINK" (Help) 13 Notes on Header Files 14 Notes on "Microsoft Class Libraries" Quick-Reference Card See also ERRATA1.TXT. =============< Part 6: Notes on "Run-Time Library Reference" >=============== The CString Class and Type const char * --------------------------------------- Because the Microsoft Foundation Class's (MFC) CString class has a conversion operator for type const char *, you can pass CString objects to any run-time library function that accepts arguments of type const char *. In this version and previous versions of the run-time library documentation, the const keyword was left out of the syntax line that begins each run-time function's description. This omission was made because the const keyword served mainly as an instruction to the compiler and offered little useful information to C programmers. The run-time libraries that ship with Microsoft C/C++ contain over 175 functions that accept arguments of type const char *. Look in the associated include file if you are uncertain whether a function accepts such arguments. Page Note ---- ---- 13 Operating-System Considerations ------------------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 32 Text and Binary Modes --------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 36 Using Predefined Stream Pointers -------------------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 36 Controlling Stream Buffering ---------------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 37 Closing Streams --------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 40 Using Predefined Handles ------------------------ For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 63, _doserrno, errno, _sys_errlist, _sys_nerr 64 ----------------------------------------- For ANSI compatibility, both the _sys_nerr variable and the _sys_errlist variable require leading underscores. 129 The _cabs and _cabsl Functions ------------------------------ For these functions, replace the Return Value section with the following: "These functions return the absolute value if successful, or HUGE_VAL on overflow (_LHUGEVAL for _cabsl). The errno variable is set to ERANGE on overflow. Error handling can be changed with the _matherrl function." 163 The cos Functions ----------------- For the cos functions, replace the Return Value section with the following: "If <x> is large, a partial loss of significance in the result may occur in a call to cos, in which case the function generates a _PLOSS error. If <x> is so large that significance is completely lost, cos prints a _TLOSS message to stderr and returns 0. In both cases, errno is set to ERANGE. "If the result is too large in a cosh call, the functions return HUGE_VAL (_LHUGE_VAL for _coshl) and set errno to ERANGE. This behavior can be changed with _matherr." 177 The difftime Function --------------------- The final printf statement in the example program DIFFTIME.C uses a format specifier, "%6.2f", which incorrectly implies accuracy to a hundredth of a second. The difftime function always returns a whole number. Replace the printf statement with the following: printf( "\nProgram takes %6.0f seconds.\n", elapsed_time ); Replace the last line of the output with the following: Program takes 4 seconds. 236 The _dup and _dup2 Functions ---------------------------- Replace the third paragraph in the Remarks section with the following: "Both _dup and _dup2 accept file handles as arguments. If you want to pass a stream (FILE *) to either of these functions, use the _fileno function. This function associates a stream with a file handle. The following example shows how to associate stderr (defined as FILE * in STDIO.H) with a handle: cstderr = _dup( _fileno( stderr )); "Note that in a QuickWin application, you cannot use either _dup or _dup2 with _fileno on the stdin, stdout, or stderr streams. You can, however, use the combination of the _dup functions and the _fileno function on other streams." 253 The exp and _expl Functions --------------------------- For these functions, replace the Return Value section with the following: "These functions return the exponential value of <x> if successful. On overflow, expl returns _HUGE_VAL and _expl returns _LHUGE_VAL. On underflow, they return 0. The errno variable is set to ERANGE on overflow but is not set on underflow." 260 The fclose and _fcloseall Functions ----------------------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 277 The fgets Function ------------------ In the Description section, replace the description of the int n argument with "Maximum number of characters to read". In the Remarks section, remove the following sentence: "If <n> is equal to 1, <string> is empty ("")." 312 The FREOPEN.C Example --------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 318 The fseek Function ------------------ Starting with Microsoft C/C++, the default action of the fseek function is to not check whether the file pointer is positioned before the beginning of the specified file. This change was made for performance reasons. If you want to receive an error message when an attempt is made to seek before the beginning of a file, you must explicitly link with LSEEKCHK.OBJ. This new object enables the code that checks for and reports the occurrence of the error condition. Because the error check slows fseek, you may wish to use LSEEKCHK.OBJ during the development cycle but not in the final product. Replace the last sentence in the third paragraph of the Remarks section with the following: "An attempt to position the pointer before the beginning of the file causes an error only if you explicitly link with LSEEKCHK.OBJ." 326 The _fstat Function ------------------- The last paragraph in the Remarks section for the _fstat function should read as follows: "In the File Allocation Table (FAT) file system supported by MS-DOS, the creation and last access times of a file are not kept separately. Therefore st_atime, st_ctime, and st_mtime are always the same. These fields have different values under UNIX. "Note that if <handle> refers to a device, the size and time fields in the _stat structure are not meaningful. Also, because STAT.H uses the _dev_t type, which is defined in TYPES.H, you must include TYPES.H before STAT.H in your code." 335, The _fwopen Function 337 -------------------- Change all occurrences of "_WINVER" to "_QWINVER". 348 _getch, _getche --------------- These functions cannot be used in QuickWin programs. 366 The _getgtextvector Function ---------------------------- Change the first word in the Description sentence from "Changes" to "Gets". 367 The _getimage Functions ----------------------- Remove the following sentence fragment from the final paragraph of the Remarks section: "or by using the formula described on the _imagesize reference page." 424 The _hypot and _hypotl Functions -------------------------------- For these functions, replace the Return Value section with the following: "These functions return the length of the hypotenuse if successful or _HUGE_VAL (_LHUGE_VAL for _hypotl) on overflow. The errno variable is set to ERANGE on overflow." 447 The _ldexp, and _ldexpl Functions --------------------------------- For these functions, replace the first paragraph of the Return section with the following: "These functions return an exponential value if successful. On overflow (depending on the sign of <x>), _ldexp returns +/-_HUGE_VAL, and _ldexpl returns +/-_LHUGE_VAL; the errno variable is set to ERANGE." 462 The _locking Function --------------------- The LOCKING.C example should include a call to the lseek function (between the call to printf and the call to _locking in the if block). Add the following line to correctly reset the file pointer before unlocking the file: lseek( fh, 0L, SEEK_SET ); 463 The log Functions ----------------- For the _logl and _logl10 functions, replace the first paragraph of the Return Value section with the following: "The log functions return the logarithm of <x> if successful. If <x> is negative, the functions print a _DOMAIN error message to stderr, return the value -HUGE_VAL (-_LHUGE_VAL for the long double functions), and set errno to EDOM. If <x> is 0, they print a _SING error message to stderr, and set errno to ERANGE." 471 The _lseek Function ------------------- Starting with Microsoft C/C++, the default action of the _lseek function is to not check whether the file pointer is positioned before the beginning of the specified file. This change was made for performance reasons. If you want to receive an error message when an attempt is made to seek before the beginning of a file, you must explicitly link with LSEEKCHK.OBJ. This new object enables the code that checks for and reports the occurrence of the error condition. Because the error check slows _lseek, you may wish to use LSEEKCHK.OBJ during the development cycle but not in the final product. Replace the last sentence in the Remarks section with: "An attempt to position the pointer before the beginning of the file causes an error only if you explicitly link with LSEEKCHK.OBJ." 480 malloc Functions ---------------- The list of functions that can call malloc should include the realloc and ::operator new functions. 484 The _matherr and _matherrl Functions ------------------------------------ The description of the _UNDERFLOW value for the type field of the _exception structure should read as follows: _UNDERFLOW The result is too small to be represented. (This condition is not currently supported.) 521 The mktime Function ------------------- The fourth paragraph in the Remarks section should read as follows: "The mktime function handles dates in any time zone from midnight, December 31, 1899 to midnight, February 5, 2036. If timeptr references a date before midnight, December 31, 1899, mktime returns -1 cast to type time_t." 547 perror ------ For ANSI compatibility, both the _sys_nerr variable and the _sys_errlist variable require leading underscores. 654 The setbuf Function ------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 570 The _pg_initchart Function -------------------------- The first paragraph under the Remarks section should end with the following sentence: "Also, _pg_initchart must be called after each change of the video mode." 613 The realloc Functions --------------------- Add the following sentence to the last sentence of the second paragraph in the Remarks section: "Because the new block can be in a new memory location, the pointer returned by the realloc functions is not guaranteed to be the pointer passed through the <memblock> argument." 670 The setmod Function ------------------- For ANSI compatibility, change all occurrences of "stdaux" and "stdprn" to "_stdaux" and "_stdprn". 638 scanf ----- The format specifier %U listed in table R.5 is obsolete and is no longer supported. 712 The sinh Functions ------------------ For the sin functions, replace the second paragraph of the Return Value section with the following: "The sinh function returns the hyperbolic sine of <x>. If the result is too large, sinh sets errno to ERANGE and returns +/-HUGE_VAL (+/-_LHUGE_VAL for the long double functions). Error handling can be changed with the _matherr function." 734 _stat ----- For ANSI compatibility, type _dev_t requires a leading underscore. 755 strerror, _strerror ------------------- For ANSI compatibility, both the _sys_nerr variable and the _sys_errlist variable require a leading underscore. 799 strxfrm ------- At the end of the Remarks section, this line: return( strlen( _string2 ) ); should be: return strlen( _string1 ); 866 The _wclose Function -------------------- Change all occurrences of "_WINVER" to "_QWINVER". 872 The _wgetexit Function ---------------------- Change all occurrences of "_WINVER" to "_QWINVER". 879 The _wgetsize Function ---------------------- Change all occurrences of "_WINVER" to "_QWINVER". 882, The _wopen Function 884 ------------------- Change all occurrences of "_WINVER" to "_QWINVER". 890 The _wsetexit Function ---------------------- Change all occurrences of "_WINVER" to "_QWINVER". 897 The _wsetsize Function ---------------------- Change all occurrences of "_WINVER" to "_QWINVER". 906 Index ----- The _daylight variable is missing its leading underscore in the index to the "Run-Time Library Reference." 915 Index ----- Under O, insert: OLDNAMES.LIB, x (Introduction) 922 Index ----- The _timezone and _tzname variables are missing their leading underscores in the index to the "Run-Time Library Reference." =============< Part 7: Notes on "Class Libraries Reference" >=============== The LPSTR parameter type in the following functions was changed to LPCSTR: CWinApp::LoadCursor CWinApp::LoadStandardCursor CWinApp::LoadIcon CWinApp::LoadStandardIcon CDC::Escape(..., LPSTR lpInData, ...); The LPSTR parameter type in the following functions was changed to const void FAR*: CBitmap::CreateBitmap CBitmap::SetBitmapBits CDC::CreateDC CDC::CreateIC CMenu::LoadMenuIndirect CDialog::CreateIndirect The const char FAR* parameter type in the following functions was changed to LPSTR: CWnd::DlgDirList CWnd::DlgDirListComboBox CDC::GetTextFace The function CWnd::SetWindowPos returns a Boolean value, which is nonzero if successful. The function CGdiObject::DeleteObject now returns a Boolean value. The return type has changed for the following global functions and member function: HINSTANCE AfxGetInstanceHandle(); HINSTANCE AfxGetResourceHandle(); WNDPROC* CWnd::GetSuperWndProcAddr(); The LPSTR parameter type has been replaced in the syntax of the following member functions, as follows: DWORD CBitmap::GetBitmapBits( DWORD dwCount, void FAR* lpBits) const; int CDC::Escape( int nEscape, int nCount, LPCSTR lpInData, void FAR* lpOutData); int CGdiObject::GetObject( int nCount, void FAR* lpObject) const; The CWnd* parameter type in the syntax of the following member functions has changed to the CListBox* or CScrollBar* type: afx_msg int CWnd::OnCharToItem(UINT nChar, CListBox* pListBox, UINT nIndex); afx_msg void CWnd::OnHScroll(UINT nSBCode, UINT nPos, CScrollBar* pScrollBar); afx_msg int CWnd::OnVKeyToItem(UINT nKey, CListBox* pListBox, UINT nIndex); afx_msg void CWnd::OnVScroll(UINT nSBCode, UINT nPos, CScrollBar* pScrollBar); Both the parameter and parameter type have changed for the following member function: afx_msg void CWnd::OnNcCalcSize( NCCALCSIZE_PARAMS FAR* lpncsp); Changes should be made to the "Class Libraries Reference" manual on the following pages: Page Note ---- ---- 74 WM_PAINTICON Message is No Longer Supported ------------------------------------------- The WM_PAINTICON message is no longer supported by Windows 3.1. Therefore, the OnPaintIcon member function (described on page 773) is no longer part of the CWnd class. This also means that the message-map entry ON_WM_PAINTICON() on page 74 and the corresponding function prototype are not supported. Windows now generates the WM_PAINT message for events that would have triggered WM_PAINTICON messages, so CWnd::OnPaint can be used instead of OnPaintIcon. 90 Class Hierarchy Chart --------------------- The chart should show that class CMDIFrameWnd is derived from CFrameWnd instead of CWnd. This change should also be made to the Microsoft Class Libraries quick-reference card. 271 CDialog::OnInitDialog -------------------- Replace the paragraph under Return Value with the following: "Specifies whether the application has set the input focus to one of the controls in the dialog box. If OnInitDialog returns TRUE, Windows sets the input focus to the first control in the dialog box. The application can return FALSE only if it has explicitly set the input focus to one of the controls in the dialog box." 437 The description of the lpRectReserved parameter of CMenu::TrackPopupMenu should say that lpRectReserved points to a RECT structure or CRect object instead of a RECT structure or CPoint object. 514 class CPoint ------------ The description of the return value for member function Offset of class CPoint should read as follows: "A CPoint offset by a POINT, CPoint, SIZE, or CSize." 606 class CTime ----------- Add the following sentence to the third paragraph: "Note that CTime uses the strftime function, which is not supported for Windows DLLs. Therefore, CTime cannot be used in Windows DLLs." 773 WM_PAINTICON Message is No Longer Supported ------------------------------------------- The WM_PAINTICON message is no longer supported by Windows 3.1. Therefore, the OnPaintIcon member function is no longer part of the CWnd class. ============< Part 8: Notes on "Class Libraries User's Guide" >============= Page Note ---- ---- 9 Makefile Defaults ----------------- The third paragraph in the "Makefile Defaults" section should read as follows: "Both release mode and debug mode sample programs build into the same directory. DMTEST and CMDBOOK build into the MFC\SAMPLES\TUTORIAL\CHAR directory. PHBOOK builds into the MFC\SAMPLES\TUTORIAL\WIN directory. HELLO builds into the MFC\SAMPLES\HELLO directory." The fourth paragraph in the same section should be completely deleted. This paragraph begins as follows: "For debug mode builds, DMTEST and CMDBOOK build into the MFC\SAMPLES\TUTORIAL\CHARD directory...." 11 How to Build with PWB --------------------- At the top of the page, item 7 in the list should be replaced by the following: "If you get errors, examine the error messages in the Build Results window. You can get information on the error messages in Help or in the 'Comprehensive Index and Error Reference' manual. After correcting these errors, rebuild your program." 12 Debug Mode for PWB ------------------ Item 3 under "To switch from release builds to debug- ging builds with PWB" should read simply as follows: 3. Select "Use Debug Options" Delete the explanatory paragraph below item 3. 12 Debug Mode for NMAKE -------------------- Add the following lines just after "To switch from release builds to debugging builds with NMAKE": - Remove all existing release-mode object, compiled resource, and executable files by invoking NMAKE on the sample makefile while specifying the 'clean' target. For example: NMAKE /F PHBOOK clean (case is significant) You must always move or delete old compiled files before switching build modes (from release to debug or from debug to release). 59 Incorrect Output Listing ------------------------ The line following the second instance of "List contains:" in the program output should read: Meyers, Brian 236-1234 The corresponding output from the sample program (DMTEST.EXE) is correct. 116 RESOURCE.H Source Code Listing ------------------------------ The file listing for file RESOURCE.H is missing an end- comment delimiter. The first line of the file should end with a "*/" comment delimiter. The code is correct on disk. 165 How Dialog Objects Work ----------------------- The last sentence in the first paragraph of this section should read "For a distinction between modal and modeless dialog boxes, see 'Modal vs. Modeless Dialog Boxes' on page 162." 171 VIEW.CPP Source Code Listing ---------------------------- Change the one occurrence of the following: extern "C" { #include <commdlg.h> } to the following: #include <commdlg.h> 176 VIEW.CPP Source Code Listing ---------------------------- Change both occurrences of the following: char ErrorMsg[25]; to the following: char ErrorMsg[SIZESTRING]; 178 VIEW.CPP Source Code Listing ---------------------------- Change the one occurrence of the following: char szError[50]; to the following: char szError[SIZESTRING]; Also on page 178, change the following: UINT nExtPage = ( dc->GetDeviceCaps(VERTRES) - nCharHeight ) / nCharHeight; to the following: int nExtPage = ( dc->GetDeviceCaps(VERTRES) - nCharHeight ) / nCharHeight; 180 VIEW.CPP Source Code Listing ---------------------------- After the line that reads: pMenu -> EnableMenuItem( IDM_FINDALL, MF_ENABLED ); add the line: pMenu -> EnableMenuItem( IDM_FIND, MF_GRAYED ); 181 VIEW.CPP Source Code Listing ---------------------------- After the line that reads: pMenu -> EnableMenuItem( IDM_FINDALL, MF_GRAYED ); add the line: pMenu -> EnableMenuItem( IDM_FIND, MF_ENABLED ); 184 VIEW.CPP Source Code Listing ---------------------------- Change the occurrence of the following: void CMainWindow::OnVScroll( UINT wParam, UINT pos, CWnd* ) to the following: void CMainWindow::OnVScroll( UINT wParam, UINT pos, CScrollBar* ) 185 VIEW.CPP Source Code Listing ---------------------------- Change the occurrence of the following: void CMainWindow::OnHScroll( UINT wParam, UINT pos, CWnd* ) to the following: void CMainWindow::OnHScroll( UINT wParam, UINT pos, CScrollBar* ) 192 VIEW.CPP Source Code Listing ---------------------------- Change the occurrences of the following: of.lpstrFilter = (LPSTR)szFilter; of.lpstrCustomFilter = (LPSTR)NULL; of.lpstrTitle = (LPSTR)NULL; of.lpstrDefExt = (LPSTR)"pb"; to the following: of.lpstrFilter = szFilter; of.lpstrCustomFilter = NULL; of.lpstrTitle = NULL; of.lpstrDefExt = "pb"; 193 VIEW.CPP Source Code Listing ---------------------------- Change both occurrences of the following: char ErrorMsg[25]; to the following: char ErrorMsg[SIZESTRING]; 203 To add a constructor definition: -------------------------------- The last sentence in the first paragraph on this page should read "For more information about the VERIFY macro and its companion, the ASSERT macro, see Chapter 11 in this manual." 364 Microsoft C/C++ Input/Output Alternatives ----------------------------------------- Add the following sentence to the last paragraph on the page: "You cannot use iostreams in tiny-model programs. The reason is that tiny-model programs cannot contain static objects like the cin and cout stream objects." ======< Part 9: Notes on "Comprehensive Index and Errors Reference" >======= NOTE: The Comprehensive Index does not include entries for books ordered as fulfillment items for MS C/C++. Page Note ---- ---- 77, Index 252 ----- The _daylight and _tzname variables are missing their ANSI-mandated underscores. 255 Index ----- The _timezone variable is not listed. The _timezone variable is described on page 62 of the "Run-Time Library Reference." Changes to "Error Messages" --------------------------- If there is a difference between the printed description of an error or warning and its description in Help, then the information in Help supersedes information in the printed documentation. The printed descriptions for the following errors are superseded by the descriptions given in Help: DOS Extender Run-Time Errors R6901, R6903, R6907, R6914, R6918, R6921, R6950 CVPACK Error CK1003 The CVPACK error message CK1002 is incorrect: CK1002 out of virtual memory The printed documentation and Help for this error incorrectly mention a DPMI server named MSDPMI.EXE. To make extended memory available to CVPACK, run it in a DOS session under Windows 3.x enhanced mode, or under another DPMI server such as the version of Qualitas 386-Max that ships with Microsoft C/C++. Neither the printed nor the online documents describe the following new CVPACK error message: CK1022 Symbol table for <filename> is too large. The corrective action is to compile <filename> without CodeView information, reduce the number of symbols in the file, or split the file into two or more pieces. The following CodeView error messages are new: CV1060 command not supported for current configuration If you have specified the two-monitors option (/2), you cannot specify the flip/swap option (/of- or /of+) from the CodeView Command Window. CV3627 no nonnative execution model You must specify a nonnative execution model in order to debug a p-code program. If you do not, CodeView displays a message box that reminds you that you must. Add the following line to your TOOLS.INI file: model:nmd1pcd.dll Compiler error C1060 should include the following information: C1060 compiler is out of far heap space In addition to the solutions described in the printed documentation and Help, increasing the amount of virtual memory that Windows provides can solve this problem. To do this, run the Control Panel program in the Windows Main group, and then open the 386 Enhanced dialog box. Next, click the "Virtual Memory" button and then click the "Change" button. Enter a larger value in the "New Size" field. You may need to free some disk space before you can use more virtual memory. The printed description of the following errors and warnings should be replaced: C1004 unexpected end of file found The default disk drive did not contain sufficient space for compiler-generated temporary files. The space required is approximately two times the size of the source file. This message appears: - When #if directives evaluate to false without corresponding closing #endif directives. - If the precompiled headers option (/Yu) is enabled, but the file doesn't contain a hdrstop pragma. - When the /Yu<filename> option is used and the source file being compiled does not contain an #include "<filename>" directive that uses the same filename used with /Yu. C4010 single-line comment contains line-continuation character A single-line comment (introduced by //) contains a line-continuation character (\). The next line after the line-continuation character is ignored; it is considered to be the continuation of the previous line and is treated as a comment. The following compiler error is not described in the printed documentation or in Help: C2599 'function' : local functions are not supported The specified function was defined inside the body of another function. This error can be caused by omitting the closing curly brace (}) in a function definition. The following is an example of this error: void func1() { // Missing a closing curly brace here void func2() {} // Error detected on this line void func3() { void func4() {} // Error detected on this line } // Local functions not allowed The following errors have been renumbered: C2852 is now C1852 C2853 is now C1853 The following compiler errors are not documented in Help or in the printed documentation. C2675 unary 'operator' : 'class' does not define this operator or a conversion to a type acceptable to the builtin operator The specified operator was used on a user-defined type that did not define the operator or allow the type to be converted to a fundamental type. This error can be caused by specifying a parameter type in a unary operator definition. A unary operator is always passed an implicit this pointer to the object the operator is applied to, so the correct definition syntax for a unary operator uses empty parentheses. The following is an example of this error: class C { public: operator +(int); // binary +, not unary } c; class D { public: operator int(); // conversion to int } d; class E { public: operator +(); // correct unary syntax } e; int i; void main() { i = +c; // error i = +d; // OK, d is converted to int first i = +e; // OK, unary + is defined } C2676 binary 'operator' : 'class' does not define this operator or a conversion to a type acceptable to the builtin operator The specified operator was used on a user-defined type that did not define the operator or allow the type to be converted to a fundamental type. This error can be caused by defining the unary form of an operator but not the binary form. The correct syntax for a binary operator uses one parameter. The following is an example of this error: class C { public: operator +(); // unary +, but not binary } c; class D { public: operator int(); // conversion to int } d; class E { public: operator +(int); // correct binary syntax } e; int i; void main() { i = c + 1; // error i = d + 1; // OK, d is converted to int first i = e + 1; // OK, binary + is defined C2677 binary 'operator' : no global operator defined which takes type 'class' (or there is no acceptable conversion) The specified operator was used on a user-defined type that did not define the operator or allow the type to be converted to a fundamental type. This error can be caused by an incorrect definition of the operator, or by using the parameters in the wrong order. The following is an example of this error: class C { public: operator +(int); // OK for C + int } c; void main() { c = 1 + c; // error for int + C } The following example shows correct declaration of a global operator for a user-defined type: class C { public: friend C operator+( int, C& ); } c; void main() { c = 1 + c; // OK } Compiler Warning C4793 is not documented in Help or in the printed documentation: C4793 native code generated for p-code function 'function' with inline assembly Because the specified function contained inline assembly, it was not compiled as a p-code function. It is not possible to mix p-code and assembly in a function. Additions to "Error Messages" ----------------------------- The following errors and warnings are described in Help but are not described in the printed documentation. Fatal Errors: C1014 too many include files : depth = 'level' C1038 compiler limit : 'function' : control flow state too complex; simplify function C1506 unrecoverable block scoping error Compiler Errors: C2080 illegal __far __fastcall function or __saveregs function C2081 'identifier' : name in formal parameter list illegal C2117 'identifier' : array bounds overflow C2128 'function' : alloc_text/same_seg applicable only to functions with C linkage C2138 cannot define an enumeration with no members C2147 'identifier' : const array must be fully initialized C2201 'identifier' : cannot export static declarations C2225 'function' : function allocation must be in a named segment C2230 'operator' : indirection to different types C2240 unexpected 'token' following formal list. (only modifiers are allowed.) C2265 'identifier' : reference to a zero sized array is illegal C2266 'identifier' : reference to a nonconstant bounded array is illegal C2267 'function' : block scoped static functions are illegal C2275 'identifier' : illegal use of this type as an expression C2276 illegal address of bound member function expression C2277 'identifier' : cannot take the address of a constructor C2278 'function' : no function with C linkage found C2279 cannot use braces to initialize default arguments C2280 missing '{' before identifier 'identifier'? C2281 'identifier1' : is not a function, but contains <function returning>. 'identifier2' is expected. C2282 'identifier1' is followed by 'identifier2'. (missing ','?) C2283 'identifier' : pure specifier not allowed on unnamed class-key C2290 C++ 'keyword' syntax is reserved for future use C2379 formal parameter 'number' has different type when promoted C2379 formal parameter 'number' has different type when promoted C2387 ambiguous conversion from 'type1&' to 'type2&' C2423 'number' : illegal scale C2466 cannot allocate an array of constant size 0 C2513 'type' : decl-specifier is missing a declarator before '=' C2602 'class::identifier' is not a member of a base of 'class' C2603 illegal access declaration: 'class1' is not a direct base of 'class2' C2604 cannot declare 'modifier' access to 'modifier' member C2605 overloaded functions 'class::function' do not have same access C2606 'class::identifier': illegal private access declaration C2610 identifier 'identifier' can never be instantiated; user defined constructor is required C2634 '&class::member' : pointer to reference member is illegal C2723 'function' : 'specifier' storage class specifier illegal on function definition C2758 'const' : must be initialized in constructor base/member C2855 command-line option 'option' inconsistant with precompiled header C2856 #pragma hdrstop cannot be inside an #if block C2857 '#include filename' statement specified with the /Yc'filename'command-line option was not found in the source file Compiler Warnings: C4011 'identifier' : identifier was truncated to 'number' characters C4032 formal parameter 'number' has different type when promoted C4056 overflow in floating point constant arithmetic C4057 'operator' : indirection to slightly different base types C4058 unions are now aligned on alignment requirement, not size C4097 nonstandard extension : typedef-name 'identifier1' used as synonym for class-name 'identifier2' C4126 'option' : unknown memory-model command-line option C4127 conditional expression is constant C4146 unary minus operator applied to unsigned type, result still unsigned C4147 'identifier' : 'modifier' ignored on pointer and reference declarations C4152 nonstandard extension, function/data ptr conversion in expression C4153 function/data pointer conversion in expression C4154 deletion of an array expression; conversion to pointer supplied C4155 deletion of an array expression without using the array form of 'delete' C4156 deletion of an array expression without using the array form of 'delete'; array form substituted C4251 'identifier' : base type 'type' should be exported to export 'class-key' 'identifier' C4260 no override, 'class1::function1' has 'model' 'this' memory model whereas 'class2::function2' has 'model 'this' memory model C4261 no override, 'class1::function1' has 'model' calling convention whereas 'class2::function2' has 'model' calling convention C4307 'operator' : integral constant overflow; result truncated C4308 negative integral constant converted to unsigned type C4309 'identifier' : truncation of constant value C4355 'this' : used in base member initializer list C4385 'text' the 'identifier1' in base 'identifier2' C4386 'text' to the 'identifier1' in base 'identifier2' C4622 overwriting debug information formed during creation of the precompiled header in object file: 'file' C4741 /Oq option ignored for __fastcall function 'function' Command-Line Warnings: D4023 option 'option' forces use of optimizing compiler LINK Warnings: L4086 /r not first on command line; ignored DOS Extender Run-Time Errors: R6950 VMCPD.386 has been superseded by a later version of Windows. Change the [386Enh] section in SYSTEM.INI by replacing this line: device=[path\]vmcpd.386 with this line: device=*vmcpd ================< Part 10: Notes on "CLANG.HLP" (Help) >================== Example Program for vsprintf, vprintf, and vfprintf --------------------------------------------------- In the WPRINTF.C example program for the vsprintf, vprintf, and vfprintf functions, the following line is incorrect: size = (size > 512) ? 512 : size * 5; The greater-than operator should be a less-than operator: size = (size < 512) ? 512 : size * 5; __near, __far, __huge, and __based with Static Data Objects ----------------------------------------------------------- Online Help for __near, __far, __huge, and __based says that these keywords modify data objects. More precisely, they modify static data objects (variables with file scope, or static local variables), not automatic variables. ==================< Part 11: Notes on "CL.HLP" (Help) >=================== The second paragraph in the following MPC error message should read as follows: MP1083 segment overflow writing p-code tables in segment <segment> Reduce the size of the given segment and relink. If you are generating a Windows executable or are using the LINK /PACKC option, you must reserve enough space for the p-code tables with the LINK /PACKC:<number> option. The <number> specifies the maximum size of a code segment--63488 bytes (64K minus 512 bytes) is recommended. The following NMAKE error is not documented in Help or in the printed documentation: U1059 syntax error : '}' missing in dependent A search path for a dependent was incorrectly specified. Either a space existed in the path or the closing brace (}) was omitted. The syntax for a directory specification for a dependent is: {directories}dependent where <directories> specifies one or more paths, each separated by a semicolon (;). No spaces are allowed. If part or all of a search path is replaced by a macro, be sure that no spaces exist in the macro expansion. =======================< Part 12: Notes on "LINK" >======================= The /PCODE option for LINK is not documented in Help for LINK or in the printed documentation. The /PCODE option tells LINK to call the Make P-Code (MPC) utility after linking. If you link a p-code program using LINK instead of using CL, you must use /PCODE to generate a valid executable file. Section 3.4 of the "Programming Techniques" manual (page 53) contains information on how to manually link p-code programs. ===================< Part 13: Notes on Header Files >===================== Correction to Function Definition --------------------------------- The inline definition of istream::getline(signed char *, int, char) is missing "_fGline++;" This omission will cause the getline function to behave like the get function in all cases. Replace the definition with the following: inline istream& istream::getline(char _HFAR_ * _b,int \ _lim,char _delim) { _fGline++; return get(_b, _lim, _delim); } inline istream& istream::getline(unsigned char _HFAR_ * \ _b,int _lim,char _delim) { _fGline++; return get((char _HFAR_ \ *)_b, _lim, _delim); } inline istream& istream::getline(signed char _HFAR_ * _b,int _lim,char _delim) { _fGline++; return get((char _HFAR_ *)_b, \ _lim, _delim); } Missing Modifier in Prototypes ------------------------------ The set_new_handler prototypes in the NEW.H include file are missing the __cdecl modifier. The correct prototypes are: _PNH __cdecl _set_new_handler( _PNH ); _PNH __cdecl _set_nnew_handler( _PNH ); _PNH __cdecl _set_fnew_handler( _PNH ); _PNHH __cdecl _set_hnew_handler( _PNHH ); _PNHB __cdecl _set_bnew_handler( _PNHB ); ===< Part 14: Notes on "Microsoft Class Libraries" Quick-Reference Card >=== The chart should show that class CMDIFrameWnd is derived from CFrameWnd instead of CWnd. This change should also be made to the chart on pages 90 and 91 of the "Class Libraries Reference" book. ========================================================================== Microsoft, MS, MS-DOS, CodeView and QuickC are registered trademarks, and Windows is a trademark of Microsoft Corporation. UNIX is a registered trademark of American Telephone & Telegraph Company.