mirror of
https://sourceware.org/git/binutils-gdb.git
synced 2024-12-05 08:14:19 +08:00
3666a04883
This commits the result of running gdb/copyright.py as per our Start of New Year procedure... gdb/ChangeLog Update copyright year range in copyright header of all GDB files.
289 lines
7.8 KiB
C++
289 lines
7.8 KiB
C++
/* Miscellaneous routines making it easier to use GMP within GDB's framework.
|
|
|
|
Copyright (C) 2019-2021 Free Software Foundation, Inc.
|
|
|
|
This file is part of GDB.
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>. */
|
|
|
|
#ifndef GMP_UTILS_H
|
|
#define GMP_UTILS_H
|
|
|
|
#include "defs.h"
|
|
|
|
/* Include <stdio.h> and <stdarg.h> ahead of <gmp.h>, so as to get
|
|
access to GMP's various formatting functions. */
|
|
#include <stdio.h>
|
|
#include <stdarg.h>
|
|
#include <gmp.h>
|
|
#include "gdbsupport/traits.h"
|
|
|
|
/* Same as gmp_asprintf, but returning an std::string. */
|
|
|
|
std::string gmp_string_printf (const char *fmt, ...);
|
|
|
|
/* A class to make it easier to use GMP's mpz_t values within GDB. */
|
|
|
|
struct gdb_mpz
|
|
{
|
|
mpz_t val;
|
|
|
|
/* Constructors. */
|
|
gdb_mpz () { mpz_init (val); }
|
|
|
|
explicit gdb_mpz (const mpz_t &from_val)
|
|
{
|
|
mpz_init (val);
|
|
mpz_set (val, from_val);
|
|
}
|
|
|
|
gdb_mpz (const gdb_mpz &from)
|
|
{
|
|
mpz_init (val);
|
|
mpz_set (val, from.val);
|
|
}
|
|
|
|
/* Initialize using the given integral value.
|
|
|
|
The main advantage of this method is that it handles both signed
|
|
and unsigned types, with no size restriction. */
|
|
template<typename T, typename = gdb::Requires<std::is_integral<T>>>
|
|
explicit gdb_mpz (T src)
|
|
{
|
|
mpz_init (val);
|
|
set (src);
|
|
}
|
|
|
|
explicit gdb_mpz (gdb_mpz &&from)
|
|
{
|
|
mpz_init (val);
|
|
mpz_swap (val, from.val);
|
|
}
|
|
|
|
|
|
gdb_mpz &operator= (const gdb_mpz &from)
|
|
{
|
|
mpz_set (val, from.val);
|
|
return *this;
|
|
}
|
|
|
|
gdb_mpz &operator= (gdb_mpz &&other)
|
|
{
|
|
mpz_swap (val, other.val);
|
|
return *this;
|
|
}
|
|
|
|
template<typename T, typename = gdb::Requires<std::is_integral<T>>>
|
|
gdb_mpz &operator= (T src)
|
|
{
|
|
set (src);
|
|
return *this;
|
|
}
|
|
|
|
/* Convert VAL to an integer of the given type.
|
|
|
|
The return type can signed or unsigned, with no size restriction. */
|
|
template<typename T> T as_integer () const;
|
|
|
|
/* Set VAL by importing the number stored in the byte array (BUF),
|
|
using the given BYTE_ORDER. The size of the data to read is
|
|
the byte array's size.
|
|
|
|
UNSIGNED_P indicates whether the number has an unsigned type. */
|
|
void read (gdb::array_view<const gdb_byte> buf, enum bfd_endian byte_order,
|
|
bool unsigned_p);
|
|
|
|
/* Write VAL into BUF as a number whose byte size is the size of BUF,
|
|
using the given BYTE_ORDER.
|
|
|
|
UNSIGNED_P indicates whether the number has an unsigned type. */
|
|
void write (gdb::array_view<gdb_byte> buf, enum bfd_endian byte_order,
|
|
bool unsigned_p) const;
|
|
|
|
/* Return a string containing VAL. */
|
|
std::string str () const { return gmp_string_printf ("%Zd", val); }
|
|
|
|
/* The destructor. */
|
|
~gdb_mpz () { mpz_clear (val); }
|
|
|
|
private:
|
|
|
|
/* Helper template for constructor and operator=. */
|
|
template<typename T> void set (T src);
|
|
|
|
/* Low-level function to export VAL into BUF as a number whose byte size
|
|
is the size of BUF.
|
|
|
|
If UNSIGNED_P is true, then export VAL into BUF as an unsigned value.
|
|
Otherwise, export it as a signed value.
|
|
|
|
The API is inspired from GMP's mpz_export, hence the naming and types
|
|
of the following parameter:
|
|
- ENDIAN should be:
|
|
. 1 for most significant byte first; or
|
|
. -1 for least significant byte first; or
|
|
. 0 for native endianness.
|
|
|
|
An error is raised if BUF is not large enough to contain the value
|
|
being exported. */
|
|
void safe_export (gdb::array_view<gdb_byte> buf,
|
|
int endian, bool unsigned_p) const;
|
|
};
|
|
|
|
/* A class to make it easier to use GMP's mpq_t values within GDB. */
|
|
|
|
struct gdb_mpq
|
|
{
|
|
mpq_t val;
|
|
|
|
/* Constructors. */
|
|
gdb_mpq () { mpq_init (val); }
|
|
|
|
explicit gdb_mpq (const mpq_t &from_val)
|
|
{
|
|
mpq_init (val);
|
|
mpq_set (val, from_val);
|
|
}
|
|
|
|
gdb_mpq (const gdb_mpq &from)
|
|
{
|
|
mpq_init (val);
|
|
mpq_set (val, from.val);
|
|
}
|
|
|
|
explicit gdb_mpq (gdb_mpq &&from)
|
|
{
|
|
mpq_init (val);
|
|
mpq_swap (val, from.val);
|
|
}
|
|
|
|
/* Copy assignment operator. */
|
|
gdb_mpq &operator= (const gdb_mpq &from)
|
|
{
|
|
mpq_set (val, from.val);
|
|
return *this;
|
|
}
|
|
|
|
gdb_mpq &operator= (gdb_mpq &&from)
|
|
{
|
|
mpq_swap (val, from.val);
|
|
return *this;
|
|
}
|
|
|
|
/* Return a string representing VAL as "<numerator> / <denominator>". */
|
|
std::string str () const { return gmp_string_printf ("%Qd", val); }
|
|
|
|
/* Return VAL rounded to the nearest integer. */
|
|
gdb_mpz get_rounded () const;
|
|
|
|
/* Set VAL from the contents of the given byte array (BUF), which
|
|
contains the unscaled value of a fixed point type object.
|
|
The byte size of the data is the size of BUF.
|
|
|
|
BYTE_ORDER provides the byte_order to use when reading the data.
|
|
|
|
UNSIGNED_P indicates whether the number has an unsigned type.
|
|
SCALING_FACTOR is the scaling factor to apply after having
|
|
read the unscaled value from our buffer. */
|
|
void read_fixed_point (gdb::array_view<const gdb_byte> buf,
|
|
enum bfd_endian byte_order, bool unsigned_p,
|
|
const gdb_mpq &scaling_factor);
|
|
|
|
/* Write VAL into BUF as fixed point value following the given BYTE_ORDER.
|
|
The size of BUF is used as the length to write the value into.
|
|
|
|
UNSIGNED_P indicates whether the number has an unsigned type.
|
|
SCALING_FACTOR is the scaling factor to apply before writing
|
|
the unscaled value to our buffer. */
|
|
void write_fixed_point (gdb::array_view<gdb_byte> buf,
|
|
enum bfd_endian byte_order, bool unsigned_p,
|
|
const gdb_mpq &scaling_factor) const;
|
|
|
|
/* The destructor. */
|
|
~gdb_mpq () { mpq_clear (val); }
|
|
};
|
|
|
|
/* A class to make it easier to use GMP's mpf_t values within GDB.
|
|
|
|
Should MPFR become a required dependency, we should probably
|
|
drop this class in favor of using MPFR. */
|
|
|
|
struct gdb_mpf
|
|
{
|
|
mpf_t val;
|
|
|
|
/* Constructors. */
|
|
gdb_mpf () { mpf_init (val); }
|
|
|
|
DISABLE_COPY_AND_ASSIGN (gdb_mpf);
|
|
|
|
/* Set VAL from the contents of the given buffer (BUF), which
|
|
contains the unscaled value of a fixed point type object
|
|
with the given size (LEN) and byte order (BYTE_ORDER).
|
|
|
|
UNSIGNED_P indicates whether the number has an unsigned type.
|
|
SCALING_FACTOR is the scaling factor to apply after having
|
|
read the unscaled value from our buffer. */
|
|
void read_fixed_point (gdb::array_view<const gdb_byte> buf,
|
|
enum bfd_endian byte_order, bool unsigned_p,
|
|
const gdb_mpq &scaling_factor)
|
|
{
|
|
gdb_mpq tmp_q;
|
|
|
|
tmp_q.read_fixed_point (buf, byte_order, unsigned_p, scaling_factor);
|
|
mpf_set_q (val, tmp_q.val);
|
|
}
|
|
|
|
/* The destructor. */
|
|
~gdb_mpf () { mpf_clear (val); }
|
|
};
|
|
|
|
/* See declaration above. */
|
|
|
|
template<typename T>
|
|
void
|
|
gdb_mpz::set (T src)
|
|
{
|
|
mpz_import (val, 1 /* count */, -1 /* order */,
|
|
sizeof (T) /* size */, 0 /* endian (0 = native) */,
|
|
0 /* nails */, &src /* op */);
|
|
if (std::is_signed<T>::value && src < 0)
|
|
{
|
|
/* mpz_import does not handle the sign, so our value was imported
|
|
as an unsigned. Adjust that imported value so as to make it
|
|
the correct negative value. */
|
|
gdb_mpz neg_offset;
|
|
|
|
mpz_ui_pow_ui (neg_offset.val, 2, sizeof (T) * HOST_CHAR_BIT);
|
|
mpz_sub (val, val, neg_offset.val);
|
|
}
|
|
}
|
|
|
|
/* See declaration above. */
|
|
|
|
template<typename T>
|
|
T
|
|
gdb_mpz::as_integer () const
|
|
{
|
|
T result;
|
|
|
|
this->safe_export ({(gdb_byte *) &result, sizeof (result)},
|
|
0 /* endian (0 = native) */,
|
|
!std::is_signed<T>::value /* unsigned_p */);
|
|
|
|
return result;
|
|
}
|
|
|
|
#endif
|