Chapter Contents

Previous

Next

Implementation-defined Behavior

In many instances, the ISO/ANSI Standard gives an implementor of the C language the freedom to choose an appropriate approach for a particular aspect of the language. The only requirement is that the choice made is explained to the user. For SAS/C software, implementation-defined behavior refers to aspects of the compiler where coding decisions have been made as a result of the SAS/C implementation of the ISO/ANSI Standard C language. The following sections follow the ISO/ANSI Standard in covering implementation-defined behavior for the compiler. The ISO/ANSI Standard conventions are given in Annex G.3 of the ISO/IEC 9899:1990 Standard for Programming Languages -- C. The relevant sections of this appendix are noted next to the topic.

Translation (G.3.1)

• Compiler diagnostics are listed in SAS/C Software Diagnostic Messages.

Environment (G.3.2)

• The compiler conforms to the ISO/ANSI Standard for a hosted environment as documented in Section 2.1.2.2 of the American National Standard for Information Systems - Programming Language C. It also conforms to the environment variables extension documented in Run-Time Argument Processing.

• The library recognizes the terminal as the only interactive device. Under OS/390 batch, the terminal is logically represented by the DDname SYSTERM.

Identifiers (G.3.3)

• The number of significant initial characters in an identifier without external linkage is 31. The significant initial characters are 65,535 if the extname option is specified.

• The compiler recognizes eight significant characters in an identifier with external linkage if the extname option is not specified.

• Case distinctions are not significant in an identifier with external linkage unless the extname option is specified.

Characters (G.3.4)

• The source and execution character sets include all EBCDIC characters.

• Two shift states are used for encoding multibyte characters; a single byte shift state and a double byte shift state. 0x0e causes entry into the double byte shift state, and 0x0f causes return to the single byte shift state.

• There are eight bits in a character in the execution character set.

• The mapping of members of the source character set (in character constants and string literals) to members of the execution character set is standard EBCDIC. Refer to Escape Sequence Values.

• All characters in the basic (or extended) source character set are also in the basic (or extended) execution character set.

• An integer character constant that contains more than one character is interpreted as a 4-byte integer. If it contains more than four characters, only the rightmost four characters are used. A wide character constant that contains more than one character is interpreted by ignoring all characters but the rightmost.

• By default, a non-DBCS locale is used to convert multibyte characters to wide characters (and in all other processing of multibyte characters within the compiler). This can be changed by supplying an environment variable to the compiler that affects its locale. (See Chapter 10, "Localization Functions," and Chapter 11, "Multibyte Character Functions," in SAS/C Library Reference, Volume 2 for more information.) If the specified locale supports DBCS, EBCDIC DBCS rules are used to convert multibyte characters to wide character codes.

• A plain char is identical to an unsigned character.

Integers (G.3.5)

• Refer to Data Type Characteristics for the representations and sets of values of the various types of integers. The IBM architecture and SAS/C software use two's complement representation.

• When an integer is converted to a shorter signed integer, the high-order bytes are discarded. The bit pattern of the remaining bytes is unchanged. The bit pattern of an unsigned integer is not changed when it is converted to a signed integer of equal length.

• The following list covers the results of bitwise operations on signed integers:

  ~ (bitwise NOT)	The bits are inverted, that is, 1 bits are set to 0, and 0 bits are set to 1.
  >> (shift right)	The bits are right shifted. The sign bit (bit 0) is used to fill the vacated bit positions on the left.
  << (shift left)	The bits are left shifted. The vacated bit positions on the right are filled with 0s.
  & (bitwise AND)	If the corresponding bits in both operands are 1, then the corresponding bit in the result is set to 1; otherwise it is set to 0.
  | (bitwise OR)	If either of the corresponding bits in the operands are 1, then the corresponding bit in the result is set to 1; otherwise it is set to 0.
  ˆ (bitwise XOR)	If the corresponding bits in the operands are not alike, then the corresponding bit in the result is set to 1; otherwise it is set to 0.

• In integer division, the remainder has the same sign as the dividend, except that a 0 remainder is always positive. (These are the results produced by the 370 DIVIDE instruction.)

• In a right shift of a negative-valued signed integral type, the sign bit is used to fill the vacated bit positions on the left; that is, the result retains the sign of the first operand.

Floating point (G.3.6)

• Data Type Characteristics provides the representations and sets of values of the various types of floating-point numbers.

• The IBM 370 representation of a double can represent all integral values exactly, so no rounding is necessary when an integer is converted to a double . When a long integer value is converted to float , the value is rounded.

• Rounding (away from 0 when either direction is equally correct) occurs when a floating-point number is converted to a narrower floating-point number. (These are the results of the IBM 370 LOAD ROUNDED instruction.)

Arrays and pointers (G.3.7)

• The type of integer required to hold the maximum size of an array (for example, the type of the result of the sizeof operator, size_t ) is unsigned int .

• When a pointer is converted to integer, the integer contains the logical address of a data object. (Function pointers may use an indirect representation. See Local Function Pointers for more information.) If the logical address is not representable in the destination integral type, the conversion is analogous to the conversion from unsigned long to the destination type. Nonaddress bits in the pointer will be preserved.

• When an integer is converted to a pointer, the value used is that of the integer converted to unsigned long . The pointer contains the logical address corresponding to the integral value. Nonaddress bits are preserved.

• ptrdiff_t is signed long .

Registers (G.3.8)

• There can be up to six integer or pointer register variables and up to two floating-point register variables. The register keyword is ignored when the optimize option is used because registers are allocated to variables by the compiler when the optimize option is specified. The number of registers allocated can be controlled by the greg and freg options. See Optimizations for more information.

Structures, unions, enumerations, and bitfields (G.3.9)

• If a member of a union object is accessed using a member of a different type, the result is undefined.

• Each element is aligned according to Data Type Characteristics unless the bytealign option is used or the __noalignmem keyword is used. Padding is introduced as necessary to maintain correct alignment. For bitfields, see Noninteger bitfields. By default, bitfields are aligned on word boundaries.

• Plain int bitfields are treated as unsigned int bitfields.

• The order of allocation of bitfields within an int is left to right.

• A bitfield can straddle a storage unit boundary if the bitfield option specifies a storage allocation unit size of less than 4 bytes.

• The values of an enumeration type are represented as ints .

Qualifiers (G.3.10)

• Any reference to an object that has volatile -qualified type is considered an access to that object.

Declarators (G.3.11)

• There is no limit on the number of declarators that can modify an arithmetic, structure, or union type.

Statements (G.3.12)

• There is no limit to the number of case values in a switch statement.

Preprocessing directives (G.3.13)

• A value of a single-character character constant in a constant expression that controls conditional inclusion matches the value of the same character constant in the execution character set. However, such a character cannot have a negative value.

• The method for locating includable source files is outlined in Code Generation Conventions.

• The support of quoted names for includable source files is outlined in Code Generation Conventions.

• The mapping of source file character sequences is found in Code Generation Conventions.

• The behavior of each recognized #pragma directive is covered earlier in this chapter.

• The compiler does not create the __DATE__ and __TIME__ macros when the date and time are not available.

Library functions (G.3.14)

• The null pointer constant to which the macro NULL expands is 0.

• For information about the behavior of the assert function see Chapter 6, "Function Descriptions" in SAS/C Library Reference, Volume 1. The diagnostic printed by the assert diagnostic control function is

  Assertion failed: expr
  Interrupted while: Executing line <number > of <source file >

  where number is the current value of __LINE__ and source file is the current value of __FILE__ .

• The sets of characters tested for by the isalnum , isalpha , iscntrl , islower , isprint , and isupper functions are described in Chapter 6, "Function Descriptions," in SAS/C Library Reference, Volume 1 .

• The values returned by the mathematical functions on domain errors are covered in Chapter 6, "Function Descriptions," in SAS/C Library Reference, Volume 1.

• On underflow range errors the mathematical functions set the integer expression errno to the macro ERANGE .

• A value of 0 is returned when the fmod function has a second argument of 0.

• The following items are documented in Chapter 5, "Signal-Handling Functions" in SAS/C Library Reference, Volume 1:
  • the set of signals for the signal function

  • the semantics for each signal recognized

  • default handling and handling at program start-up for each signal recognized

  • the treatment of a signal on entry to a handler

  • whether default handling is reset if the SIGILL signal is received by a handler specified to the signal function.

• The last line of a text stream requires a terminating new-line character, unless the stream is an USS HFS file. One is generated if not explicitly written.

• Depending on file format, space characters that are written out to a text stream immediately before a new-line character may or may not appear when read in.

  Note:    See Chapter 3, "I/O Functions" in SAS/C Library Reference, Volume 1 for additional information concerning the pad amparm.  
  

• As many null characters can be appended to data written to a binary stream as it takes to fill out a complete logical record.

• When a file is opened with open mode "a" or "ab", it is initially positioned to the end of file. When opened with "a+" or "a+b", it is initially positioned to the start of the file.

• Whether a write on a text stream causes the associated file to be truncated beyond that point depends on the file type and options. See Chapter 3, "I/O Functions" in SAS/C Library Reference, Volume 1.

• For the characteristics of file buffering, refer to the discussion of setbuf and setvbuf in Chapter 6, "Function Descriptions" in SAS/C Library Reference, Volume 1.

• Refer to the section "IBM 370 I/O Concepts" in Chapter 3, "I/O Functions" in SAS/C Library Reference, Volume 1 for information on the existence of a zero-length file.

• Refer to "Opening Files" in Chapter 3, "I/O Functions" in SAS/C Library Reference, Volume 1 for the rules for composing valid filenames.

• The terminal can be opened multiple times, as can files open for input only. Opening a nonterminal file for output multiple times or for both input and output gives unpredictable results.

• The effect of the remove function on an open file is unpredictable.

• rename will fail if a file with the new name exists prior to a call to the rename function.

• The output for %p conversion in the fprintf function is the pointer value in hexadecimal.

• The input for the %p conversion in the fscanf function is a hexadecimal string, optionally preceded by 0x or 0p.

• In the fscanf function, the hyphen (-) character has no special meaning. It is interpreted the same way as any other character in the scan list.

• Refer to the documentation for fgetpos and ftell in Chapter 6, "Function Descriptions" in SAS/C Library Reference, Volume 1 for information on failure of these functions.

• The messages generated by the perror function are listed in SAS/C Software Diagnostic Messages.

• calloc , malloc , and realloc return a NULL pointer if the size requested is 0.

• Open files are not closed if the abort function is called. The effect on temporary files is unpredictable and system dependent.

• If the value of the argument to exit is other than 0, EXIT_SUCCESS , or EXIT_FAILURE , the status returned by exit is passed unchanged to CMS or reduced modulo 4096 in OS/390. Under the USS shell, a negative exit status or one greater than 255 is set by the USS kernel to 255.

• See the documentation for the getenv function in Chapter 6, "Function Descriptions," in SAS/C Library Reference, Volume 1 for information on the set of environment names and the method for altering the environment list.

• See the documentation for the system function in Chapter 6, "Function Descriptions," in SAS/C Library Reference, Volume 1 for information on the contents and mode of execution of the string involved.

• The contents of the error message strings returned by the strerror function are the same as the perror messages.

• The local time zone can be defined by setting the TZ environment variable. See the description of tzset in Chapter 6, "Function Descriptions," in SAS/C Library Reference, Volume 1. If TZ is not set, the offset from Greenwich time is determined by the operating system, and no Daylight Savings Time information is available.

• The era for the clock function under OS/390 is set at the first call to clock , under CMS, at virtual machine log on.

Locale-specific behavior (G.4)

• The execution character set contains all EBCDIC characters in all locales.

• The direction of printing is left-to-right.

• A period (.) is the decimal point character.

• See Chapter 3 in SAS/C Library Reference, Volume 1 for the documentation of the ctype functions for the implementation-defined aspects of character-testing and case-mapping functions.

• The collation sequence for EBCDIC characters is used as the collation sequence of the execution character set.

• SAS/C software uses the normal U.S. time and date conventions for the "C" locale. The exact lisitngs are shown in C Locale U.S. Date/Time Conversions (both locales are the same).

C Locale U.S. Date/Time Conversions

Type	Values
Abbreviated weekday name	Sun	Thu
	Mon	Fri
	Tue	Sat
	Wed
Full weekday name	Sunday	Thursday
	Monday	Friday
	Tuesday	Saturday
	Wednesday
Abbreviated month name	Jan	Jul
	Feb	Aug
	Mar	Sep
	Apr	Oct
	May	Nov
	Jun	Dec
Full month name	January	July
	February	August
	March	September
	April	October
	May	November
	June	December
Date representation	Mnn dd yyyy
	May 01 1989
Time representation	hh:mm:ss
	01:22:45

Note that conventions for user-written locales are defined by the user.

Chapter Contents

Previous

Next

Top of Page

Copyright © 2001 by SAS Institute Inc., Cary, NC, USA. All rights reserved.