1 .\" Copyright (c) 2008, Linux Foundation, written by Michael Kerrisk
2 .\" <mtk.manpages@gmail.com>
4 .\" %%%LICENSE_START(VERBATIM)
5 .\" Permission is granted to make and distribute verbatim copies of this
6 .\" manual provided the copyright notice and this permission notice are
7 .\" preserved on all copies.
9 .\" Permission is granted to copy and distribute modified versions of this
10 .\" manual under the conditions for verbatim copying, provided that the
11 .\" entire resulting derived work is distributed under the terms of a
12 .\" permission notice identical to this one.
14 .\" Since the Linux kernel and libraries are constantly changing, this
15 .\" manual page may be incorrect or out-of-date. The author(s) assume no
16 .\" responsibility for errors or omissions, or for damages resulting from
17 .\" the use of the information contained herein. The author(s) may not
18 .\" have taken the same level of care in the production of this manual,
19 .\" which is licensed free of charge, as they might when working
22 .\" Formatted or processed versions of this manual, if unaccompanied by
23 .\" the source, must acknowledge the copyright and authors of this work.
26 .TH MATHERR 3 2021-03-22 "Linux" "Linux Programmer's Manual"
28 matherr \- SVID math library exception handling
33 .BI "int matherr(struct exception *" exc );
35 .B extern _LIB_VERSION_TYPE _LIB_VERSION;
41 the mechanism described in this page is no longer supported by glibc.
42 Before glibc 2.27, it had been marked as obsolete.
44 .\" glibc commit 813378e9fe17e029caf627cab76fe23eb46815fa
45 the mechanism has been removed altogether.
46 New applications should use the techniques described in
50 This page documents the
52 mechanism as an aid for maintaining and porting older applications.
54 The System V Interface Definition (SVID) specifies that various
55 math functions should invoke a function called
57 if a math exception is detected.
58 This function is called before the math function returns;
61 returns, the system then returns to the math function,
62 which in turn returns to the caller.
66 the programmer must define the
74 to the external variable
77 The system provides a default version of
79 This version does nothing, and returns zero
80 (see below for the significance of this).
83 can be overridden by a programmer-defined
84 version, which will be invoked when an exception occurs.
85 The function is invoked with one argument, a pointer to an
87 structure, defined as follows:
92 int type; /* Exception type */
93 char *name; /* Name of function causing exception */
94 double arg1; /* 1st argument to function */
95 double arg2; /* 2nd argument to function */
96 double retval; /* Function return value */
103 field has one of the following values:
106 A domain error occurred (the function argument was outside the range
107 for which the function is defined).
108 The return value depends on the function;
114 A pole error occurred (the function result is an infinity).
115 The return value in most cases is
117 (the largest single precision floating-point number),
118 appropriately signed.
125 An overflow occurred.
126 In most cases, the value
134 An underflow occurred.
141 Total loss of significance.
148 Partial loss of significance.
149 This value is unused on glibc
150 (and many other systems).
156 fields are the arguments supplied to the function
158 is undefined for functions that take only one argument).
162 field specifies the return value that the math
163 function will return to its caller.
164 The programmer-defined
166 can modify this field to change the return value of the math function.
170 function returns zero, then the system sets
172 as described above, and may print an error message on standard error
177 function returns a nonzero value, then the system does not set
179 and doesn't print an error message.
180 .SS Math functions that employ matherr()
181 The table below lists the functions and circumstances in which
184 The "Type" column indicates the value assigned to
188 The "Result" column is the default return value assigned to
191 The "Msg?" and "errno" columns describe the default behavior if
194 If the "Msg?" columns contains "y",
195 then the system prints an error message on standard error.
197 The table uses the following notations and abbreviations:
202 x first argument to function
203 y second argument to function
204 fin finite value for argument
205 neg negative value for argument
206 int integral value for argument
207 o/f result overflowed
208 u/f result underflowed
209 |x| absolute value of x
210 X_TLOSS is a constant defined in \fI<math.h>\fP
213 .\" Details below from glibc 2.8's sysdeps/ieee754/k_standard.c
214 .\" A subset of cases were test by experimental programs.
218 Function Type Result Msg? errno
219 acos(|x|>1) DOMAIN HUGE y EDOM
220 asin(|x|>1) DOMAIN HUGE y EDOM
221 atan2(0,0) DOMAIN HUGE y EDOM
222 acosh(x<1) DOMAIN NAN y EDOM \" retval is 0.0/0.0
223 atanh(|x|>1) DOMAIN NAN y EDOM \" retval is 0.0/0.0
224 atanh(|x|==1) SING (x>0.0)? y EDOM \" retval is x/0.0
227 cosh(fin) o/f OVERFLOW HUGE n ERANGE
228 sinh(fin) o/f OVERFLOW (x>0.0) ? n ERANGE
230 sqrt(x<0) DOMAIN 0.0 y EDOM
231 hypot(fin,fin) o/f OVERFLOW HUGE n ERANGE
232 exp(fin) o/f OVERFLOW HUGE n ERANGE
233 exp(fin) u/f UNDERFLOW 0.0 n ERANGE
234 exp2(fin) o/f OVERFLOW HUGE n ERANGE
235 exp2(fin) u/f UNDERFLOW 0.0 n ERANGE
236 exp10(fin) o/f OVERFLOW HUGE n ERANGE
237 exp10(fin) u/f UNDERFLOW 0.0 n ERANGE
238 j0(|x|>X_TLOSS) TLOSS 0.0 y ERANGE
239 j1(|x|>X_TLOSS) TLOSS 0.0 y ERANGE
240 jn(|x|>X_TLOSS) TLOSS 0.0 y ERANGE
241 y0(x>X_TLOSS) TLOSS 0.0 y ERANGE
242 y1(x>X_TLOSS) TLOSS 0.0 y ERANGE
243 yn(x>X_TLOSS) TLOSS 0.0 y ERANGE
244 y0(0) DOMAIN \-HUGE y EDOM
245 y0(x<0) DOMAIN \-HUGE y EDOM
246 y1(0) DOMAIN \-HUGE y EDOM
247 y1(x<0) DOMAIN \-HUGE y EDOM
248 yn(n,0) DOMAIN \-HUGE y EDOM
249 yn(x<0) DOMAIN \-HUGE y EDOM
250 lgamma(fin) o/f OVERFLOW HUGE n ERANGE
251 lgamma(\-int) or SING HUGE y EDOM
253 tgamma(fin) o/f OVERFLOW HUGE_VAL n ERANGE
254 tgamma(\-int) SING NAN y EDOM
255 tgamma(0) SING copysign( y ERANGE
257 log(0) SING \-HUGE y EDOM
258 log(x<0) DOMAIN \-HUGE y EDOM
259 log2(0) SING \-HUGE n EDOM \" different from log()
260 log2(x<0) DOMAIN \-HUGE n EDOM \" different from log()
261 log10(0) SING \-HUGE y EDOM
262 log10(x<0) DOMAIN \-HUGE y EDOM
263 pow(0.0,0.0) DOMAIN 0.0 y EDOM
264 pow(x,y) o/f OVERFLOW HUGE n ERANGE
265 pow(x,y) u/f UNDERFLOW 0.0 n ERANGE
266 pow(NaN,0.0) DOMAIN x n EDOM
267 0**neg DOMAIN 0.0 y EDOM \" +0 and -0
268 neg**non-int DOMAIN 0.0 y EDOM
269 scalb() o/f OVERFLOW (x>0.0) ? n ERANGE
272 scalb() u/f UNDERFLOW copysign( n ERANGE
274 fmod(x,0) DOMAIN x y EDOM
275 remainder(x,0) DOMAIN NAN y EDOM \" retval is 0.0/0.0
278 For an explanation of the terms used in this section, see
286 Interface Attribute Value
289 T} Thread safety MT-Safe
295 The example program demonstrates the use of
299 The program takes up to three command-line arguments.
300 The first argument is the floating-point number to be given to
302 If the optional second argument is provided, then
308 is called, and the integer supplied in the
309 command-line argument is used as the return value from
311 If the optional third command-line argument is supplied,
312 then it specifies an alternative return value that
314 should assign as the return value of the math function.
316 The following example run, where
318 is given an argument of 0.0, does not use
323 .RB "$" " ./a.out 0.0"
324 errno: Numerical result out of range
329 In the following run,
331 is called, and returns 0:
335 .RB "$" " ./a.out 0.0 0"
336 matherr SING exception in log() function
337 args: 0.000000, 0.000000
338 retval: \-340282346638528859811704183484516925440.000000
340 errno: Numerical argument out of domain
341 x=\-340282346638528859811704183484516925440.000000
345 The message "log: SING error" was printed by the C library.
347 In the following run,
349 is called, and returns a nonzero value:
353 .RB "$" " ./a.out 0.0 1"
354 matherr SING exception in log() function
355 args: 0.000000, 0.000000
356 retval: \-340282346638528859811704183484516925440.000000
357 x=\-340282346638528859811704183484516925440.000000
361 In this case, the C library did not print a message, and
365 In the following run,
367 is called, changes the return value of the math function,
368 and returns a nonzero value:
372 .RB "$" " ./a.out 0.0 1 12345.0"
373 matherr SING exception in log() function
374 args: 0.000000, 0.000000
375 retval: \-340282346638528859811704183484516925440.000000
388 static int matherr_ret = 0; /* Value that matherr()
390 static int change_retval = 0; /* Should matherr() change
391 function\(aqs return value? */
392 static double new_retval; /* New function return value */
395 matherr(struct exception *exc)
397 fprintf(stderr, "matherr %s exception in %s() function\en",
398 (exc\->type == DOMAIN) ? "DOMAIN" :
399 (exc\->type == OVERFLOW) ? "OVERFLOW" :
400 (exc\->type == UNDERFLOW) ? "UNDERFLOW" :
401 (exc\->type == SING) ? "SING" :
402 (exc\->type == TLOSS) ? "TLOSS" :
403 (exc\->type == PLOSS) ? "PLOSS" : "???",
405 fprintf(stderr, " args: %f, %f\en",
406 exc\->arg1, exc\->arg2);
407 fprintf(stderr, " retval: %f\en", exc\->retval);
410 exc\->retval = new_retval;
416 main(int argc, char *argv[])
421 fprintf(stderr, "Usage: %s <argval>"
422 " [<matherr\-ret> [<new\-func\-retval>]]\en", argv[0]);
427 _LIB_VERSION = _SVID_;
428 matherr_ret = atoi(argv[2]);
433 new_retval = atof(argv[3]);
436 x = log(atof(argv[1]));
440 printf("x=%f\en", x);