Header File
signal.h
Category
Process Control Routines
Prototype
void (_USERENTRY *signal(int sig, void (_USERENTRY *func)(int sig[, int subcode])))(int);
Description
Specifies signal-handling actions.
signal determines how receipt of signal number sig will subsequently be treated. You can install a user-specified handler routine (specified by the argument func) or use one of the two predefined handlers, SIG_DFL and SIG_IGN, in signal.h. The function func must be declared with the _USERENTRY calling convention.
A routine that catches a signal (such as a floating point) also clears the signal. To continue to receive signals, a signal handler must be reinstalled by calling signal again.
SIG_DFL |
Terminates the program |
SIG_ERR |
Indicates an error return from signal |
SIG_IGN |
Ignore this type signal |
The following table shows signal types and their defaults:
SIGBREAK |
Keyboard must be in raw mode. |
SIGABRT |
Abnormal termination. Default action is equivalent to calling _exit(3). |
SIGFPE |
Arithmetic error caused by division by 0, invalid operation, and the like. Default action is equivalent to calling _exit(1). |
SIGILL |
Illegal operation. Default action is equivalent to calling _exit(1). |
SIGINT |
Ctrl-C interrupt. Default action is equivalent to calling _exit(3). |
SIGSEGV |
Illegal storage access. Default action is equivalent to calling _exit(1). |
SIGTERM |
Request for program termination. Default action is equivalent to calling _exit(1). |
|
|
|
User-defined signals can be generated only by calling raise. Default action is to ignore the signal. |
signal.h defines a type called sig_atomic_t, the largest integer type the processor can load or store atomically in the presence of asynchronous interrupts (this is a 32-bit integer -- a Borland C++ integer).
When a signal is generated by the raise function or by an external event, the following two things happen:
Borland C++ implements an extension to ANSI C when the signal type is SIGFPE, SIGSEGV, or SIGILL. The user-specified handler function is called with one or two extra parameters. If SIGFPE, SIGSEGV, or SIGILL has been raised as the result of an explicit call to the raise function, the user-specified handler is called with one extra parameter, an integer specifying that the handler is being explicitly invoked. The explicit activation values for SIGFPE, SIGSEGV and SIGILL are as follows
SIGFPE |
FPE_EXPLICITGEN |
SIGSEGV |
SEGV_EXPLICITGEN |
SIGILL |
ILL_EXPLICITGEN |
If SIGFPE is raised because of a floating-point exception, the user handler is called with one extra parameter that specifies the FPE_xxx type of the signal. If SIGSEGV, SIGILL, or the integer-related variants of SIGFPE signals (FPE_INTOVFLOW or FPE_INTDIV0) are raised as the result of a processor exception, the user handler is called with two extra parameters:
1.The SIGFPE, SIGSEGV, or SIGILL exception type (see float.h for all these types). This first parameter is the usual ANSI signal type.
2.An integer pointer into the stack of the interrupt handler that called the user-specified handler. This pointer points to a list of the processor registers saved when the exception occurred. The registers are in the same order as the parameters to an interrupt function; that is, EBP, EDI, ESI, EDS, ES, EDX, ECX, EBX, EAX, EIP, CS, EFL. To have a register value changed when the handler returns, change one of the locations in this list.
For example, to have a new SI value on return, do something like this:
*((int*)list_pointer + 2) = new_SI_value;
In this way, the handler can examine and make any adjustments to the registers that you want.
The following SIGFPE-type signals can occur (or be generated). They correspond to the exceptions that the 80x87 family is capable of detecting, as well as the "INTEGER DIVIDE BY ZERO" and the "INTERRUPT ON OVERFLOW" on the main CPU. (The declarations for these are in float.h.)
FPE_INTOVFLOW |
INTO executed with OF flag set |
FPE_INTDIV0 |
Integer divide by zero |
FPE_INVALID |
Invalid operation |
FPE_ZERODIVIDE |
Division by zero |
FPE_OVERFLOW |
Numeric overflow |
FPE_UNDERFLOW |
Numeric underflow |
FPE_INEXACT |
Precision |
FPE_EXPLICITGEN |
User program executed raise(SIGFPE) |
FPE_STACKFAULT |
Floating-point stack overflow or underflow |
FPE_STACKFAULT |
Stack overflow |
The FPE_INTOVFLOW and FPE_INTDIV0 signals are generated by integer operations, and the others are generated by floating-point operations. Whether the floating-point exceptions are generated depends on the coprocessor control word, which can be modified with _control87. Denormal exceptions are handled by Borland C++ and not passed to a signal handler.
The following SIGSEGV-type signals can occur:
SEGV_BOUND |
Bound constraint exception |
SEGV_EXPLICITGEN |
raise(SIGSEGV) was executed |
The following SIGILL-type signals can occur:
ILL_EXECUTION |
Illegal operation attempted |
ILL_EXPLICITGEN |
raise(SIGILL) was executed |
When the signal type is SIGFPE, SIGSEGV, or SIGILL, a return from a signal handler is generally not advisable if the state of the floating point processor is corrupt, the results of an integer division are wrong, an operation that shouldn't have overflowed did, a bound instruction failed, or an illegal operation was attempted. The only time a return is reasonable is when the handler alters the registers so that a reasonable return context exists or the signal type indicates that the signal was generated explicitly (for example, FPE_EXPLICITGEN, SEGV_EXPLICITGEN, or ILL_EXPLICITGEN). Generally in this case you would print an error message and terminate the program using _exit, exit, or abort. If a return is executed under any other conditions, the program's action will probably be unpredictable.
On success, signal returns a pointer to the previous handler routine for the specified signal type.
On error, signal returns SIG_ERR, and the external variable errno is set to EINVAL.
Example
/* signal example */ /* This example installs a signal handler routine for SIGFPE, catches an integer overflow condition, makes an adjustment to AX register, and returns. This example program MAY cause your computer to crash, and will produce runtime errors depending on which memory model is used. */ #pragma inline #include <stdio.h> #include <signal.h> #ifdef __cplusplus typedef void (*fptr)(int); #else typedef void (*fptr)(); #endif void Catcher(int *reglist) { signal(SIGFPE, (fptr)Catcher); // ******reinstall signal handler printf("Caught it!\n"); *(reglist + 8) = 3; /* make return AX = 3 */ } int main(void) { signal(SIGFPE, (fptr)Catcher); /* cast Catcher to appropriate type */ asm mov ax,07FFFH /* AX = 32767 */ asm inc ax /* cause overflow */ asm into /* activate handler */ /* The handler set AX to 3 on return. If that had not happened, there would have been another exception when the next 'into' executed after the 'dec' instruction. */ asm dec ax /* no overflow now */ asm into /* doesn't activate */ return 0; }
Portability
POSIX |
Win32 |
ANSI C |
ANSI C++ |
+ |
+ |
+ |
+ |
Copyright(C) 2008 CodeGear(TM). All Rights Reserved.
|
What do you think about this topic? Send feedback!
|