1997-09-24 20:57:14 +04:00
|
|
|
|
/* Print values for GDB, the GNU debugger.
|
|
|
|
|
Copyright 1986, 1988, 1989, 1991, 1992, 1993, 1994
|
|
|
|
|
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 2 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, write to the Free Software
|
|
|
|
|
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
|
|
|
|
|
|
|
|
|
|
#include "defs.h"
|
|
|
|
|
#include "gdb_string.h"
|
|
|
|
|
#include "symtab.h"
|
|
|
|
|
#include "gdbtypes.h"
|
|
|
|
|
#include "value.h"
|
|
|
|
|
#include "gdbcore.h"
|
|
|
|
|
#include "gdbcmd.h"
|
|
|
|
|
#include "target.h"
|
|
|
|
|
#include "obstack.h"
|
|
|
|
|
#include "language.h"
|
|
|
|
|
#include "demangle.h"
|
|
|
|
|
#include "annotate.h"
|
1999-02-11 01:06:10 +03:00
|
|
|
|
#include "valprint.h"
|
1997-09-24 20:57:14 +04:00
|
|
|
|
|
|
|
|
|
#include <errno.h>
|
|
|
|
|
|
|
|
|
|
/* Prototypes for local functions */
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
print_hex_chars PARAMS ((GDB_FILE *, unsigned char *, unsigned int));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
show_print PARAMS ((char *, int));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_print PARAMS ((char *, int));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_radix PARAMS ((char *, int));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
show_radix PARAMS ((char *, int));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_input_radix PARAMS ((char *, int, struct cmd_list_element *));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_input_radix_1 PARAMS ((int, unsigned));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_output_radix PARAMS ((char *, int, struct cmd_list_element *));
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_output_radix_1 PARAMS ((int, unsigned));
|
|
|
|
|
|
|
|
|
|
/* Maximum number of chars to print for a string pointer value or vector
|
|
|
|
|
contents, or UINT_MAX for no limit. Note that "set print elements 0"
|
|
|
|
|
stores UINT_MAX in print_max, which displays in a show command as
|
|
|
|
|
"unlimited". */
|
|
|
|
|
|
|
|
|
|
unsigned int print_max;
|
|
|
|
|
#define PRINT_MAX_DEFAULT 200 /* Start print_max off at this value. */
|
|
|
|
|
|
|
|
|
|
/* Default input and output radixes, and output format letter. */
|
|
|
|
|
|
|
|
|
|
unsigned input_radix = 10;
|
|
|
|
|
unsigned output_radix = 10;
|
|
|
|
|
int output_format = 0;
|
|
|
|
|
|
|
|
|
|
/* Print repeat counts if there are more than this many repetitions of an
|
|
|
|
|
element in an array. Referenced by the low level language dependent
|
|
|
|
|
print routines. */
|
|
|
|
|
|
|
|
|
|
unsigned int repeat_count_threshold = 10;
|
|
|
|
|
|
|
|
|
|
/* If nonzero, stops printing of char arrays at first null. */
|
|
|
|
|
|
|
|
|
|
int stop_print_at_null;
|
|
|
|
|
|
|
|
|
|
/* Controls pretty printing of structures. */
|
|
|
|
|
|
|
|
|
|
int prettyprint_structs;
|
|
|
|
|
|
|
|
|
|
/* Controls pretty printing of arrays. */
|
|
|
|
|
|
|
|
|
|
int prettyprint_arrays;
|
|
|
|
|
|
|
|
|
|
/* If nonzero, causes unions inside structures or other unions to be
|
|
|
|
|
printed. */
|
|
|
|
|
|
|
|
|
|
int unionprint; /* Controls printing of nested unions. */
|
|
|
|
|
|
|
|
|
|
/* If nonzero, causes machine addresses to be printed in certain contexts. */
|
|
|
|
|
|
|
|
|
|
int addressprint; /* Controls printing of machine addresses */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Print data of type TYPE located at VALADDR (within GDB), which came from
|
|
|
|
|
the inferior at address ADDRESS, onto stdio stream STREAM according to
|
|
|
|
|
FORMAT (a letter, or 0 for natural format using TYPE).
|
|
|
|
|
|
|
|
|
|
If DEREF_REF is nonzero, then dereference references, otherwise just print
|
|
|
|
|
them like pointers.
|
|
|
|
|
|
|
|
|
|
The PRETTY parameter controls prettyprinting.
|
|
|
|
|
|
|
|
|
|
If the data are a string pointer, returns the number of string characters
|
|
|
|
|
printed.
|
|
|
|
|
|
|
|
|
|
FIXME: The data at VALADDR is in target byte order. If gdb is ever
|
|
|
|
|
enhanced to be able to debug more than the single target it was compiled
|
|
|
|
|
for (specific CPU type and thus specific target byte ordering), then
|
|
|
|
|
either the print routines are going to have to take this into account,
|
|
|
|
|
or the data is going to have to be passed into here already converted
|
|
|
|
|
to the host byte ordering, whichever is more convenient. */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
int
|
|
|
|
|
val_print (type, valaddr, address, stream, format, deref_ref, recurse, pretty)
|
|
|
|
|
struct type *type;
|
|
|
|
|
char *valaddr;
|
|
|
|
|
CORE_ADDR address;
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
int format;
|
|
|
|
|
int deref_ref;
|
|
|
|
|
int recurse;
|
|
|
|
|
enum val_prettyprint pretty;
|
|
|
|
|
{
|
|
|
|
|
struct type *real_type = check_typedef (type);
|
|
|
|
|
if (pretty == Val_pretty_default)
|
|
|
|
|
{
|
|
|
|
|
pretty = prettyprint_structs ? Val_prettyprint : Val_no_prettyprint;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
QUIT;
|
|
|
|
|
|
|
|
|
|
/* Ensure that the type is complete and not just a stub. If the type is
|
|
|
|
|
only a stub and we can't find and substitute its complete type, then
|
|
|
|
|
print appropriate string and return. */
|
|
|
|
|
|
|
|
|
|
if (TYPE_FLAGS (real_type) & TYPE_FLAG_STUB)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, "<incomplete type>");
|
|
|
|
|
gdb_flush (stream);
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (LA_VAL_PRINT (type, valaddr, address, stream, format, deref_ref,
|
|
|
|
|
recurse, pretty));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Print the value VAL in C-ish syntax on stream STREAM.
|
|
|
|
|
FORMAT is a format-letter, or 0 for print in natural format of data type.
|
|
|
|
|
If the object printed is a string pointer, returns
|
|
|
|
|
the number of string bytes printed. */
|
|
|
|
|
|
|
|
|
|
int
|
|
|
|
|
value_print (val, stream, format, pretty)
|
|
|
|
|
value_ptr val;
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
int format;
|
|
|
|
|
enum val_prettyprint pretty;
|
|
|
|
|
{
|
|
|
|
|
if (val == 0)
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("<address of value unknown>");
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
if (VALUE_OPTIMIZED_OUT (val))
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("<value optimized out>");
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
return LA_VALUE_PRINT (val, stream, format, pretty);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Called by various <lang>_val_print routines to print
|
|
|
|
|
TYPE_CODE_INT's. TYPE is the type. VALADDR is the address of the
|
|
|
|
|
value. STREAM is where to print the value. */
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
val_print_type_code_int (type, valaddr, stream)
|
|
|
|
|
struct type *type;
|
|
|
|
|
char *valaddr;
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
{
|
|
|
|
|
if (TYPE_LENGTH (type) > sizeof (LONGEST))
|
|
|
|
|
{
|
|
|
|
|
LONGEST val;
|
|
|
|
|
|
|
|
|
|
if (TYPE_UNSIGNED (type)
|
|
|
|
|
&& extract_long_unsigned_integer (valaddr, TYPE_LENGTH (type),
|
|
|
|
|
&val))
|
|
|
|
|
{
|
|
|
|
|
print_longest (stream, 'u', 0, val);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Signed, or we couldn't turn an unsigned value into a
|
|
|
|
|
LONGEST. For signed values, one could assume two's
|
|
|
|
|
complement (a reasonable assumption, I think) and do
|
|
|
|
|
better than this. */
|
|
|
|
|
print_hex_chars (stream, (unsigned char *) valaddr,
|
|
|
|
|
TYPE_LENGTH (type));
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
#ifdef PRINT_TYPELESS_INTEGER
|
|
|
|
|
PRINT_TYPELESS_INTEGER (stream, type, unpack_long (type, valaddr));
|
|
|
|
|
#else
|
|
|
|
|
print_longest (stream, TYPE_UNSIGNED (type) ? 'u' : 'd', 0,
|
|
|
|
|
unpack_long (type, valaddr));
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Print a number according to FORMAT which is one of d,u,x,o,b,h,w,g.
|
1999-02-11 01:06:10 +03:00
|
|
|
|
The raison d'etre of this function is to consolidate printing of
|
|
|
|
|
LONG_LONG's into this one function. Some platforms have long longs but
|
|
|
|
|
don't have a printf() that supports "ll" in the format string. We handle
|
|
|
|
|
these by seeing if the number is representable as either a signed or
|
|
|
|
|
unsigned long, depending upon what format is desired, and if not we just
|
|
|
|
|
bail out and print the number in hex.
|
|
|
|
|
|
|
|
|
|
The format chars b,h,w,g are from print_scalar_formatted(). If USE_LOCAL,
|
|
|
|
|
format it according to the current language (this should be used for most
|
|
|
|
|
integers which GDB prints, the exception is things like protocols where
|
|
|
|
|
the format of the integer is a protocol thing, not a user-visible thing).
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#if defined (CC_HAS_LONG_LONG) && !defined (PRINTF_HAS_LONG_LONG)
|
|
|
|
|
static void
|
|
|
|
|
print_decimal (stream, sign, use_local, val_ulong)
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
char *sign;
|
|
|
|
|
int use_local;
|
|
|
|
|
ULONGEST val_ulong;
|
|
|
|
|
{
|
|
|
|
|
unsigned long temp[3];
|
|
|
|
|
int i = 0;
|
|
|
|
|
do
|
|
|
|
|
{
|
|
|
|
|
temp[i] = val_ulong % (1000 * 1000 * 1000);
|
|
|
|
|
val_ulong /= (1000 * 1000 * 1000);
|
|
|
|
|
i++;
|
|
|
|
|
}
|
|
|
|
|
while (val_ulong != 0 && i < (sizeof (temp) / sizeof (temp[0])));
|
|
|
|
|
switch (i)
|
|
|
|
|
{
|
|
|
|
|
case 1:
|
|
|
|
|
fprintf_filtered (stream, "%s%lu",
|
|
|
|
|
sign, temp[0]);
|
|
|
|
|
break;
|
|
|
|
|
case 2:
|
|
|
|
|
fprintf_filtered (stream, "%s%lu%09lu",
|
|
|
|
|
sign, temp[1], temp[0]);
|
|
|
|
|
break;
|
|
|
|
|
case 3:
|
|
|
|
|
fprintf_filtered (stream, "%s%lu%09lu%09lu",
|
|
|
|
|
sign, temp[2], temp[1], temp[0]);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
abort ();
|
|
|
|
|
}
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
#endif
|
1997-09-24 20:57:14 +04:00
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
print_longest (stream, format, use_local, val_long)
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
int format;
|
|
|
|
|
int use_local;
|
|
|
|
|
LONGEST val_long;
|
|
|
|
|
{
|
|
|
|
|
#if defined (CC_HAS_LONG_LONG) && !defined (PRINTF_HAS_LONG_LONG)
|
1999-02-11 01:06:10 +03:00
|
|
|
|
if (sizeof (long) < sizeof (LONGEST))
|
1997-09-24 20:57:14 +04:00
|
|
|
|
{
|
1999-02-11 01:06:10 +03:00
|
|
|
|
switch (format)
|
|
|
|
|
{
|
|
|
|
|
case 'd':
|
|
|
|
|
{
|
|
|
|
|
/* Print a signed value, that doesn't fit in a long */
|
|
|
|
|
if ((long) val_long != val_long)
|
|
|
|
|
{
|
|
|
|
|
if (val_long < 0)
|
|
|
|
|
print_decimal (stream, "-", use_local, -val_long);
|
|
|
|
|
else
|
|
|
|
|
print_decimal (stream, "", use_local, val_long);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
case 'u':
|
|
|
|
|
{
|
|
|
|
|
/* Print an unsigned value, that doesn't fit in a long */
|
|
|
|
|
if ((unsigned long) val_long != (ULONGEST) val_long)
|
|
|
|
|
{
|
|
|
|
|
print_decimal (stream, "", use_local, val_long);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
case 'x':
|
|
|
|
|
case 'o':
|
|
|
|
|
case 'b':
|
|
|
|
|
case 'h':
|
|
|
|
|
case 'w':
|
|
|
|
|
case 'g':
|
|
|
|
|
/* Print as unsigned value, must fit completely in unsigned long */
|
|
|
|
|
{
|
|
|
|
|
unsigned long temp = val_long;
|
|
|
|
|
if (temp != val_long)
|
|
|
|
|
{
|
|
|
|
|
/* Urk, can't represent value in long so print in hex.
|
|
|
|
|
Do shift in two operations so that if sizeof (long)
|
|
|
|
|
== sizeof (LONGEST) we can avoid warnings from
|
|
|
|
|
picky compilers about shifts >= the size of the
|
|
|
|
|
shiftee in bits */
|
|
|
|
|
unsigned long vbot = (unsigned long) val_long;
|
|
|
|
|
LONGEST temp = (val_long >> (sizeof (long) * HOST_CHAR_BIT - 1));
|
|
|
|
|
unsigned long vtop = temp >> 1;
|
|
|
|
|
fprintf_filtered (stream, "0x%lx%08lx", vtop, vbot);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
1997-09-24 20:57:14 +04:00
|
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
1999-02-11 01:06:10 +03:00
|
|
|
|
#if defined (CC_HAS_LONG_LONG) && defined (PRINTF_HAS_LONG_LONG)
|
1997-09-24 20:57:14 +04:00
|
|
|
|
switch (format)
|
|
|
|
|
{
|
|
|
|
|
case 'd':
|
|
|
|
|
fprintf_filtered (stream,
|
|
|
|
|
use_local ? local_decimal_format_custom ("ll")
|
|
|
|
|
: "%lld",
|
|
|
|
|
val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'u':
|
|
|
|
|
fprintf_filtered (stream, "%llu", val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'x':
|
|
|
|
|
fprintf_filtered (stream,
|
|
|
|
|
use_local ? local_hex_format_custom ("ll")
|
|
|
|
|
: "%llx",
|
|
|
|
|
val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'o':
|
|
|
|
|
fprintf_filtered (stream,
|
|
|
|
|
use_local ? local_octal_format_custom ("ll")
|
|
|
|
|
: "%llo",
|
|
|
|
|
val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'b':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("02ll"), val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'h':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("04ll"), val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'w':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("08ll"), val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'g':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("016ll"), val_long);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
abort ();
|
|
|
|
|
}
|
1999-02-11 01:06:10 +03:00
|
|
|
|
#else /* !CC_HAS_LONG_LONG || !PRINTF_HAS_LONG_LONG*/
|
1997-09-24 20:57:14 +04:00
|
|
|
|
/* In the following it is important to coerce (val_long) to a long. It does
|
|
|
|
|
nothing if !LONG_LONG, but it will chop off the top half (which we know
|
|
|
|
|
we can ignore) if the host supports long longs. */
|
|
|
|
|
|
|
|
|
|
switch (format)
|
|
|
|
|
{
|
|
|
|
|
case 'd':
|
|
|
|
|
fprintf_filtered (stream,
|
|
|
|
|
use_local ? local_decimal_format_custom ("l")
|
|
|
|
|
: "%ld",
|
|
|
|
|
(long) val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'u':
|
|
|
|
|
fprintf_filtered (stream, "%lu", (unsigned long) val_long);
|
|
|
|
|
break;
|
|
|
|
|
case 'x':
|
|
|
|
|
fprintf_filtered (stream,
|
|
|
|
|
use_local ? local_hex_format_custom ("l")
|
|
|
|
|
: "%lx",
|
1999-02-11 01:06:10 +03:00
|
|
|
|
(unsigned long) val_long);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
break;
|
|
|
|
|
case 'o':
|
|
|
|
|
fprintf_filtered (stream,
|
|
|
|
|
use_local ? local_octal_format_custom ("l")
|
|
|
|
|
: "%lo",
|
1999-02-11 01:06:10 +03:00
|
|
|
|
(unsigned long) val_long);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
break;
|
|
|
|
|
case 'b':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("02l"),
|
1999-02-11 01:06:10 +03:00
|
|
|
|
(unsigned long) val_long);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
break;
|
|
|
|
|
case 'h':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("04l"),
|
1999-02-11 01:06:10 +03:00
|
|
|
|
(unsigned long) val_long);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
break;
|
|
|
|
|
case 'w':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("08l"),
|
1999-02-11 01:06:10 +03:00
|
|
|
|
(unsigned long) val_long);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
break;
|
|
|
|
|
case 'g':
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_custom ("016l"),
|
1999-02-11 01:06:10 +03:00
|
|
|
|
(unsigned long) val_long);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
abort ();
|
|
|
|
|
}
|
1999-02-11 01:06:10 +03:00
|
|
|
|
#endif /* CC_HAS_LONG_LONG || PRINTF_HAS_LONG_LONG */
|
1997-09-24 20:57:14 +04:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* This used to be a macro, but I don't think it is called often enough
|
|
|
|
|
to merit such treatment. */
|
|
|
|
|
/* Convert a LONGEST to an int. This is used in contexts (e.g. number of
|
|
|
|
|
arguments to a function, number in a value history, register number, etc.)
|
|
|
|
|
where the value must not be larger than can fit in an int. */
|
|
|
|
|
|
|
|
|
|
int
|
|
|
|
|
longest_to_int (arg)
|
|
|
|
|
LONGEST arg;
|
|
|
|
|
{
|
1999-02-11 01:06:10 +03:00
|
|
|
|
/* Let the compiler do the work */
|
|
|
|
|
int rtnval = (int) arg;
|
1997-09-24 20:57:14 +04:00
|
|
|
|
|
1999-02-11 01:06:10 +03:00
|
|
|
|
/* Check for overflows or underflows */
|
|
|
|
|
if (sizeof (LONGEST) > sizeof (int))
|
|
|
|
|
{
|
|
|
|
|
if (rtnval != arg)
|
|
|
|
|
{
|
|
|
|
|
error ("Value out of range.");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return (rtnval);
|
1997-09-24 20:57:14 +04:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Print a floating point value of type TYPE, pointed to in GDB by VALADDR,
|
|
|
|
|
on STREAM. */
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
print_floating (valaddr, type, stream)
|
|
|
|
|
char *valaddr;
|
|
|
|
|
struct type *type;
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
{
|
|
|
|
|
DOUBLEST doub;
|
|
|
|
|
int inv;
|
|
|
|
|
unsigned len = TYPE_LENGTH (type);
|
|
|
|
|
|
|
|
|
|
#if defined (IEEE_FLOAT)
|
|
|
|
|
|
|
|
|
|
/* Check for NaN's. Note that this code does not depend on us being
|
|
|
|
|
on an IEEE conforming system. It only depends on the target
|
|
|
|
|
machine using IEEE representation. This means (a)
|
|
|
|
|
cross-debugging works right, and (2) IEEE_FLOAT can (and should)
|
|
|
|
|
be defined for systems like the 68881, which uses IEEE
|
|
|
|
|
representation, but is not IEEE conforming. */
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
unsigned long low, high;
|
|
|
|
|
/* Is the sign bit 0? */
|
|
|
|
|
int nonnegative;
|
|
|
|
|
/* Is it is a NaN (i.e. the exponent is all ones and
|
|
|
|
|
the fraction is nonzero)? */
|
|
|
|
|
int is_nan;
|
|
|
|
|
|
|
|
|
|
if (len == 4)
|
|
|
|
|
{
|
|
|
|
|
/* It's single precision. */
|
|
|
|
|
/* Assume that floating point byte order is the same as
|
|
|
|
|
integer byte order. */
|
|
|
|
|
low = extract_unsigned_integer (valaddr, 4);
|
|
|
|
|
nonnegative = ((low & 0x80000000) == 0);
|
|
|
|
|
is_nan = ((((low >> 23) & 0xFF) == 0xFF)
|
|
|
|
|
&& 0 != (low & 0x7FFFFF));
|
|
|
|
|
low &= 0x7fffff;
|
|
|
|
|
high = 0;
|
|
|
|
|
}
|
|
|
|
|
else if (len == 8)
|
|
|
|
|
{
|
|
|
|
|
/* It's double precision. Get the high and low words. */
|
|
|
|
|
|
|
|
|
|
/* Assume that floating point byte order is the same as
|
|
|
|
|
integer byte order. */
|
|
|
|
|
if (TARGET_BYTE_ORDER == BIG_ENDIAN)
|
|
|
|
|
{
|
|
|
|
|
low = extract_unsigned_integer (valaddr + 4, 4);
|
|
|
|
|
high = extract_unsigned_integer (valaddr, 4);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
low = extract_unsigned_integer (valaddr, 4);
|
|
|
|
|
high = extract_unsigned_integer (valaddr + 4, 4);
|
|
|
|
|
}
|
|
|
|
|
nonnegative = ((high & 0x80000000) == 0);
|
|
|
|
|
is_nan = (((high >> 20) & 0x7ff) == 0x7ff
|
|
|
|
|
&& ! ((((high & 0xfffff) == 0)) && (low == 0)));
|
|
|
|
|
high &= 0xfffff;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
/* Extended. We can't detect NaNs for extendeds yet. Also note
|
|
|
|
|
that currently extendeds get nuked to double in
|
|
|
|
|
REGISTER_CONVERTIBLE. */
|
|
|
|
|
is_nan = 0;
|
|
|
|
|
|
|
|
|
|
if (is_nan)
|
|
|
|
|
{
|
|
|
|
|
/* The meaning of the sign and fraction is not defined by IEEE.
|
|
|
|
|
But the user might know what they mean. For example, they
|
|
|
|
|
(in an implementation-defined manner) distinguish between
|
|
|
|
|
signaling and quiet NaN's. */
|
|
|
|
|
if (high)
|
|
|
|
|
fprintf_filtered (stream, "-NaN(0x%lx%.8lx)" + nonnegative,
|
|
|
|
|
high, low);
|
|
|
|
|
else
|
|
|
|
|
fprintf_filtered (stream, "-NaN(0x%lx)" + nonnegative, low);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
#endif /* IEEE_FLOAT. */
|
|
|
|
|
|
|
|
|
|
doub = unpack_double (type, valaddr, &inv);
|
|
|
|
|
if (inv)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, "<invalid float value>");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (len < sizeof (double))
|
|
|
|
|
fprintf_filtered (stream, "%.9g", (double) doub);
|
|
|
|
|
else if (len == sizeof (double))
|
|
|
|
|
fprintf_filtered (stream, "%.17g", (double) doub);
|
|
|
|
|
else
|
|
|
|
|
#ifdef PRINTF_HAS_LONG_DOUBLE
|
|
|
|
|
fprintf_filtered (stream, "%.35Lg", doub);
|
|
|
|
|
#else
|
|
|
|
|
/* This at least wins with values that are representable as doubles */
|
|
|
|
|
fprintf_filtered (stream, "%.17g", (double) doub);
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* VALADDR points to an integer of LEN bytes. Print it in hex on stream. */
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
print_hex_chars (stream, valaddr, len)
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
unsigned char *valaddr;
|
|
|
|
|
unsigned len;
|
|
|
|
|
{
|
|
|
|
|
unsigned char *p;
|
|
|
|
|
|
|
|
|
|
/* FIXME: We should be not printing leading zeroes in most cases. */
|
|
|
|
|
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_prefix ());
|
|
|
|
|
if (TARGET_BYTE_ORDER == BIG_ENDIAN)
|
|
|
|
|
{
|
|
|
|
|
for (p = valaddr;
|
|
|
|
|
p < valaddr + len;
|
|
|
|
|
p++)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, "%02x", *p);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
for (p = valaddr + len - 1;
|
|
|
|
|
p >= valaddr;
|
|
|
|
|
p--)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, "%02x", *p);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
fprintf_filtered (stream, local_hex_format_suffix ());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Called by various <lang>_val_print routines to print elements of an
|
|
|
|
|
array in the form "<elem1>, <elem2>, <elem3>, ...".
|
|
|
|
|
|
|
|
|
|
(FIXME?) Assumes array element separator is a comma, which is correct
|
|
|
|
|
for all languages currently handled.
|
|
|
|
|
(FIXME?) Some languages have a notation for repeated array elements,
|
|
|
|
|
perhaps we should try to use that notation when appropriate.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
val_print_array_elements (type, valaddr, address, stream, format, deref_ref,
|
|
|
|
|
recurse, pretty, i)
|
|
|
|
|
struct type *type;
|
|
|
|
|
char *valaddr;
|
|
|
|
|
CORE_ADDR address;
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
int format;
|
|
|
|
|
int deref_ref;
|
|
|
|
|
int recurse;
|
|
|
|
|
enum val_prettyprint pretty;
|
|
|
|
|
unsigned int i;
|
|
|
|
|
{
|
|
|
|
|
unsigned int things_printed = 0;
|
|
|
|
|
unsigned len;
|
|
|
|
|
struct type *elttype;
|
|
|
|
|
unsigned eltlen;
|
|
|
|
|
/* Position of the array element we are examining to see
|
|
|
|
|
whether it is repeated. */
|
|
|
|
|
unsigned int rep1;
|
|
|
|
|
/* Number of repetitions we have detected so far. */
|
|
|
|
|
unsigned int reps;
|
|
|
|
|
|
|
|
|
|
elttype = TYPE_TARGET_TYPE (type);
|
|
|
|
|
eltlen = TYPE_LENGTH (check_typedef (elttype));
|
|
|
|
|
len = TYPE_LENGTH (type) / eltlen;
|
|
|
|
|
|
|
|
|
|
annotate_array_section_begin (i, elttype);
|
|
|
|
|
|
|
|
|
|
for (; i < len && things_printed < print_max; i++)
|
|
|
|
|
{
|
|
|
|
|
if (i != 0)
|
|
|
|
|
{
|
|
|
|
|
if (prettyprint_arrays)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, ",\n");
|
|
|
|
|
print_spaces_filtered (2 + 2 * recurse, stream);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, ", ");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
wrap_here (n_spaces (2 + 2 * recurse));
|
|
|
|
|
|
|
|
|
|
rep1 = i + 1;
|
|
|
|
|
reps = 1;
|
|
|
|
|
while ((rep1 < len) &&
|
|
|
|
|
!memcmp (valaddr + i * eltlen, valaddr + rep1 * eltlen, eltlen))
|
|
|
|
|
{
|
|
|
|
|
++reps;
|
|
|
|
|
++rep1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (reps > repeat_count_threshold)
|
|
|
|
|
{
|
|
|
|
|
val_print (elttype, valaddr + i * eltlen, 0, stream, format,
|
|
|
|
|
deref_ref, recurse + 1, pretty);
|
|
|
|
|
annotate_elt_rep (reps);
|
|
|
|
|
fprintf_filtered (stream, " <repeats %u times>", reps);
|
|
|
|
|
annotate_elt_rep_end ();
|
|
|
|
|
|
|
|
|
|
i = rep1 - 1;
|
|
|
|
|
things_printed += repeat_count_threshold;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
val_print (elttype, valaddr + i * eltlen, 0, stream, format,
|
|
|
|
|
deref_ref, recurse + 1, pretty);
|
|
|
|
|
annotate_elt ();
|
|
|
|
|
things_printed++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
annotate_array_section_end ();
|
|
|
|
|
if (i < len)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, "...");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Print a string from the inferior, starting at ADDR and printing up to LEN
|
|
|
|
|
characters, to STREAM. If LEN is zero, printing stops at the first null
|
|
|
|
|
byte, otherwise printing proceeds (including null bytes) until either
|
|
|
|
|
print_max or LEN characters have been printed, whichever is smaller. */
|
|
|
|
|
|
|
|
|
|
/* FIXME: All callers supply LEN of zero. Supplying a non-zero LEN is
|
|
|
|
|
pointless, this routine just then becomes a convoluted version of
|
|
|
|
|
target_read_memory_partial. Removing all the LEN stuff would simplify
|
|
|
|
|
this routine enormously.
|
|
|
|
|
|
|
|
|
|
FIXME: Use target_read_string. */
|
|
|
|
|
|
|
|
|
|
int
|
|
|
|
|
val_print_string (addr, len, stream)
|
|
|
|
|
CORE_ADDR addr;
|
|
|
|
|
unsigned int len;
|
|
|
|
|
GDB_FILE *stream;
|
|
|
|
|
{
|
|
|
|
|
int force_ellipsis = 0; /* Force ellipsis to be printed if nonzero. */
|
|
|
|
|
int errcode; /* Errno returned from bad reads. */
|
|
|
|
|
unsigned int fetchlimit; /* Maximum number of bytes to fetch. */
|
|
|
|
|
unsigned int nfetch; /* Bytes to fetch / bytes fetched. */
|
|
|
|
|
unsigned int chunksize; /* Size of each fetch, in bytes. */
|
|
|
|
|
unsigned int bufsize; /* Size of current fetch buffer. */
|
|
|
|
|
char *buffer = NULL; /* Dynamically growable fetch buffer. */
|
|
|
|
|
char *bufptr; /* Pointer to next available byte in buffer. */
|
|
|
|
|
char *limit; /* First location past end of fetch buffer. */
|
|
|
|
|
struct cleanup *old_chain = NULL; /* Top of the old cleanup chain. */
|
|
|
|
|
char peekchar; /* Place into which we can read one char. */
|
|
|
|
|
|
|
|
|
|
/* First we need to figure out the limit on the number of characters we are
|
|
|
|
|
going to attempt to fetch and print. This is actually pretty simple. If
|
|
|
|
|
LEN is nonzero, then the limit is the minimum of LEN and print_max. If
|
|
|
|
|
LEN is zero, then the limit is print_max. This is true regardless of
|
|
|
|
|
whether print_max is zero, UINT_MAX (unlimited), or something in between,
|
|
|
|
|
because finding the null byte (or available memory) is what actually
|
|
|
|
|
limits the fetch. */
|
|
|
|
|
|
|
|
|
|
fetchlimit = (len == 0 ? print_max : min (len, print_max));
|
|
|
|
|
|
|
|
|
|
/* Now decide how large of chunks to try to read in one operation. This
|
|
|
|
|
is also pretty simple. If LEN is nonzero, then we want fetchlimit bytes,
|
|
|
|
|
so we might as well read them all in one operation. If LEN is zero, we
|
|
|
|
|
are looking for a null terminator to end the fetching, so we might as
|
|
|
|
|
well read in blocks that are large enough to be efficient, but not so
|
|
|
|
|
large as to be slow if fetchlimit happens to be large. So we choose the
|
|
|
|
|
minimum of 8 and fetchlimit. We used to use 200 instead of 8 but
|
|
|
|
|
200 is way too big for remote debugging over a serial line. */
|
|
|
|
|
|
|
|
|
|
chunksize = (len == 0 ? min (8, fetchlimit) : fetchlimit);
|
|
|
|
|
|
|
|
|
|
/* Loop until we either have all the characters to print, or we encounter
|
|
|
|
|
some error, such as bumping into the end of the address space. */
|
|
|
|
|
|
|
|
|
|
bufsize = 0;
|
|
|
|
|
do {
|
|
|
|
|
QUIT;
|
|
|
|
|
/* Figure out how much to fetch this time, and grow the buffer to fit. */
|
|
|
|
|
nfetch = min (chunksize, fetchlimit - bufsize);
|
|
|
|
|
bufsize += nfetch;
|
|
|
|
|
if (buffer == NULL)
|
|
|
|
|
{
|
|
|
|
|
buffer = (char *) xmalloc (bufsize);
|
|
|
|
|
bufptr = buffer;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
discard_cleanups (old_chain);
|
|
|
|
|
buffer = (char *) xrealloc (buffer, bufsize);
|
|
|
|
|
bufptr = buffer + bufsize - nfetch;
|
|
|
|
|
}
|
|
|
|
|
old_chain = make_cleanup (free, buffer);
|
|
|
|
|
|
|
|
|
|
/* Read as much as we can. */
|
|
|
|
|
nfetch = target_read_memory_partial (addr, bufptr, nfetch, &errcode);
|
|
|
|
|
if (len != 0)
|
|
|
|
|
{
|
|
|
|
|
addr += nfetch;
|
|
|
|
|
bufptr += nfetch;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Scan this chunk for the null byte that terminates the string
|
|
|
|
|
to print. If found, we don't need to fetch any more. Note
|
|
|
|
|
that bufptr is explicitly left pointing at the next character
|
|
|
|
|
after the null byte, or at the next character after the end of
|
|
|
|
|
the buffer. */
|
|
|
|
|
limit = bufptr + nfetch;
|
|
|
|
|
while (bufptr < limit)
|
|
|
|
|
{
|
|
|
|
|
++addr;
|
|
|
|
|
++bufptr;
|
|
|
|
|
if (bufptr[-1] == '\0')
|
|
|
|
|
{
|
|
|
|
|
/* We don't care about any error which happened after
|
|
|
|
|
the NULL terminator. */
|
|
|
|
|
errcode = 0;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
} while (errcode == 0 /* no error */
|
|
|
|
|
&& bufsize < fetchlimit /* no overrun */
|
|
|
|
|
&& !(len == 0 && *(bufptr - 1) == '\0')); /* no null term */
|
|
|
|
|
|
|
|
|
|
/* bufptr and addr now point immediately beyond the last byte which we
|
|
|
|
|
consider part of the string (including a '\0' which ends the string). */
|
|
|
|
|
|
|
|
|
|
/* We now have either successfully filled the buffer to fetchlimit, or
|
|
|
|
|
terminated early due to an error or finding a null byte when LEN is
|
|
|
|
|
zero. */
|
|
|
|
|
|
|
|
|
|
if (len == 0 && bufptr > buffer && *(bufptr - 1) != '\0')
|
|
|
|
|
{
|
|
|
|
|
/* We didn't find a null terminator we were looking for. Attempt
|
|
|
|
|
to peek at the next character. If not successful, or it is not
|
|
|
|
|
a null byte, then force ellipsis to be printed. */
|
|
|
|
|
if (target_read_memory (addr, &peekchar, 1) != 0 || peekchar != '\0')
|
|
|
|
|
{
|
|
|
|
|
force_ellipsis = 1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
else if ((len != 0 && errcode != 0) || (len > bufptr - buffer))
|
|
|
|
|
{
|
|
|
|
|
/* Getting an error when we have a requested length, or fetching less
|
|
|
|
|
than the number of characters actually requested, always make us
|
|
|
|
|
print ellipsis. */
|
|
|
|
|
force_ellipsis = 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
QUIT;
|
|
|
|
|
|
|
|
|
|
/* If we get an error before fetching anything, don't print a string.
|
|
|
|
|
But if we fetch something and then get an error, print the string
|
|
|
|
|
and then the error message. */
|
|
|
|
|
if (errcode == 0 || bufptr > buffer)
|
|
|
|
|
{
|
|
|
|
|
if (addressprint)
|
|
|
|
|
{
|
|
|
|
|
fputs_filtered (" ", stream);
|
|
|
|
|
}
|
|
|
|
|
LA_PRINT_STRING (stream, buffer, bufptr - buffer, force_ellipsis);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (errcode != 0)
|
|
|
|
|
{
|
|
|
|
|
if (errcode == EIO)
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, " <Address ");
|
|
|
|
|
print_address_numeric (addr, 1, stream);
|
|
|
|
|
fprintf_filtered (stream, " out of bounds>");
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
fprintf_filtered (stream, " <Error reading address ");
|
|
|
|
|
print_address_numeric (addr, 1, stream);
|
|
|
|
|
fprintf_filtered (stream, ": %s>", safe_strerror (errcode));
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
gdb_flush (stream);
|
|
|
|
|
do_cleanups (old_chain);
|
|
|
|
|
return (bufptr - buffer);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Validate an input or output radix setting, and make sure the user
|
|
|
|
|
knows what they really did here. Radix setting is confusing, e.g.
|
|
|
|
|
setting the input radix to "10" never changes it! */
|
|
|
|
|
|
|
|
|
|
/* ARGSUSED */
|
|
|
|
|
static void
|
|
|
|
|
set_input_radix (args, from_tty, c)
|
|
|
|
|
char *args;
|
|
|
|
|
int from_tty;
|
|
|
|
|
struct cmd_list_element *c;
|
|
|
|
|
{
|
|
|
|
|
set_input_radix_1 (from_tty, *(unsigned *)c->var);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ARGSUSED */
|
|
|
|
|
static void
|
|
|
|
|
set_input_radix_1 (from_tty, radix)
|
|
|
|
|
int from_tty;
|
|
|
|
|
unsigned radix;
|
|
|
|
|
{
|
|
|
|
|
/* We don't currently disallow any input radix except 0 or 1, which don't
|
|
|
|
|
make any mathematical sense. In theory, we can deal with any input
|
|
|
|
|
radix greater than 1, even if we don't have unique digits for every
|
|
|
|
|
value from 0 to radix-1, but in practice we lose on large radix values.
|
|
|
|
|
We should either fix the lossage or restrict the radix range more.
|
|
|
|
|
(FIXME). */
|
|
|
|
|
|
|
|
|
|
if (radix < 2)
|
|
|
|
|
{
|
|
|
|
|
error ("Nonsense input radix ``decimal %u''; input radix unchanged.",
|
|
|
|
|
radix);
|
|
|
|
|
}
|
|
|
|
|
input_radix = radix;
|
|
|
|
|
if (from_tty)
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("Input radix now set to decimal %u, hex %x, octal %o.\n",
|
|
|
|
|
radix, radix, radix);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ARGSUSED */
|
|
|
|
|
static void
|
|
|
|
|
set_output_radix (args, from_tty, c)
|
|
|
|
|
char *args;
|
|
|
|
|
int from_tty;
|
|
|
|
|
struct cmd_list_element *c;
|
|
|
|
|
{
|
|
|
|
|
set_output_radix_1 (from_tty, *(unsigned *)c->var);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_output_radix_1 (from_tty, radix)
|
|
|
|
|
int from_tty;
|
|
|
|
|
unsigned radix;
|
|
|
|
|
{
|
|
|
|
|
/* Validate the radix and disallow ones that we aren't prepared to
|
|
|
|
|
handle correctly, leaving the radix unchanged. */
|
|
|
|
|
switch (radix)
|
|
|
|
|
{
|
|
|
|
|
case 16:
|
|
|
|
|
output_format = 'x'; /* hex */
|
|
|
|
|
break;
|
|
|
|
|
case 10:
|
|
|
|
|
output_format = 0; /* decimal */
|
|
|
|
|
break;
|
|
|
|
|
case 8:
|
|
|
|
|
output_format = 'o'; /* octal */
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
error ("Unsupported output radix ``decimal %u''; output radix unchanged.",
|
|
|
|
|
radix);
|
|
|
|
|
}
|
|
|
|
|
output_radix = radix;
|
|
|
|
|
if (from_tty)
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("Output radix now set to decimal %u, hex %x, octal %o.\n",
|
|
|
|
|
radix, radix, radix);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Set both the input and output radix at once. Try to set the output radix
|
|
|
|
|
first, since it has the most restrictive range. An radix that is valid as
|
|
|
|
|
an output radix is also valid as an input radix.
|
|
|
|
|
|
|
|
|
|
It may be useful to have an unusual input radix. If the user wishes to
|
|
|
|
|
set an input radix that is not valid as an output radix, he needs to use
|
|
|
|
|
the 'set input-radix' command. */
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
set_radix (arg, from_tty)
|
|
|
|
|
char *arg;
|
|
|
|
|
int from_tty;
|
|
|
|
|
{
|
|
|
|
|
unsigned radix;
|
|
|
|
|
|
|
|
|
|
radix = (arg == NULL) ? 10 : parse_and_eval_address (arg);
|
|
|
|
|
set_output_radix_1 (0, radix);
|
|
|
|
|
set_input_radix_1 (0, radix);
|
|
|
|
|
if (from_tty)
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("Input and output radices now set to decimal %u, hex %x, octal %o.\n",
|
|
|
|
|
radix, radix, radix);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Show both the input and output radices. */
|
|
|
|
|
|
|
|
|
|
/*ARGSUSED*/
|
|
|
|
|
static void
|
|
|
|
|
show_radix (arg, from_tty)
|
|
|
|
|
char *arg;
|
|
|
|
|
int from_tty;
|
|
|
|
|
{
|
|
|
|
|
if (from_tty)
|
|
|
|
|
{
|
|
|
|
|
if (input_radix == output_radix)
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("Input and output radices set to decimal %u, hex %x, octal %o.\n",
|
|
|
|
|
input_radix, input_radix, input_radix);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
printf_filtered ("Input radix set to decimal %u, hex %x, octal %o.\n",
|
|
|
|
|
input_radix, input_radix, input_radix);
|
|
|
|
|
printf_filtered ("Output radix set to decimal %u, hex %x, octal %o.\n",
|
|
|
|
|
output_radix, output_radix, output_radix);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*ARGSUSED*/
|
|
|
|
|
static void
|
|
|
|
|
set_print (arg, from_tty)
|
|
|
|
|
char *arg;
|
|
|
|
|
int from_tty;
|
|
|
|
|
{
|
|
|
|
|
printf_unfiltered (
|
|
|
|
|
"\"set print\" must be followed by the name of a print subcommand.\n");
|
|
|
|
|
help_list (setprintlist, "set print ", -1, gdb_stdout);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*ARGSUSED*/
|
|
|
|
|
static void
|
|
|
|
|
show_print (args, from_tty)
|
|
|
|
|
char *args;
|
|
|
|
|
int from_tty;
|
|
|
|
|
{
|
|
|
|
|
cmd_show_list (showprintlist, from_tty, "");
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
_initialize_valprint ()
|
|
|
|
|
{
|
|
|
|
|
struct cmd_list_element *c;
|
|
|
|
|
|
|
|
|
|
add_prefix_cmd ("print", no_class, set_print,
|
|
|
|
|
"Generic command for setting how things print.",
|
|
|
|
|
&setprintlist, "set print ", 0, &setlist);
|
|
|
|
|
add_alias_cmd ("p", "print", no_class, 1, &setlist);
|
|
|
|
|
/* prefer set print to set prompt */
|
|
|
|
|
add_alias_cmd ("pr", "print", no_class, 1, &setlist);
|
|
|
|
|
|
|
|
|
|
add_prefix_cmd ("print", no_class, show_print,
|
|
|
|
|
"Generic command for showing print settings.",
|
|
|
|
|
&showprintlist, "show print ", 0, &showlist);
|
|
|
|
|
add_alias_cmd ("p", "print", no_class, 1, &showlist);
|
|
|
|
|
add_alias_cmd ("pr", "print", no_class, 1, &showlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("elements", no_class, var_uinteger, (char *)&print_max,
|
|
|
|
|
"Set limit on string chars or array elements to print.\n\
|
|
|
|
|
\"set print elements 0\" causes there to be no limit.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("null-stop", no_class, var_boolean,
|
|
|
|
|
(char *)&stop_print_at_null,
|
|
|
|
|
"Set printing of char arrays to stop at first null char.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("repeats", no_class, var_uinteger,
|
|
|
|
|
(char *)&repeat_count_threshold,
|
|
|
|
|
"Set threshold for repeated print elements.\n\
|
|
|
|
|
\"set print repeats 0\" causes all elements to be individually printed.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("pretty", class_support, var_boolean,
|
|
|
|
|
(char *)&prettyprint_structs,
|
|
|
|
|
"Set prettyprinting of structures.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("union", class_support, var_boolean, (char *)&unionprint,
|
|
|
|
|
"Set printing of unions interior to structures.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("array", class_support, var_boolean,
|
|
|
|
|
(char *)&prettyprint_arrays,
|
|
|
|
|
"Set prettyprinting of arrays.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
add_show_from_set
|
|
|
|
|
(add_set_cmd ("address", class_support, var_boolean, (char *)&addressprint,
|
|
|
|
|
"Set printing of addresses.",
|
|
|
|
|
&setprintlist),
|
|
|
|
|
&showprintlist);
|
|
|
|
|
|
|
|
|
|
c = add_set_cmd ("input-radix", class_support, var_uinteger,
|
|
|
|
|
(char *)&input_radix,
|
|
|
|
|
"Set default input radix for entering numbers.",
|
|
|
|
|
&setlist);
|
|
|
|
|
add_show_from_set (c, &showlist);
|
|
|
|
|
c->function.sfunc = set_input_radix;
|
|
|
|
|
|
|
|
|
|
c = add_set_cmd ("output-radix", class_support, var_uinteger,
|
|
|
|
|
(char *)&output_radix,
|
|
|
|
|
"Set default output radix for printing of values.",
|
|
|
|
|
&setlist);
|
|
|
|
|
add_show_from_set (c, &showlist);
|
|
|
|
|
c->function.sfunc = set_output_radix;
|
|
|
|
|
|
|
|
|
|
/* The "set radix" and "show radix" commands are special in that they are
|
|
|
|
|
like normal set and show commands but allow two normally independent
|
|
|
|
|
variables to be either set or shown with a single command. So the
|
|
|
|
|
usual add_set_cmd() and add_show_from_set() commands aren't really
|
|
|
|
|
appropriate. */
|
|
|
|
|
add_cmd ("radix", class_support, set_radix,
|
|
|
|
|
"Set default input and output number radices.\n\
|
|
|
|
|
Use 'set input-radix' or 'set output-radix' to independently set each.\n\
|
|
|
|
|
Without an argument, sets both radices back to the default value of 10.",
|
|
|
|
|
&setlist);
|
|
|
|
|
add_cmd ("radix", class_support, show_radix,
|
|
|
|
|
"Show the default input and output number radices.\n\
|
|
|
|
|
Use 'show input-radix' or 'show output-radix' to independently show each.",
|
|
|
|
|
&showlist);
|
|
|
|
|
|
|
|
|
|
/* Give people the defaults which they are used to. */
|
|
|
|
|
prettyprint_structs = 0;
|
|
|
|
|
prettyprint_arrays = 0;
|
|
|
|
|
unionprint = 1;
|
|
|
|
|
addressprint = 1;
|
|
|
|
|
print_max = PRINT_MAX_DEFAULT;
|
|
|
|
|
}
|