14.4 Detecting programming errors by linguistic formalism

The front end of a compiler performs lexical, syntactic, and semantic analysis. Not all compilers find the same defects. The more information a compiler has, the more defects it can find. Some compilers operate in a “forgiving” mode but have a “strict” or “pedantic” mode, if you request it.

14.4.1 Regular grammars

Lexical analyzers can find the following errors:

Characters in the source that aren’t in the alphabet of the language
Words in the source that aren’t in the vocabulary of the language

14.4.2 Context-free grammars

Syntactic analyzers for fixed-form languages such as Fortran and COBOL can find the following errors:

Required fields that haven’t been used
Incorrect values in a field
Invalid continuations of statements across multiple lines
Keywords that have been misspelled
Required punctuation that is missing
Delimiters such as parentheses that are missing
Blank or tab characters that are missing
Blank or tab characters that shouldn’t occur where they’re found

Syntactic analyzers for free-form languages such as C++ and Java can find the following errors:

Comment delimiters that have been put in the wrong place or omitted
Literal delimiters that have been put in the wrong place or omitted
Keywords that have been misspelled
Required punctuation that is missing
Construct delimiters such as parentheses or braces that have been misplaced
Blank or tab characters that are missing
Blank or tab characters that shouldn’t occur where they’re found

14.4.3 Semantic analysis

If a semantic analyzer has a symbol table for each separate procedure, it can find semantic errors that occur because of the following mistakes:

Names that aren’t declared
Operands of the wrong type for the operator they’re used with
Values that have the wrong type for the name to which they’re assigned

If a semantic analyzer has a symbol table for the program as a whole, it can find semantic errors that occur because of the following mistakes:

Procedures that are invoked with the wrong number of arguments
Procedures that are invoked with the wrong type of arguments
Function return values that are the wrong type for the context in which they’re used

If a semantic analyzer has control-flow and data-flow information for each separate procedure, it can find semantic errors that occur because of the following mistakes:

Code blocks that are unreachable
Code blocks that have no effect
Local variables that are used before being initialized or assigned
Local variables that are initialized or assigned but not used

If a semantic analyzer has control-flow and data-flow information for the program as a whole, it can find semantic errors that occur because of the following mistakes:

Procedures that are never invoked
Procedures that have no effect
Global variables that are used before being initialized or assigned
Global variables that are initialized or assigned, but not used