matherr(3M)
NAME
matherr − math library exception-handling function
SYNOPSIS
#include <math.h>
int matherr(struct exception ∗exc);
MT-LEVEL
MT-Safe
DESCRIPTION
The SVID3 (System V Interface Definition Third Edition) specifies that certain libm functions call matherr() when exceptions are detected. Users may define their own mechanisms for handling exceptions, by including a function named matherr() in their programs. matherr() is of the form described above. When an exception occurs, a pointer to the exception structure exc will be passed to the user-supplied matherr() function. This structure, which is defined in the <math.h> header file, is as follows:
struct exception {
int type;
char ∗name;
double arg1, arg2, retval;
};
The element type is an integer describing the type of exception that has occurred, from the following list of constants (defined in the header file):
DOMAIN argument domain exception
SING argument singularity
OVERFLOW overflow range exception
UNDERFLOW underflow range exception
TLOSS total loss of significance
PLOSS partial loss of significance
Note that both TLOSS and PLOSS reflect limitations of particular algorithms for trigonometric functions that suffer abrupt declines in accuracy at definite boundaries. Since the Sun implementation does not suffer such abrupt declines, PLOSS is never signaled. TLOSS is signaled for Bessel functions only to satisfy SVID3 requirements.
The element name points to a string containing the name of the function that incurred the exception. The elements arg1 and arg2 are the arguments with which the function was invoked. retval is set to the default value that will be returned by the function unless the user’s matherr() sets it to a different value.
If the user’s matherr() function returns non-zero, no exception message will be printed, and errno will not be set.
SVID3 STANDARD CONFORMANCE
If matherr() is not supplied by the user, the default matherr exception-handling mechanisms, summarized in the table below, will be invoked upon exception:
DOMAIN
0.0 is usually returned, errno is set to EDOM, and a message is usually printed on standard error.
SING
The largest finite single-precision number, HUGE of appropriate sign is returned, errno is set to EDOM, and a message is printed on standard error.
OVERFLOW
The largest finite single-precision number, HUGE of appropriate sign is usually returned, errno is set to ERANGE.
UNDERFLOW
0.0 is returned, and errno is set to ERANGE.
TLOSS
0.0 is returned, errno is set to ERANGE, and a message is printed on standard error.
In general, errno is not a reliable error indicator in that it may be unexpectedly set by a function in a handler for an asynchronous signal.
| DEFAULT ERROR HANDLING PROCEDURES (SVID3) | |||||
| Types of Errors | |||||
| <math.h> type | DOMAIN | SING | OVERFLOW | UNDERFLOW | TLOSS |
| errno | EDOM | EDOM | ERANGE | ERANGE | ERANGE |
| IEEE Exception | Invalid Operation | Division by Zero | Overflow | Underflow | − |
| fp_exception_type | fp_invalid | fp_division | fp_overflow | fp_underflow | − |
| ACOS, ASIN(|x| > 1): | Md, 0.0 | − | − | − | − |
| ACOSH(x < 1), ATANH(|x| > 1): | NaN | − | − | − | − |
| ATAN2(0,0): | Md, 0.0 | − | − | − | − |
| COSH, SINH: | − | − | ±HUGE | − | − |
| EXP: | − | − | +HUGE | 0.0 | − |
| FMOD(x,0): | x | − | − | − | − |
| HYPOT: | − | − | +HUGE | − | − |
| J0, J1, JN(|x| > X_TLOSS): | − | − | − | − | Mt, 0.0 |
| LGAMMA: | |||||
| usual cases | − | − | +HUGE | − | − |
| (x = 0 or −integer) | − | Ms, +HUGE | − | − | − |
| LOG, LOG10: | |||||
| (x < 0) | Md, −HUGE | − | − | − | − |
| (x = 0) | − | Ms, −HUGE | − | − | − |
| POW: | |||||
| usual cases | − | − | ±HUGE | ±0.0 | − |
| (x < 0) ∗∗ (y not an integer) | Md, 0.0 | − | − | − | − |
| 0 ∗∗ 0 | Md, 0.0 | − | − | − | − |
| 0 ∗∗ (y < 0) | Md, 0.0 | − | − | − | − |
| REMAINDER(x,0): | NaN | − | − | − | − |
| SCALB: | − | − | ±HUGE_VAL | ±0.0 | − |
| SQRT(x < 0): | Md, 0.0 | − | − | − | − |
| Y0, Y1, YN: | |||||
| (x < 0) | Md, −HUGE | − | − | − | − |
| (x = 0) | − | Md, −HUGE | − | − | − |
| (x > X_TLOSS) | − | − | − | − | Mt, 0.0 |
| ABBREVIATIONS | |
| Md | Message is printed (DOMAIN error). |
| Ms | Message is printed (SING error). |
| Mt | Message is printed (TLOSS error). |
| NaN | IEEE NaN result and invalid operation exception. |
| HUGE | Maximum finite single-precision floating-point number. |
| HUGE_VAL | IEEE ∞ result and division-by-zero exception. |
| X_TLOSS | The value X_TLOSS is defined in <values.h>. |
The interaction of IEEE arithmetic and matherr() is not defined when executing under IEEE rounding modes other than the default round to nearest: matherr() may not be called on overflow or underflow, and the SUN-provided matherr() may return results that differ from those in this table.
X/OPEN (XPG3) STANDARD CONFORMANCE
XPG3 (X/Open Portability Guide Issue 3) no longer sanctions the use of the matherr() interface. The following table summarizes the values returned in the exceptional cases. In general, XPG3 dictates that as long as one of the input argument(s) is a NaN, NaN shall be returned. In particular, pow(NaN,0) = NaN.
| X/Open (XPG3) ERROR HANDLING PROCEDURES (compile with "cc −Xa") | |||||
| Types of Errors | |||||
| <math.h> type | DOMAIN | SING | OVERFLOW | UNDERFLOW | TLOSS |
| errno | EDOM | EDOM | ERANGE | ERANGE | ERANGE |
| ACOS, ASIN(|x| > 1): | 0.0 | − | − | − | − |
| ATAN2(0,0): | 0.0 | − | − | − | − |
| COSH, SINH: | − | − | {±HUGE_VAL} | − | − |
| EXP: | − | − | {+HUGE_VAL} | {0.0} | − |
| FMOD(x,0): | {NaN} | − | − | − | − |
| HYPOT: | − | − | {+HUGE_VAL} | − | − |
| J0, J1, JN(|x| > X_TLOSS): | − | − | − | − | {0.0} |
| LGAMMA: | |||||
| usual cases | − | − | {+HUGE_VAL} | − | − |
| (x = 0 or −integer) | − | +HUGE_VAL | − | − | − |
| LOG, LOG10: | |||||
| (x < 0) | −HUGE_VAL | − | − | − | − |
| (x = 0) | − | −HUGE_VAL | − | − | − |
| POW: | |||||
| usual cases | − | − | ±HUGE_VAL | ±0.0 | − |
| (x < 0) ∗∗ (y not an integer) | 0.0 | − | − | − | − |
| 0 ∗∗ 0 | {1.0} | − | − | − | − |
| 0 ∗∗ (y < 0) | {−HUGE_VAL} | − | − | − | − |
| SQRT(x < 0): | 0.0 | − | − | − | − |
| Y0, Y1, YN: | |||||
| (x < 0) | {−HUGE_VAL} | − | − | − | − |
| (x = 0) | − | {−HUGE_VAL} | − | − | − |
| (x > X_TLOSS) | − | − | − | − | 0.0 |
| ABBREVIATIONS | |
| {...} | errno is not to be relied upon in all braced cases. |
| NaN | IEEE NaN result and invalid operation exception. |
| HUGE_VAL | IEEE ∞ result and division-by-zero exception. |
| X_TLOSS | The value X_TLOSS is defined in <values.h>. |
ANSI/ISO-C STANDARD CONFORMANCE
The ANSI/ISO-C standard covers a small subset of XPG3.
The following table summarizes the values returned in the exceptional cases.
| ANSI/ISO-C ERROR HANDLING PROCEDURES (compile with "cc −Xc") | ||||
| Types of Errors | ||||
| <math.h> type | DOMAIN | SING | OVERFLOW | UNDERFLOW |
| errno | EDOM | EDOM | ERANGE | ERANGE |
| ACOS, ASIN(|x| > 1): | 0.0 | − | − | − |
| ATAN2(0,0): | 0.0 | − | − | − |
| EXP: | − | − | +HUGE_VAL | 0.0 |
| FMOD(x,0): | NaN | − | − | − |
| LOG, LOG10: | ||||
| (x < 0) | −HUGE_VAL | − | − | − |
| (x = 0) | − | −HUGE_VAL | − | − |
| POW: | ||||
| usual cases | − | − | ±HUGE_VAL | ±0.0 |
| (x < 0) ∗∗ (y not an integer) | 0.0 | − | − | − |
| 0 ∗∗ (y < 0) | −HUGE_VAL | − | − | − |
| SQRT(x < 0): | 0.0 | − | − | − |
| ABBREVIATIONS | |
| NaN | IEEE NaN result and invalid operation exception. |
| HUGE_VAL | IEEE ∞ result and division-by-zero exception. |
EXAMPLES
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
int
matherr(struct exception ∗x) {
switch (x−>type) {
case DOMAIN:
/∗ change sqrt to return sqrt(−arg1), not NaN ∗/
if (!strcmp(x−>name, "sqrt")) {
x−>retval = sqrt(−x−>arg1);
return (0); /∗ print message and set errno ∗/
} /∗ FALLTHRU ∗/
case SING:
/∗ all other domain or sing exceptions, print message and ∗/
/∗ abort ∗/
fprintf(stderr, "domain exception in %s\n", x−>name);
abort( );
break;
}
return (0); /∗ all other exceptions, execute default procedure ∗/
}
SunOS 5.4 — Last change: 4 Mar 1994