More Books
C++ Gotchas: Avoiding Common Problems in Coding and Design
Main Page
Table of content
Copyright
Addison-Wesley Professional Computing Series
Preface
Acknowledgments
Chapter 1. Basics
Gotcha #1: Excessive Commenting
Gotcha #2: Magic Numbers
Gotcha #3: Global Variables
Gotcha #4: Failure to Distinguish Overloading from Default Initialization
Gotcha #5: Misunderstanding References
Gotcha #6: Misunderstanding Const
Gotcha #7: Ignorance of Base Language Subtleties
Gotcha #8: Failure to Distinguish Access and Visibility
Gotcha #9: Using Bad Language
Gotcha #10: Ignorance of Idiom
Gotcha #11: Unnecessary Cleverness
Gotcha #12: Adolescent Behavior
Chapter 2. Syntax
Gotcha #13: Array/Initializer Confusion
Gotcha #14: Evaluation Order Indecision
Gotcha #15: Precedence Problems
Gotcha #16: 'for' Statement Debacle
Gotcha #17: Maximal Munch Problems
Gotcha #18: Creative Declaration-Specifier Ordering
Gotcha #19: Function/Object Ambiguity
Gotcha #20: Migrating Type-Qualifiers
Gotcha #21: Self-Initialization
Gotcha #22: Static and Extern Types
Gotcha #23: Operator Function Lookup Anomaly
Gotcha #24: Operator '->' Subtleties
Chapter 3. The Preprocessor
Gotcha #25: '#define' Literals
Gotcha #26: '#define' Pseudofunctions
Gotcha #27: Overuse of '#if'
Gotcha #28: Side Effects in Assertions
Chapter 4. Conversions
Gotcha #29: Converting through 'void *'
Gotcha #30: Slicing
Gotcha #31: Misunderstanding Pointer-to-Const Conversion
Gotcha #32: Misunderstanding Pointer-to-Pointer-to-Const Conversion
Gotcha #33: Misunderstanding Pointer-to-Pointer-to-Base Conversion
Gotcha #34: Pointer-to-Multidimensional-Array Problems
Gotcha #35: Unchecked Downcasting
Gotcha #36: Misusing Conversion Operators
Gotcha #37: Unintended Constructor Conversion
Gotcha #38: Casting under Multiple Inheritance
Gotcha #39: Casting Incomplete Types
Gotcha #40: Old-Style Casts
Gotcha #41: Static Casts
Gotcha #42: Temporary Initialization of Formal Arguments
Gotcha #43: Temporary Lifetime
Gotcha #44: References and Temporaries
Gotcha #45: Ambiguity Failure of 'dynamic_cast'
Gotcha #46: Misunderstanding Contravariance
Chapter 5. Initialization
Gotcha #47: Assignment/Initialization Confusion
Gotcha #48: Improperly Scoped Variables
Gotcha #49: Failure to Appreciate C++'s Fixation on Copy Operations
Gotcha #50: Bitwise Copy of Class Objects
Gotcha #51: Confusing Initialization and Assignment in Constructors
Gotcha #52: Inconsistent Ordering of the Member Initialization List
Gotcha #53: Virtual Base Default Initialization
Gotcha #54: Copy Constructor Base Initialization
Gotcha #55: Runtime Static Initialization Order
Gotcha #56: Direct versus Copy Initialization
Gotcha #57: Direct Argument Initialization
Gotcha #58: Ignorance of the Return Value Optimizations
Gotcha #59: Initializing a Static Member in a Constructor
Chapter 6. Memory and Resource Management
Gotcha #60: Failure to Distinguish Scalar and Array Allocation
Gotcha #61: Checking for Allocation Failure
Gotcha #62: Replacing Global New and Delete
Gotcha #63: Confusing Scope and Activation of Member 'new' and 'delete'
Gotcha #64: Throwing String Literals
Gotcha #65: Improper Exception Mechanics
Gotcha #66: Abusing Local Addresses
Gotcha #67: Failure to Employ Resource Acquisition Is Initialization
Gotcha #68: Improper Use of 'auto_ptr'
Chapter 7. Polymorphism
Gotcha #69: Type Codes
Gotcha #70: Nonvirtual Base Class Destructor
Gotcha #71: Hiding Nonvirtual Functions
Gotcha #72: Making Template Methods Too Flexible
Gotcha #73: Overloading Virtual Functions
Gotcha #74: Virtual Functions with Default Argument Initializers
Gotcha #75: Calling Virtual Functions in Constructors and Destructors
Gotcha #76: Virtual Assignment
Gotcha #77: Failure to Distinguish among Overloading, Overriding, and Hiding
Gotcha #78: Failure to Grok Virtual Functions and Overriding
Gotcha #79: Dominance Issues
Chapter 8. Class Design
Gotcha #80: Get/Set Interfaces
Gotcha #81: Const and Reference Data Members
Gotcha #82: Not Understanding the Meaning of Const Member Functions
Gotcha #83: Failure to Distinguish Aggregation and Acquaintance
Gotcha #84: Improper Operator Overloading
Gotcha #85: Precedence and Overloading
Gotcha #86: Friend versus Member Operators
Gotcha #87: Problems with Increment and Decrement
Gotcha #88: Misunderstanding Templated Copy Operations
Chapter 9. Hierarchy Design
Gotcha #89: Arrays of Class Objects
Gotcha #90: Improper Container Substitutability
Gotcha #91: Failure to Understand Protected Access
Gotcha #92: Public Inheritance for Code Reuse
Gotcha #93: Concrete Public Base Classes
Gotcha #94: Failure to Employ Degenerate Hierarchies
Gotcha #95: Overuse of Inheritance
Gotcha #96: Type-Based Control Structures
Gotcha #97: Cosmic Hierarchies
Gotcha #98: Asking Personal Questions of an Object
Gotcha #99: Capability Queries
Bibliography

Gotcha #55: Runtime Static Initialization Order

All static data in a C++ program are initialized before access. Most of these static initializations are accomplished when the program image is loaded, before execution begins. If no explicit initializer is provided, the data are initialized to "all zeros":

static int question; // 0 
extern int answer = 42;
const char *terminalType; // null
bool isVT100; // false
const char **ptt = &terminalType;

These initializations all take place "simultaneously," with no issue of initializer ordering.

We can also employ runtime static initialization. In this case, there is no guarantee of initialization order between translation units. (A translation unit is basically a preprocessed file.) This is a frequent source of bugs, since initialization order may change without source code change:

// in file term.cpp 
const char *terminalType = getenv( "TERM" );

// in file vt100.cpp
extern const char *terminalType;
bool isVT100 = strcmp( terminalType, "vt100" )==0; // error?

There is an implicit ordering dependency between the initializations of terminalType and isVT100, but the C++ language does not, and cannot, guarantee a particular initialization order. This gotcha typically occurs when an existing, working program is ported to a different platform that happens to implement a different translation unit ordering for runtime static initializations. It may also pop up without source changes due to changes in a build procedure or if a facility that was formerly statically linked is changed to use dynamic linking.

Keep in mind that default initialization of static class objects also constitutes a runtime static initialization:

class TermInfo { 
 public:
   TermInfo()
       : type_( ::terminalType )
       {}
 private:
   std::string type_;
};
// . . .
TermInfo myTerm; // runtime static init!

The best way to avoid runtime static initialization difficulties is to minimize the use of external variables, including static class data members (see Gotcha #3).

Failing that, another possibility is to depend only on the initialization order within a given translation unit. This ordering is well defined, and the static variables within a translation unit are initialized in the order in which they are defined. For example, if the definitions for terminalType and isVT100 occurred in that order within the same file, there would be no portability issue. Even with this procedure, however, an initialization order problem may occur if an external function, including member functions, uses a static variable, since that function may be called, directly or indirectly, from runtime static initializations of other translation units:

extern const char *termType() 
   { return terminalType; }

Failing that, another approach might be to substitute lazy evaluation for initialization. Typically, this is accomplished with some variation of the Singleton pattern (see Gotcha #3).

As a last resort, we can code the initialization order explicitly, using standard techniques. One such standard technique is a Schwarz counter, so called because it was devised by Jerry Schwarz and is employed in his implementation of the iostream library:

gotcha55/term.h

extern const char *terminalType; 
//other things to initialize . . .
class InitMgr { // Schwarz counter
 public:
   InitMgr()
       { if( !count_++ ) init(); }
   ~InitMgr()
       { if( !--count_ ) cleanup(); }
   void init();
   void cleanup();
 private:
   static long count_; // one per process
};
namespace { InitMgr initMgr; } // one per file inclusion

gotcha55/term.cpp

extern const char *terminalType = 0; 
long InitMgr::count_ = 0;
void InitMgr::init() {
   if( !(terminalType = getenv( "TERM" )) )
       terminalType = "VT100";
   // other initializations . . .
}
void InitMgr::cleanup() {
   // any required cleanup . . .
}

A Schwarz counter counts how many times the header file in which it resides is #included. There is a single instance, per process, of the static member count_ of InitMgr. However, every time the header file term.h is included, a new object of type InitMgr is allocated, and each of these requires a runtime static initialization. The InitMgr constructor checks the count_ member to see if this is the "first" initialization of an InitMgr object of the process. If it is, the initializations are performed.

Conversely, when the process terminates normally, static objects that have destructors will be destroyed. With each InitMgr object destruction, the InitMgr destructor decrements the count_. When count_ reaches zero, any required cleanup is performed.

Although they are robust, particularly boneheaded coding can defeat even Schwarz counters. In general, it's best to minimize use of static variables and avoid runtime static initializations.