6

Fmt

2004.08.01

2

In numbered paragraph 1, "correct" shouldn't be in code font.

8

Cor

2004.08.22
Carl Daniel

2

In paragraph 5, change:
The new buffer initially holds zeroes

To:
The new buffer initially holds undefined values (often zeroes)

11

Typ

2004.08.22
Carl Daniel

2

In paragraph 1, the way it was written unintentionally implied that the C++ standard is based on C99 (the C++ standard is based on C95 = C90 + some later corrigenda + the 1995 amendment).

Change:
Throwing off the shackles and limitations of the C standard [C99], on which the C++ standard [C++03] is based,

To:
Throwing off the shackles and limitations of the C standard (currently [C99]; the C++ standard [C++03] is based on an earlier version of C),

12

Typ

2004.11.02
Chris Breen

2

In each of the two code examples, change:
PrettyFormat( value, buf );

To:
PrettyFormat( value, smallBuf );

27-28

Cor

2004.11.30
Igor Krassikov and Alexander Krotov

2

Remove spurious parentheses that actually are not valid C++ syntax for taking a pointer to member.

In the Question 2 code on page 27, and in two places on page 28, change:
&(std::vector<int>::clear)

To:
&std::vector<int>::clear

39

Typ

2004.09.07
Jason Shirk

2

In Example 6-2(b), there is a missing return type. Change:
template<> swap<MyClass>( MyClass& a, MyClass& b ) { // throw()

To:
template<> void swap<MyClass>( MyClass& a, MyClass& b ) { // throw()

39

Typ

2004.09.17
Bronek Kozicki

2

In Example 6-2(c), there is the same missing return type. Change:
swap( MyClass& a, MyClass& b ) /* throw() */ {

To:
void swap( MyClass& a, MyClass& b ) /* throw() */ {

41

Typ

2004.09.28
Igor Krassikov

2

In the first line of each of the first two paragraphs, change:
copy assignment

To:
copy construction

58

Fmt

2004.08.01

2

In footnote 12, "boost" should be in variable-width code font.

72

Fmt

2004.08.01

2

In footnote 18, the final "export" should be in code font.

102

Fmt

2004.09.14
Ramsay Jones

2

To make Example 14-2 and Figure 14-1 a little more consistent to avoid needless confusion, change the order of the base classes (this has no effect), change:
class D2 : public B3, virtual public V2 { };

To:
class D2 : virtual public V2, public B3 { };

131-2

Typ

2004.09.07
Jason Shirk

2

In Examples 18-1 and 18-2, there is missing whitespace between the "int" return type and the function name.

In both examples, change:
intProcess

To:
int Process

And twice in Example 18-2, change:
intDoProcess...

To:
int DoProcess...

144

Typ

2004.09.14
Ramsay Jones

2

Some cut-and-paste typos.

In paragraph 1, change:
An implicitly declared default constructor

To:
An implicitly declared copy constructor

In paragraph 4, change:
An implicitly declared default constructor

To:
An implicitly declared copy assignment operator

In paragraph 5, change:
can throw anything that a base or member copy constructor could throw

To:
can throw anything that a base or member copy assignment operator could throw

145

Typ

2004.09.17
Bronek Kozicki

2

In the second code block, there is a missing parameter name. Change:
inline X& X::operator=( X& ) throw() { i_ = other.i_; return *this; }

To:
inline X& X::operator=( X& other ) throw() { i_ = other.i_; return *this; }

164
166
168
169

Cor

2004.11.02
Josh Lehrer

2

Change FastMemory() temporary to a named variable.

On pages 164 and 168, in the Question 3 code, change:
Derived* p4 = new (FastMemory()) Derived; // 4

To:
FastMemory f;
Derived* p4 = new (f) Derived; // 4

On page 166, Example 22-1(b), change:
new (FastMemory()) T; // calls some user-supplied

To:
FastMemory f
new (f) T; // calls some user-supplied

On page 169, in the code example, change:
Derived* p4 = new (FastMemory()) Derived; // calls Base::operator new()

To:
FastMemory f;
Derived* p4 = new (f) Derived; // calls Base::operator new()

176

Enh

2004.08.20
Stephan Lavavej and Ryan Myers

2

In numbered paragraph 1, I've reworded some text for greater clarity. Change:
1. Checking new failure is useless on systems that don’t commit memory until the memory is used. On some operating systems, including but not limited to Linuxes,²⁹ memory allocation always succeeds. Full stop.

To:
1. Checking new failure can be useless on systems that don’t commit memory until the memory is used. On some operating systems,²⁹ the operating system’s underlying memory allocation function always succeeds. Full stop.

And in footnote 29, change:
Lazy-commit is also a configurable feature on some other operating systems.

To:
Lazy-commit is a default or configurable feature on some versions of Linux, AIX, and other operating systems.

Some more detail that doesn't fit in the book:
Something you need to know about and turn off: Specifically, Linux kernels before 2.6 have this as their default and unconfigurable behavior, and apparently the kernel team suggested that gcc should rewrite their malloc to do exactly what I described in this Item as an unacceptable workaround -- namely that gcc should install a SIGSEGV signal handler and touch the pages themselves before returning a pointer. Of course, gcc would take the blame for the performance hit. Still, many people provided patches for glibc to do so, or personally wrapped malloc/new for their own apps, because even such a performance hit was preferable to unpredictable behavior and the possibility of crashing anytime you touch memory you think you own. If you are using a kernel older than 2.6, I strongly recommend you do this too. Then in 2.6 the kernel team relented and provided a flag to turn this off. As of this erratum entry writing, however, I do not know of any Linux distribution that does not have this unpredictable behavior on by default. If you are using a Linux kernel 2.6 or later (and especially if you're using Linux as a server), I urge you to run (don't walk) to find the flag for this behavior and turn it off. It's not a good idea to make unpredictable behavior the required behavior (pre-2.6), or even the default behavior (2.6 and later).

191

Typ

2004.11.02
Josh Lehrer

2

In paragraph 4, for consistency, change:
3.14159 * 2.71828 * 3.14159 * 2.71828

To:
(3.14159 * 2.71828) * (3.14159 * 2.71828)

202

Typ

2004.11.26
Igor Krassikov

2

In paragraph 3, change:
although it wastes two possible values

To:
although it wastes a possible value

204-5

Typ

2004.09.14
Ramsay Jones

2

In the Question 1 code, "p" should be "dest" -- and while I'm at it I'll also change "dest" to "dst" just for consistency because I call that parameter "dst" later on (this has no effect on the code's correctness because the formal parameter name doesn't matter, but might as well be consistent).

On page 204, in Question 1's code, change:
// Get num bits starting with the start-th bit,
// and store the result starting with the first
// bit of p.
//
void Get( size_t start, size_t num, unsigned char* dest ) const;

To:
// Get num bits starting with the start-th bit,
// and store the result starting with the first
// bit of dst.
//
void Get( size_t start, size_t num, unsigned char* dst ) const;

On page 205, where Question 1's code is quoted, change:
void Get( size_t start, size_t num, unsigned char* dest ) const;

To:
void Get( size_t start, size_t num, unsigned char* dst ) const;

233

Typ

2004.09.14
Ramsay Jones

2

In paragraph 6, change:
that member's name

To:
that variable’s name

250

Typ

2004.11.27
Igor Krassikov

2

In numbered paragraph 4, a 9 turned into a 6 in the cross-reference. Change:
Item 36 in More Exceptional C++ [Sutter02]

To:
Item 39 in More Exceptional C++ [Sutter02]

253

Typ

2004.11.30
Igor Krassikov

2

In numbered paragraph 14, the cross-reference got mistyped. Change:
see [Sutter00, Item 39]

To:
see [Sutter00, Item 6]

255-6

Typ

2004.11.30
Igor Krassikov

2

When reformatting the code for the book I didn't update all the line counts.

On page 255, change:
// Solution 2 uses a pair instead of reinventing a pair-like helper class. Now we’re
// down to 13 lines, from the original 23. Of the 14 lines, 9 are purpose-specific,
// and 5 are directly reusable in other contexts.

To:
// Solution 2 uses a pair instead of reinventing a pair-like helper class. Now we’re
// down to 12 lines, from the original 23. Of the 12 lines, 8 are purpose-specific,
// and 4 are directly reusable in other contexts.

On page 256, change:
// Here we still have 15 lines, but more are reusable. This version uses more space

To:
// Here we still have 13 lines, but more are reusable. This version uses more space

259

Typ

2004.11.30
Igor Krassikov

2

In numbered paragraph 1, change:
(For a complete example, see Item 11 of Exceptional C++ [Sutter00].)

To:
(For a complete example, see Item 11 and its context in Exceptional C++ [Sutter00].)

263

Cor

2004.09.14
Ramsay Jones

2

In paragraph 3, something appeared as a template parameter that should have been a constructor parameter. Change:
Now it’s possible to have, for example, a list< callback< Widget, &Widget::SomeFunc > >

To:
Now it’s possible to write, for example:
  list< callback< Widget > > lcw( w, &Widget::SomeFunc );

287

Typ

2004.09.14
Ramsay Jones

2

In the last paragraph, change:
1998

To:
2003

297

Cor

2004.09.27
Jim Leahy

2

In the first two code examples, change:
!=

To:
==