0
0
mirror of https://github.com/python/cpython.git synced 2024-11-24 17:47:13 +01:00
cpython/Include/pyfpe.h
Guido van Rossum 1f06beeedd Change PyFPE_END_PROTECT to PyFPE_END_PROTECT(v). v should be the
last variable to which a floating point expression is assigned.  The
macro passes its address to a dummy function so that the optimizer
can't delay calculating its value until after the macro.
1997-03-14 04:23:42 +00:00

165 lines
7.6 KiB
C

#ifndef Py_PYFPE_H
#define Py_PYFPE_H
#ifdef __cplusplus
extern "C" {
#endif
/*
---------------------------------------------------------------------
/ Copyright (c) 1996. \
| The Regents of the University of California. |
| All rights reserved. |
| |
| Permission to use, copy, modify, and distribute this software for |
| any purpose without fee is hereby granted, provided that this en- |
| tire notice is included in all copies of any software which is or |
| includes a copy or modification of this software and in all |
| copies of the supporting documentation for such software. |
| |
| This work was produced at the University of California, Lawrence |
| Livermore National Laboratory under contract no. W-7405-ENG-48 |
| between the U.S. Department of Energy and The Regents of the |
| University of California for the operation of UC LLNL. |
| |
| DISCLAIMER |
| |
| This software was prepared as an account of work sponsored by an |
| agency of the United States Government. Neither the United States |
| Government nor the University of California nor any of their em- |
| ployees, makes any warranty, express or implied, or assumes any |
| liability or responsibility for the accuracy, completeness, or |
| usefulness of any information, apparatus, product, or process |
| disclosed, or represents that its use would not infringe |
| privately-owned rights. Reference herein to any specific commer- |
| cial products, process, or service by trade name, trademark, |
| manufacturer, or otherwise, does not necessarily constitute or |
| imply its endorsement, recommendation, or favoring by the United |
| States Government or the University of California. The views and |
| opinions of authors expressed herein do not necessarily state or |
| reflect those of the United States Government or the University |
| of California, and shall not be used for advertising or product |
\ endorsement purposes. /
---------------------------------------------------------------------
*/
/*
* Define macros for handling SIGFPE.
* Lee Busby, LLNL, November, 1996
* busby1@llnl.gov
*
*********************************************
* Overview of the system for handling SIGFPE:
*
* This file (Include/pyfpe.h) defines a couple of "wrapper" macros for
* insertion into your Python C code of choice. Their proper use is
* discussed below. The file Python/pyfpe.c defines a pair of global
* variables PyFPE_jbuf and PyFPE_counter which are used by the signal
* handler for SIGFPE to decide if a particular exception was protected
* by the macros. The signal handler itself, and code for enabling the
* generation of SIGFPE in the first place, is in a (new) Python module
* named fpectl. This module is standard in every respect. It can be loaded
* either statically or dynamically as you choose, and like any other
* Python module, has no effect until you import it.
*
* In the general case, there are three steps toward handling SIGFPE in any
* Python code:
*
* 1) Add the *_PROTECT macros to your C code as required to protect
* dangerous floating point sections.
*
* 2) Turn on the inclusion of the code by #defining WANT_SIGFPE_HANDLER in
* config.h.in before you configure, compile, and install Python, and the
* fpectl module, and any other modules which may have conditional code.
*
* 3) When python is built and running, import fpectl, and execute
* fpectl.turnon_sigfpe(). This sets up the signal handler and enables
* generation of SIGFPE whenever an exception occurs. From this point
* on, any properly trapped SIGFPE should result in the Python
* FloatingPointError exception.
*
* Step 1 has been done already for the Python kernel code, and will be
* done soon for Hugunin's NumPy array package and my Gist graphics module.
* Step 2 is usually done once at python install time. Python's behavior
* with respect to SIGFPE is not changed unless you also do step 3. Thus
* you can control this new facility at compile time, or run time, or both.
*
********************************
* Using the macros in your code:
*
* static PyObject *foobar(PyObject *self,PyObject *args)
* {
* ....
* PyFPE_START_PROTECT("Error in foobar", return 0)
* dangerous_op(somearg1, somearg2, ...);
* PyFPE_END_PROTECT
* ....
* }
*
* If a floating point error occurs in dangerous_op, foobar returns 0
* (NULL), after setting the associated value of the FloatingPointError
* exception to "Error in foobar". ``Dangerous_op'' can be a single
* operation, or a block, or function calls, or any combination, so long as
* no alternate return is possible before the PyFPE_END_PROTECT macro is
* reached.
*
* The macros can only be used in a function context where an error return
* can be recognized as signaling a Python exception. (Generally, most
* functions that return a PyObject * will qualify.)
*
* Guido's original design suggestion for PyFPE_START_PROTECT and
* PyFPE_END_PROTECT had them open and close a local block, with a locally
* defined jmp_buf and jmp_buf pointer. This would allow recursive nesting
* of the macros. The Ansi C standard makes it clear that such local
* variables need to be declared with the "volatile" type qualifier to keep
* setjmp from corrupting their values. Some current implementations seem
* to be more restrictive. For example, the HPUX man page for setjmp says
*
* Upon the return from a setjmp() call caused by a longjmp(), the
* values of any non-static local variables belonging to the routine
* from which setjmp() was called are undefined. Code which depends on
* such values is not guaranteed to be portable.
*
* I therefore decided on a more limited form of nesting, using a counter
* variable (PyFPE_counter) to keep track of any recursion. If an exception
* occurs in an ``inner'' pair of macros, the return will apparently
* come from the top level.
*
*/
#ifdef WANT_SIGFPE_HANDLER
#include <signal.h>
#include <setjmp.h>
#include <math.h>
extern jmp_buf PyFPE_jbuf;
extern int PyFPE_counter;
extern double PyFPE_dummy();
#define PyFPE_START_PROTECT(err_string, leave_stmt) \
if (!PyFPE_counter++ && setjmp(PyFPE_jbuf)) { \
PyFPE_counter = 0; \
PyErr_SetString(PyExc_FloatingPointError, err_string); \
leave_stmt; \
}
/*
* This (following) is a heck of a way to decrement a counter. However,
* code optimizers will sometimes move this statement so that it gets
* executed *before* the unsafe expression which we're trying to protect.
* This pretty well messes things up, of course. So the best I've been able
* to do is to put a (hopefully fast) function call into the expression
* which counts down PyFPE_counter, and thereby monkey wrench the overeager
* optimizer. Better solutions are welcomed....
*/
#define PyFPE_END_PROTECT(v) PyFPE_counter -= (int)PyFPE_dummy(&(v));
#else
#define PyFPE_START_PROTECT(err_string, leave_stmt)
#define PyFPE_END_PROTECT(v)
#endif
#ifdef __cplusplus
}
#endif
#endif /* !Py_PYFPE_H */