1@node ANSI 2@chapter @sc{gnu} C++ Conformance to @sc{ansi} C++ 3 4These changes in the @sc{gnu} C++ compiler were made to comply more 5closely with the @sc{ansi} base document, @cite{The Annotated C++ 6Reference Manual} (the @sc{arm}). Further reducing the divergences from 7@sc{ansi} C++ is a continued goal of the @sc{gnu} C++ Renovation 8Project. 9 10@b{Section 3.4}, @i{Start and Termination}. It is now invalid to take 11the address of the function @samp{main()}. 12 13@b{Section 4.8}, @i{Pointers to Members}. The compiler produces 14an error for trying to convert between a pointer to a member and the type 15@samp{void *}. 16 17@b{Section 5.2.5}, @i{Increment and Decrement}. It is an error to use 18the increment and decrement operators on an enumerated type. 19 20@b{Section 5.3.2}, @i{Sizeof}. Doing @code{sizeof} on a function is now 21an error. 22 23@b{Section 5.3.4}, @i{Delete}. The syntax of a @i{cast-expression} is 24now more strictly controlled. 25 26@b{Section 7.1.1}, @i{Storage Class Specifiers}. Using the 27@code{static} and @code{extern} specifiers can now only be applied to 28names of objects, functions, and anonymous unions. 29 30@b{Section 7.1.1}, @i{Storage Class Specifiers}. The compiler no longer complains 31about taking the address of a variable which has been declared to have @code{register} 32storage. 33 34@b{Section 7.1.2}, @i{Function Specifiers}. The compiler produces an 35error when the @code{inline} or @code{virtual} specifiers are 36used on anything other than a function. 37 38@b{Section 8.3}, @i{Function Definitions}. It is now an error to shadow 39a parameter name with a local variable; in the past, the compiler only 40gave a warning in such a situation. 41 42@b{Section 8.4.1}, @i{Aggregates}. The rules concerning declaration of 43an aggregate are now all checked in the @sc{gnu} C++ compiler; they 44include having no private or protected members and no base classes. 45 46@b{Section 8.4.3}, @i{References}. Declaring an array of references is 47now forbidden. Initializing a reference with an initializer list is 48also considered an error. 49 50@b{Section 9.5}, @i{Unions}. Global anonymous unions must be declared 51@code{static}. 52 53@b{Section 11.4}, @i{Friends}. Declaring a member to be a friend of a 54type that has not yet been defined is an error. 55 56@b{Section 12.1}, @i{Constructors}. The compiler generates a 57default copy constructor for a class if no constructor has been declared. 58 59@ignore 60@b{Section 12.4}, @i{Destructors}. In accordance with the @sc{ansi} C++ 61draft standard working paper, a pure virtual destructor must now be 62defined. 63@end ignore 64 65@b{Section 12.6.2}, @i{Special Member Functions}. When using a 66@i{mem-initializer} list, the compiler will now initialize class members 67in declaration order, not in the order in which you specify them. 68Also, the compiler enforces the rule that non-static @code{const} 69and reference members must be initialized with a @i{mem-initializer} 70list when their class does not have a constructor. 71 72@b{Section 12.8}, @i{Copying Class Objects}. The compiler generates 73default copy constructors correctly, and supplies default assignment 74operators compatible with user-defined ones. 75 76@b{Section 13.4}, @i{Overloaded Operators}. An overloaded operator may 77no longer have default arguments. 78 79@b{Section 13.4.4}, @i{Function Call}. An overloaded @samp{operator ()} 80must be a non-static member function. 81 82@b{Section 13.4.5}, @i{Subscripting}. An overloaded @samp{operator []} 83must be a non-static member function. 84 85@b{Section 13.4.6}, @i{Class Member Access}. An overloaded @samp{operator ->} 86must be a non-static member function. 87 88@b{Section 13.4.7}, @i{Increment and Decrement}. The compiler will now 89make sure a postfix @samp{@w{operator ++}} or @samp{@w{operator --}} has an 90@code{int} as its second argument. 91 92 93@node Encoding 94@chapter Name Encoding in @sc{gnu} C++ 95 96@c FIXME!! rewrite name encoding section 97@c ...to give complete rules rather than diffs from ARM. 98@c To avoid plagiarism, invent some different way of structuring the 99@c description of the rules than what ARM uses. 100 101@cindex mangling 102@cindex name encoding 103@cindex encoding information in names 104In order to support its strong typing rules and the ability to provide 105function overloading, the C++ programming language @dfn{encodes} 106information about functions and objects, so that conflicts across object 107files can be detected during linking. @footnote{This encoding is also 108sometimes called, whimsically enough, @dfn{mangling}; the corresponding 109decoding is sometimes called @dfn{demangling}.} These rules tend to be 110unique to each individual implementation of C++. 111 112The scheme detailed in the commentary for 7.2.1 of @cite{The Annotated 113Reference Manual} offers a description of a possible implementation 114which happens to closely resemble the @code{cfront} compiler. The 115design used in @sc{gnu} C++ differs from this model in a number of ways: 116 117@itemize @bullet 118@item 119In addition to the basic types @code{void}, @code{char}, @code{short}, 120@code{int}, @code{long}, @code{float}, @code{double}, and @code{long 121double}, @sc{gnu} C++ supports two additional types: @code{wchar_t}, the wide 122character type, and @code{long long} (if the host supports it). The 123encodings for these are @samp{w} and @samp{x} respectively. 124 125@item 126According to the @sc{arm}, qualified names (e.g., @samp{foo::bar::baz}) are 127encoded with a leading @samp{Q}. Followed by the number of 128qualifications (in this case, three) and the respective names, this 129might be encoded as @samp{Q33foo3bar3baz}. @sc{gnu} C++ adds a leading 130underscore to the list, producing @samp{_Q33foo3bar3baz}. 131 132@item 133The operator @samp{*=} is encoded as @samp{__aml}, not @samp{__amu}, to 134match the normal @samp{*} operator, which is encoded as @samp{__ml}. 135 136@c XXX left out ->(), __wr 137@item 138In addition to the normal operators, @sc{gnu} C++ also offers the minimum and 139maximum operators @samp{>?} and @samp{<?}, encoded as @samp{__mx} and 140@samp{__mn}, and the conditional operator @samp{?:}, encoded as @samp{__cn}. 141 142@cindex destructors, encoding of 143@cindex constructors, encoding of 144@item 145Constructors are encoded as simply @samp{__@var{name}}, where @var{name} 146is the encoded name (e.g., @code{3foo} for the @code{foo} class 147constructor). Destructors are encoded as two leading underscores 148separated by either a period or a dollar sign, depending on the 149capabilities of the local host, followed by the encoded name. For 150example, the destructor @samp{foo::~foo} is encoded as @samp{_$_3foo}. 151 152@item 153Virtual tables are encoded with a prefix of @samp{_vt}, rather than 154@samp{__vtbl}. The names of their classes are separated by dollar signs 155(or periods), and not encoded as normal: the virtual table for 156@code{foo} is @samp{__vt$foo}, and the table for @code{foo::bar} is 157named @samp{__vt$foo$bar}. 158 159@item 160Static members are encoded as a leading underscore, followed by the 161encoded name of the class in which they appear, a separating dollar sign 162or period, and finally the unencoded name of the variable. For example, 163if the class @code{foo} contains a static member @samp{bar}, its 164encoding would be @samp{_3foo$bar}. 165 166@item 167@sc{gnu} C++ is not as aggressive as other compilers when it comes to always 168generating @samp{Fv} for functions with no arguments. In particular, 169the compiler does not add the sequence to conversion operators. The 170function @samp{foo::bar()} is encoded as @samp{bar__3foo}, not 171@samp{bar__3fooFv}. 172 173@item 174The argument list for methods is not prefixed by a leading @samp{F}; it 175is considered implied. 176 177@item 178@sc{gnu} C++ approaches the task of saving space in encodings 179differently from that noted in the @sc{arm}. It does use the 180@samp{T@var{n}} and @samp{N@var{x}@var{y}} codes to signify copying the 181@var{n}th argument's type, and making the next @var{x} arguments be the 182type of the @var{y}th argument, respectively. However, the values for 183@var{n} and @var{y} begin at zero with @sc{gnu} C++, whereas the 184@sc{arm} describes them as starting at one. For the function @samp{foo 185(bartype, bartype)}, @sc{gnu} C++ uses @samp{foo__7bartypeT0}, while 186compilers following the @sc{arm} example generate @samp{foo__7bartypeT1}. 187 188@c Note it loses on `foo (int, int, int, int, int)'. 189@item 190@sc{gnu} C++ does not bother using the space-saving methods for types whose 191encoding is a single character (like an integer, encoded as @samp{i}). 192This is useful in the most common cases (two @code{int}s would result in 193using three letters, instead of just @samp{ii}). 194@end itemize 195 196@c @node Cfront 197@c @chapter @code{cfront} Compared to @sc{gnu} C++ 198@c 199@c 200@c FIXME!! Fill in. Consider points in the following: 201@c 202@c @display 203@c Date: Thu, 2 Jan 92 21:35:20 EST 204@c From: raeburn@@cygnus.com 205@c Message-Id: <9201030235.AA10999@@cambridge.cygnus.com> 206@c To: mrs@@charlie.secs.csun.edu 207@c Cc: g++@@cygnus.com 208@c Subject: Re: ARM and GNU C++ incompatabilities 209@c 210@c Along with that, we should probably describe how g++ differs from 211@c cfront, in ways that the users will notice. (E.g., cfront supposedly 212@c allows "free (new char[10])"; does g++? How do the template 213@c implementations differ? "New" placement syntax?) 214@c @end display 215@c 216@c XXX For next revision. 217@c 218@c GNU C++: 219@c * supports expanding inline functions in many situations, 220@c including those which have static objects, use `for' statements, 221@c and other situations. Part of this versatility is due to is 222@c ability to not always generate temporaries for assignments. 223@c * deliberately allows divide by 0 and mod 0, since [according 224@c to Wilson] there are actually situations where you'd like to allow 225@c such things. Note on most systems it will cause some sort of trap 226@c or bus error. Cfront considers it an error. 227@c * does [appear to] support nested classes within templates. 228@c * conversion functions among baseclasses are all usable by 229@c a class that's derived from all of those bases. 230@c * sizeof works even when the class is defined within its ()'s 231@c * conditional expressions work with member fns and pointers to 232@c members. 233@c * can handle non-trivial declarations of variables within switch 234@c statements. 235@c 236@c Cfront: 237