darling-gdb/gdb/m88k-tdep.c

857 lines
27 KiB
C
Raw Normal View History

/* Target-machine dependent code for Motorola 88000 series, for GDB.
Copyright (C) 1988, 1990, 1991 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., 675 Mass Ave, Cambridge, MA 02139, USA. */
#include "defs.h"
#include "frame.h"
#include "inferior.h"
#include "value.h"
#ifdef USG
#include <sys/types.h>
#endif
#include <sys/param.h>
#include <sys/dir.h>
#include <signal.h>
#include "gdbcore.h"
#include <sys/user.h>
#ifndef USER /* added to support BCS ptrace_user */
#define USER ptrace_user
#endif
#include <sys/ioctl.h>
#include <fcntl.h>
#include <sys/file.h>
#include <sys/stat.h>
#include "symtab.h"
#include "setjmp.h"
#include "value.h"
/* Size of an instruction */
#define BYTES_PER_88K_INSN 4
void frame_find_saved_regs ();
/* Given a GDB frame, determine the address of the calling function's frame.
This will be used to create a new GDB frame struct, and then
INIT_EXTRA_FRAME_INFO and INIT_FRAME_PC will be called for the new frame.
For us, the frame address is its stack pointer value, so we look up
the function prologue to determine the caller's sp value, and return it. */
FRAME_ADDR
frame_chain (thisframe)
FRAME thisframe;
{
frame_find_saved_regs (thisframe, (struct frame_saved_regs *) 0);
/* NOTE: this depends on frame_find_saved_regs returning the VALUE, not
the ADDRESS, of SP_REGNUM. It also depends on the cache of
frame_find_saved_regs results. */
if (thisframe->fsr->regs[SP_REGNUM])
return thisframe->fsr->regs[SP_REGNUM];
else
return thisframe->frame; /* Leaf fn -- next frame up has same SP. */
}
int
frameless_function_invocation (frame)
FRAME frame;
{
frame_find_saved_regs (frame, (struct frame_saved_regs *) 0);
/* NOTE: this depends on frame_find_saved_regs returning the VALUE, not
the ADDRESS, of SP_REGNUM. It also depends on the cache of
frame_find_saved_regs results. */
if (frame->fsr->regs[SP_REGNUM])
return 0; /* Frameful -- return addr saved somewhere */
else
return 1; /* Frameless -- no saved return address */
}
int
frame_chain_valid (chain, thisframe)
CORE_ADDR chain;
struct frame_info *thisframe;
{
return (chain != 0
&& !inside_entry_file (FRAME_SAVED_PC (thisframe)));
}
void
init_extra_frame_info (fromleaf, fi)
int fromleaf;
struct frame_info *fi;
{
fi->fsr = 0; /* Not yet allocated */
fi->args_pointer = 0; /* Unknown */
fi->locals_pointer = 0; /* Unknown */
}
/* Examine an m88k function prologue, recording the addresses at which
registers are saved explicitly by the prologue code, and returning
the address of the first instruction after the prologue (but not
after the instruction at address LIMIT, as explained below).
LIMIT places an upper bound on addresses of the instructions to be
examined. If the prologue code scan reaches LIMIT, the scan is
aborted and LIMIT is returned. This is used, when examining the
prologue for the current frame, to keep examine_prologue () from
claiming that a given register has been saved when in fact the
instruction that saves it has not yet been executed. LIMIT is used
at other times to stop the scan when we hit code after the true
function prologue (e.g. for the first source line) which might
otherwise be mistaken for function prologue.
The format of the function prologue matched by this routine is
derived from examination of the source to gcc 1.95, particularly
the routine output_prologue () in config/out-m88k.c.
subu r31,r31,n # stack pointer update
(st rn,r31,offset)? # save incoming regs
(st.d rn,r31,offset)?
(addu r30,r31,n)? # frame pointer update
(pic sequence)? # PIC code prologue
(or rn,rm,0)? # Move parameters to other regs
*/
/* Macros for extracting fields from instructions. */
#define BITMASK(pos, width) (((0x1 << (width)) - 1) << (pos))
#define EXTRACT_FIELD(val, pos, width) ((val) >> (pos) & BITMASK (0, width))
/* Prologue code that handles position-independent-code setup. */
struct pic_prologue_code {
unsigned long insn, mask;
};
static struct pic_prologue_code pic_prologue_code [] = {
/* FIXME -- until this is translated to hex, we won't match it... */
0xffffffff, 0,
/* or r10,r1,0 (if not saved) */
/* bsr.n LabN */
/* or.u r25,r0,const */
/*LabN: or r25,r25,const2 */
/* addu r25,r25,1 */
/* or r1,r10,0 (if not saved) */
};
/* Fetch the instruction at ADDR, returning 0 if ADDR is beyond LIM or
is not the address of a valid instruction, the address of the next
instruction beyond ADDR otherwise. *PWORD1 receives the first word
of the instruction. PWORD2 is ignored -- a remnant of the original
i960 version. */
#define NEXT_PROLOGUE_INSN(addr, lim, pword1, pword2) \
(((addr) < (lim)) ? next_insn (addr, pword1) : 0)
/* Read the m88k instruction at 'memaddr' and return the address of
the next instruction after that, or 0 if 'memaddr' is not the
address of a valid instruction. The instruction
is stored at 'pword1'. */
CORE_ADDR
next_insn (memaddr, pword1)
unsigned long *pword1;
CORE_ADDR memaddr;
{
unsigned long buf[1];
*pword1 = read_memory_integer (memaddr, BYTES_PER_88K_INSN);
return memaddr + BYTES_PER_88K_INSN;
}
/* Read a register from frames called by us (or from the hardware regs). */
int
read_next_frame_reg(fi, regno)
FRAME fi;
int regno;
{
for (; fi; fi = fi->next) {
if (regno == SP_REGNUM) return fi->frame;
else if (fi->fsr->regs[regno])
return read_memory_integer(fi->fsr->regs[regno], 4);
}
return read_register(regno);
}
/* Examine the prologue of a function. `ip' points to the first instruction.
`limit' is the limit of the prologue (e.g. the addr of the first
linenumber, or perhaps the program counter if we're stepping through).
`frame_sp' is the stack pointer value in use in this frame.
`fsr' is a pointer to a frame_saved_regs structure into which we put
info about the registers saved by this frame.
`fi' is a struct frame_info pointer; we fill in various fields in it
to reflect the offsets of the arg pointer and the locals pointer. */
static CORE_ADDR
examine_prologue (ip, limit, frame_sp, fsr, fi)
register CORE_ADDR ip;
register CORE_ADDR limit;
FRAME_ADDR frame_sp;
struct frame_saved_regs *fsr;
struct frame_info *fi;
{
register CORE_ADDR next_ip;
register int src;
register struct pic_prologue_code *pcode;
unsigned int insn1, insn2;
int size, offset;
char must_adjust[32]; /* If set, must adjust offsets in fsr */
int sp_offset = -1; /* -1 means not set (valid must be mult of 8) */
int fp_offset = -1; /* -1 means not set */
CORE_ADDR frame_fp;
bzero (must_adjust, sizeof (must_adjust));
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
/* Accept move of incoming registers to other registers, using
"or rd,rs,0" or "or.u rd,rs,0" or "or rd,r0,rs" or "or rd,rs,r0".
We don't have to worry about walking into the first lines of code,
since the first line number will stop us (assuming we have symbols).
What we have actually seen is "or r10,r0,r12". */
#define OR_MOVE_INSN 0x58000000 /* or/or.u with immed of 0 */
#define OR_MOVE_MASK 0xF800FFFF
#define OR_REG_MOVE1_INSN 0xF4005800 /* or rd,r0,rs */
#define OR_REG_MOVE1_MASK 0xFC1FFFE0
#define OR_REG_MOVE2_INSN 0xF4005800 /* or rd,rs,r0 */
#define OR_REG_MOVE2_MASK 0xFC00FFFF
while (next_ip &&
((insn1 & OR_MOVE_MASK) == OR_MOVE_INSN ||
(insn1 & OR_REG_MOVE1_MASK) == OR_REG_MOVE1_INSN ||
(insn1 & OR_REG_MOVE2_MASK) == OR_REG_MOVE2_INSN
)
)
{
/* We don't care what moves to where. The result of the moves
has already been reflected in what the compiler tells us is the
location of these parameters. */
ip = next_ip;
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
}
/* Accept an optional "subu sp,sp,n" to set up the stack pointer. */
#define SUBU_SP_INSN 0x67ff0000
#define SUBU_SP_MASK 0xffff0007 /* Note offset must be mult. of 8 */
#define SUBU_OFFSET(x) ((unsigned)(x & 0xFFFF))
if (next_ip &&
((insn1 & SUBU_SP_MASK) == SUBU_SP_INSN)) /* subu r31, r31, N */
{
sp_offset = -SUBU_OFFSET (insn1);
ip = next_ip;
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
}
/* The function must start with a stack-pointer adjustment, or
we don't know WHAT'S going on... */
if (sp_offset == -1)
return ip;
/* Accept zero or more instances of "st rx,sp,n" or "st.d rx,sp,n".
This may cause us to mistake the copying of a register
parameter to the frame for the saving of a callee-saved
register, but that can't be helped, since with the
"-fcall-saved" flag, any register can be made callee-saved.
This probably doesn't matter, since the ``saved'' caller's values of
non-callee-saved registers are not relevant anyway. */
#define STD_STACK_INSN 0x201f0000
#define STD_STACK_MASK 0xfc1f0000
#define ST_STACK_INSN 0x241f0000
#define ST_STACK_MASK 0xfc1f0000
#define ST_OFFSET(x) ((unsigned)((x) & 0xFFFF))
#define ST_SRC(x) EXTRACT_FIELD ((x), 21, 5)
while (next_ip)
{
if ((insn1 & ST_STACK_MASK) == ST_STACK_INSN)
size = 1;
else if ((insn1 & STD_STACK_MASK) == STD_STACK_INSN)
size = 2;
else
break;
src = ST_SRC (insn1);
offset = ST_OFFSET (insn1);
while (size--)
{
must_adjust[src] = 1;
fsr->regs[src++] = offset; /* Will be adjusted later */
offset += 4;
}
ip = next_ip;
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
}
/* Accept an optional "addu r30,r31,n" to set up the frame pointer. */
#define ADDU_FP_INSN 0x63df0000
#define ADDU_FP_MASK 0xffff0000
#define ADDU_OFFSET(x) ((unsigned)(x & 0xFFFF))
if (next_ip &&
((insn1 & ADDU_FP_MASK) == ADDU_FP_INSN)) /* addu r30, r31, N */
{
fp_offset = ADDU_OFFSET (insn1);
ip = next_ip;
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
}
/* Accept the PIC prologue code if present. */
pcode = pic_prologue_code;
size = sizeof (pic_prologue_code) / sizeof (*pic_prologue_code);
/* If return addr is saved, we don't use first or last insn of PICstuff. */
if (fsr->regs[SRP_REGNUM]) {
pcode++;
size-=2;
}
while (size-- && next_ip && (pcode->insn == (pcode->mask & insn1)))
{
pcode++;
ip = next_ip;
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
}
/* Accept moves of parameter registers to other registers, using
"or rd,rs,0" or "or.u rd,rs,0" or "or rd,r0,rs" or "or rd,rs,r0".
We don't have to worry about walking into the first lines of code,
since the first line number will stop us (assuming we have symbols).
What gcc actually seems to produce is "or rd,r0,rs". */
#define OR_MOVE_INSN 0x58000000 /* or/or.u with immed of 0 */
#define OR_MOVE_MASK 0xF800FFFF
#define OR_REG_MOVE1_INSN 0xF4005800 /* or rd,r0,rs */
#define OR_REG_MOVE1_MASK 0xFC1FFFE0
#define OR_REG_MOVE2_INSN 0xF4005800 /* or rd,rs,r0 */
#define OR_REG_MOVE2_MASK 0xFC00FFFF
while (next_ip &&
((insn1 & OR_MOVE_MASK) == OR_MOVE_INSN ||
(insn1 & OR_REG_MOVE1_MASK) == OR_REG_MOVE1_INSN ||
(insn1 & OR_REG_MOVE2_MASK) == OR_REG_MOVE2_INSN
)
)
{
/* We don't care what moves to where. The result of the moves
has already been reflected in what the compiler tells us is the
location of these parameters. */
ip = next_ip;
next_ip = NEXT_PROLOGUE_INSN (ip, limit, &insn1, &insn2);
}
/* We're done with the prologue. If we don't care about the stack
frame itself, just return. (Note that fsr->regs has been trashed,
but the one caller who calls with fi==0 passes a dummy there.) */
if (fi == 0)
return ip;
/* OK, now we have:
sp_offset original negative displacement of SP
fp_offset positive displacement between new SP and new FP, or -1
fsr->regs[0..31] offset from original SP where reg is stored
must_adjust[0..31] set if corresp. offset was set
The current SP (frame_sp) might not be the original new SP as set
by the function prologue, if alloca has been called. This can
only occur if fp_offset is set, though (the compiler allocates an
FP when it sees alloca). In that case, we have the FP,
and can calculate the original new SP from the FP.
Then, we figure out where the arguments and locals are, and
relocate the offsets in fsr->regs to absolute addresses. */
if (fp_offset != -1) {
/* We have a frame pointer, so get it, and base our calc's on it. */
frame_fp = (CORE_ADDR) read_next_frame_reg (fi->next, FP_REGNUM);
frame_sp = frame_fp - fp_offset;
} else {
/* We have no frame pointer, therefore frame_sp is still the same value
as set by prologue. But where is the frame itself? */
if (must_adjust[SRP_REGNUM]) {
/* Function header saved SRP (r1), the return address. Frame starts
4 bytes down from where it was saved. */
frame_fp = frame_sp + fsr->regs[SRP_REGNUM] - 4;
fi->locals_pointer = frame_fp;
} else {
/* Function header didn't save SRP (r1), so we are in a leaf fn or
are otherwise confused. */
frame_fp = -1;
}
}
/* The locals are relative to the FP (whether it exists as an allocated
register, or just as an assumed offset from the SP) */
fi->locals_pointer = frame_fp;
/* The arguments are just above the SP as it was before we adjusted it
on entry. */
fi->args_pointer = frame_sp - sp_offset;
/* Now that we know the SP value used by the prologue, we know where
it saved all the registers. */
for (src = 0; src < 32; src++)
if (must_adjust[src])
fsr->regs[src] += frame_sp;
/* The saved value of the SP is always known. */
/* (we hope...) */
if (fsr->regs[SP_REGNUM] != 0
&& fsr->regs[SP_REGNUM] != frame_sp - sp_offset)
fprintf(stderr, "Bad saved SP value %x != %x, offset %x!\n",
fsr->regs[SP_REGNUM],
frame_sp - sp_offset, sp_offset);
fsr->regs[SP_REGNUM] = frame_sp - sp_offset;
return (ip);
}
/* Given an ip value corresponding to the start of a function,
return the ip of the first instruction after the function
prologue. */
CORE_ADDR
skip_prologue (ip)
CORE_ADDR (ip);
{
struct frame_saved_regs saved_regs_dummy;
struct symtab_and_line sal;
CORE_ADDR limit;
sal = find_pc_line (ip, 0);
limit = (sal.end) ? sal.end : 0xffffffff;
return (examine_prologue (ip, limit, (FRAME_ADDR) 0, &saved_regs_dummy,
(struct frame_info *)0 ));
}
/* Put here the code to store, into a struct frame_saved_regs,
the addresses of the saved registers of frame described by FRAME_INFO.
This includes special registers such as pc and fp saved in special
ways in the stack frame. sp is even more special:
the address we return for it IS the sp for the next frame.
We cache the result of doing this in the frame_cache_obstack, since
it is fairly expensive. */
void
frame_find_saved_regs (fi, fsr)
struct frame_info *fi;
struct frame_saved_regs *fsr;
{
register CORE_ADDR next_addr;
register CORE_ADDR *saved_regs;
register int regnum;
register struct frame_saved_regs *cache_fsr;
extern struct obstack frame_cache_obstack;
CORE_ADDR ip;
struct symtab_and_line sal;
CORE_ADDR limit;
if (!fi->fsr)
{
cache_fsr = (struct frame_saved_regs *)
obstack_alloc (&frame_cache_obstack,
sizeof (struct frame_saved_regs));
bzero (cache_fsr, sizeof (struct frame_saved_regs));
fi->fsr = cache_fsr;
/* Find the start and end of the function prologue. If the PC
is in the function prologue, we only consider the part that
has executed already. */
ip = get_pc_function_start (fi->pc);
sal = find_pc_line (ip, 0);
limit = (sal.end && sal.end < fi->pc) ? sal.end: fi->pc;
/* This will fill in fields in *fi as well as in cache_fsr. */
examine_prologue (ip, limit, fi->frame, cache_fsr, fi);
}
if (fsr)
*fsr = *fi->fsr;
}
/* Return the address of the locals block for the frame
described by FI. Returns 0 if the address is unknown.
NOTE! Frame locals are referred to by negative offsets from the
argument pointer, so this is the same as frame_args_address(). */
CORE_ADDR
frame_locals_address (fi)
struct frame_info *fi;
{
register FRAME frame;
struct frame_saved_regs fsr;
CORE_ADDR ap;
if (fi->args_pointer) /* Cached value is likely there. */
return fi->args_pointer;
/* Nope, generate it. */
get_frame_saved_regs (fi, &fsr);
return fi->args_pointer;
}
/* Return the address of the argument block for the frame
described by FI. Returns 0 if the address is unknown. */
CORE_ADDR
frame_args_address (fi)
struct frame_info *fi;
{
register FRAME frame;
struct frame_saved_regs fsr;
CORE_ADDR ap;
if (fi->args_pointer) /* Cached value is likely there. */
return fi->args_pointer;
/* Nope, generate it. */
get_frame_saved_regs (fi, &fsr);
return fi->args_pointer;
}
/* Return the saved PC from this frame.
If the frame has a memory copy of SRP_REGNUM, use that. If not,
just use the register SRP_REGNUM itself. */
CORE_ADDR
frame_saved_pc (frame)
FRAME frame;
{
return read_next_frame_reg(frame, SRP_REGNUM);
}
#if TARGET_BYTE_ORDER != HOST_BYTE_ORDER
you lose
#else /* Host and target byte order the same. */
#define SINGLE_EXP_BITS 8
#define DOUBLE_EXP_BITS 11
int
IEEE_isNAN(fp, len)
int *fp, len;
/* fp points to a single precision OR double precision
* floating point value; len is the number of bytes, either 4 or 8.
* Returns 1 iff fp points to a valid IEEE floating point number.
* Returns 0 if fp points to a denormalized number or a NaN
*/
{
int exponent;
if (len == 4)
{
exponent = *fp;
exponent = exponent << 1 >> (32 - SINGLE_EXP_BITS - 1);
return ((exponent == -1) || (! exponent && *fp));
}
else if (len == 8)
{
exponent = *(fp+1);
exponent = exponent << 1 >> (32 - DOUBLE_EXP_BITS - 1);
return ((exponent == -1) || (! exponent && *fp * *(fp+1)));
}
else return 1;
}
#endif /* Host and target byte order the same. */
static int
pushed_size (prev_words, v)
int prev_words;
struct value *v;
{
switch (TYPE_CODE (VALUE_TYPE (v)))
{
case TYPE_CODE_VOID: /* Void type (values zero length) */
return 0; /* That was easy! */
case TYPE_CODE_PTR: /* Pointer type */
case TYPE_CODE_ENUM: /* Enumeration type */
case TYPE_CODE_INT: /* Integer type */
case TYPE_CODE_REF: /* C++ Reference types */
* gdbtypes.c (create_array_type): Complete rewrite. Now requires a optional type to decorate as an array type, the type of the index, and the bounds of the array. Records this additional info in the array type for use with languages with nonzero array bounds. * gdbtypes.h (enum type_code): Update comment for TYPE_CODE_ARRAY to note that arrays may have bounds. * gdbtypes.h (create_array_type): Update prototype. * c-exp.y (ptype production): Adjust for new create_array_type calling conventions. * coffread.c (decode_type): Call create_array_type rather than handcrafting array types. * convex-tdep.c (value_type): Remove, now use create_array_type. * convex-tdep.c (value_of_trapped_internalvar): Convert calls to vector_type into calls to create_array_type. * dwarfread.c (decode_subscr_data): Name changed to decode_subscript_data_item throughout. * dwarfread.c (decode_subscript_data_item): Rewrite to use create_array_type. Now records index type and range as well. * dwarfread.c (dwarf_read_array_type): Rewrite as part of change to use create_array_type. * dwarfread.c (read_subroutine_type): Test existing user defined types before decorating them, to ensure they are blank, and complain about it if they are not. * dwarfread.c (decode_fund_type): For unrecognized types, always return some valid type (type integer). If the unrecognized type cannot be an implementation defined type, complain as well. * m88k-tdep.c (pushed_size): Update comment for TYPE_CODE_ARRAY. * m88k-tdep.c (store_param): Update comment for TYPE_CODE_ARRAY. * mipsread.c (upgrade_type): Add FIXME comment that code to handcraft arrays should be replaced with call to create_array_type. * stabsread.c (read_array_type): Replace code to handcraft array types with call to create_array_type. * valprint.c (type_print_varspec_prefix): Minor formatting change, join lines that don't need to be split.
1992-12-15 02:52:11 +00:00
case TYPE_CODE_ARRAY: /* Array type, lower & upper bounds */
return 1;
case TYPE_CODE_FLT: /* Floating type */
if (TYPE_LENGTH (VALUE_TYPE (v)) == 4)
return 1;
else
/* Assume that it must be a double. */
if (prev_words & 1) /* at an odd-word boundary */
return 3; /* round to 8-byte boundary */
else
return 2;
case TYPE_CODE_STRUCT: /* C struct or Pascal record */
case TYPE_CODE_UNION: /* C union or Pascal variant part */
return (((TYPE_LENGTH (VALUE_TYPE (v)) + 3) / 4) * 4);
case TYPE_CODE_FUNC: /* Function type */
case TYPE_CODE_SET: /* Pascal sets */
case TYPE_CODE_RANGE: /* Range (integers within bounds) */
* c-exp.y (exp:STRING): Convert C strings into array-of-char constants with an explicit null byte terminator. OP_STRING is now used for real string types. * c-lang.c (builtin_type_*): Move declarations to lang.c since they are used by all languages. * c-lang.c (_initialize_c_language): Move initializations of builtin_type_* to lang.c. * c-typeprint.c (c_type_print_varspec_prefix, c_type_print_varspec_suffix): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * c-valprint.c (c_val_print): Change the way character arrays are printed as strings to be consistent with the way strings are printed when pointer-to-char types are dereferenced. Remove test of print_max before calling val_print_string, which now does it's own test. * eval.c (evaluate_subexp): Add case for OP_ARRAY. * expprint.c (print_subexp, dump_expression): Add case for OP_ARRAY. * expression.h (enum exp_opcode): Add OP_ARRAY and document. * gdbtypes.c (builtin_type_*): Add declarations moved from c-lang.c. * gdbtypes.c (create_string_type): New function to create real string types. * gdbtypes.c (recursive_dump_type): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * gdbtypes.c (_initialize_gdbtypes): Add initializations of builtin_type_* types moved from c-lang.c. * gdbtypes.h (enum type_code): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * gdbtypes.h (builtin_type_string): Add extern declaration. * gdbtypes.h (create_string_type): Add prototype. * m2-lang.c (m2_create_fundamental_type): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * m88k-tdep.c (pushed_size): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * mipsread.c (_initialize_mipsread): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * parse.c (length_of_subexp, prefixify_subexp): Add case for OP_ARRAY. * printcmd.c (print_formatted): Recognize TYPE_CODE_STRING. * typeprint.c (print_type_scalar): TYPE_CODE_PASCAL_ARRAY renamed to TYPE_CODE_STRING. * valops.c (allocate_space_in_inferior): New function and prototype, using code ripped out of value_string. * valops.c (value_string): Rewritten to use new function allocate_space_in_inferior, but temporarily disabled until some other support is in place. * valops.c (value_array): New function to create array constants. * valprint.c (val_print_string): Add comment to document use, complete rewrite to fix several small buglets. * value.h (value_array): Add prototype. * value.h (val_print_string): Change prototype to match rewrite. **** start-sanitize-chill **** * ch-valprint.c (chill_val_print): Add case for TYPE_CODE_STRING. * ch-exp.y (match_character_literal): Disable recognition of control sequence form of character literals and document why. **** end-sanitize-chill ****
1993-01-16 05:15:58 +00:00
case TYPE_CODE_STRING: /* String type */
case TYPE_CODE_MEMBER: /* Member type */
case TYPE_CODE_METHOD: /* Method type */
/* Don't know how to pass these yet. */
case TYPE_CODE_UNDEF: /* Not used; catches errors */
default:
abort ();
}
}
static void
store_parm_word (address, val)
CORE_ADDR address;
int val;
{
write_memory (address, (char *)&val, 4);
}
static int
store_parm (prev_words, left_parm_addr, v)
unsigned int prev_words;
CORE_ADDR left_parm_addr;
struct value *v;
{
CORE_ADDR start = left_parm_addr + (prev_words * 4);
int *val_addr = (int *)VALUE_CONTENTS(v);
switch (TYPE_CODE (VALUE_TYPE (v)))
{
case TYPE_CODE_VOID: /* Void type (values zero length) */
return 0;
case TYPE_CODE_PTR: /* Pointer type */
case TYPE_CODE_ENUM: /* Enumeration type */
case TYPE_CODE_INT: /* Integer type */
* gdbtypes.c (create_array_type): Complete rewrite. Now requires a optional type to decorate as an array type, the type of the index, and the bounds of the array. Records this additional info in the array type for use with languages with nonzero array bounds. * gdbtypes.h (enum type_code): Update comment for TYPE_CODE_ARRAY to note that arrays may have bounds. * gdbtypes.h (create_array_type): Update prototype. * c-exp.y (ptype production): Adjust for new create_array_type calling conventions. * coffread.c (decode_type): Call create_array_type rather than handcrafting array types. * convex-tdep.c (value_type): Remove, now use create_array_type. * convex-tdep.c (value_of_trapped_internalvar): Convert calls to vector_type into calls to create_array_type. * dwarfread.c (decode_subscr_data): Name changed to decode_subscript_data_item throughout. * dwarfread.c (decode_subscript_data_item): Rewrite to use create_array_type. Now records index type and range as well. * dwarfread.c (dwarf_read_array_type): Rewrite as part of change to use create_array_type. * dwarfread.c (read_subroutine_type): Test existing user defined types before decorating them, to ensure they are blank, and complain about it if they are not. * dwarfread.c (decode_fund_type): For unrecognized types, always return some valid type (type integer). If the unrecognized type cannot be an implementation defined type, complain as well. * m88k-tdep.c (pushed_size): Update comment for TYPE_CODE_ARRAY. * m88k-tdep.c (store_param): Update comment for TYPE_CODE_ARRAY. * mipsread.c (upgrade_type): Add FIXME comment that code to handcraft arrays should be replaced with call to create_array_type. * stabsread.c (read_array_type): Replace code to handcraft array types with call to create_array_type. * valprint.c (type_print_varspec_prefix): Minor formatting change, join lines that don't need to be split.
1992-12-15 02:52:11 +00:00
case TYPE_CODE_ARRAY: /* Array type, lower & upper bounds */
case TYPE_CODE_REF: /* C++ Reference types */
store_parm_word (start, *val_addr);
return 1;
case TYPE_CODE_FLT: /* Floating type */
if (TYPE_LENGTH (VALUE_TYPE (v)) == 4)
{
store_parm_word (start, *val_addr);
return 1;
}
else
{
store_parm_word (start + ((prev_words & 1) * 4), val_addr[0]);
store_parm_word (start + ((prev_words & 1) * 4) + 4, val_addr[1]);
return 2 + (prev_words & 1);
}
case TYPE_CODE_STRUCT: /* C struct or Pascal record */
case TYPE_CODE_UNION: /* C union or Pascal variant part */
{
unsigned int words = (((TYPE_LENGTH (VALUE_TYPE (v)) + 3) / 4) * 4);
unsigned int word;
for (word = 0; word < words; word++)
store_parm_word (start + (word * 4), val_addr[word]);
return words;
}
default:
abort ();
}
}
/* This routine sets up all of the parameter values needed to make a pseudo
call. The name "push_parameters" is a misnomer on some archs,
because (on the m88k) most parameters generally end up being passed in
registers rather than on the stack. In this routine however, we do
end up storing *all* parameter values onto the stack (even if we will
realize later that some of these stores were unnecessary). */
#define FIRST_PARM_REGNUM 2
void
push_parameters (return_type, struct_conv, nargs, args)
struct type *return_type;
int struct_conv;
int nargs;
value *args;
{
int parm_num;
unsigned int p_words = 0;
CORE_ADDR left_parm_addr;
/* Start out by creating a space for the return value (if need be). We
only need to do this if the return value is a struct or union. If we
do make a space for a struct or union return value, then we must also
arrange for the base address of that space to go into r12, which is the
standard place to pass the address of the return value area to the
callee. Note that only structs and unions are returned in this fashion.
Ints, enums, pointers, and floats are returned into r2. Doubles are
returned into the register pair {r2,r3}. Note also that the space
reserved for a struct or union return value only has to be word aligned
(not double-word) but it is double-word aligned here anyway (just in
case that becomes important someday). */
switch (TYPE_CODE (return_type))
{
case TYPE_CODE_STRUCT:
case TYPE_CODE_UNION:
{
int return_bytes = ((TYPE_LENGTH (return_type) + 7) / 8) * 8;
CORE_ADDR rv_addr;
rv_addr = read_register (SP_REGNUM) - return_bytes;
write_register (SP_REGNUM, rv_addr); /* push space onto the stack */
write_register (SRA_REGNUM, rv_addr);/* set return value register */
}
}
/* Here we make a pre-pass on the whole parameter list to figure out exactly
how many words worth of stuff we are going to pass. */
for (p_words = 0, parm_num = 0; parm_num < nargs; parm_num++)
p_words += pushed_size (p_words, value_arg_coerce (args[parm_num]));
/* Now, check to see if we have to round up the number of parameter words
to get up to the next 8-bytes boundary. This may be necessary because
of the software convention to always keep the stack aligned on an 8-byte
boundary. */
if (p_words & 1)
p_words++; /* round to 8-byte boundary */
/* Now figure out the absolute address of the leftmost parameter, and update
the stack pointer to point at that address. */
left_parm_addr = read_register (SP_REGNUM) - (p_words * 4);
write_register (SP_REGNUM, left_parm_addr);
/* Now we can go through all of the parameters (in left-to-right order)
and write them to their parameter stack slots. Note that we are not
really "pushing" the parameter values. The stack space for these values
was already allocated above. Now we are just filling it up. */
for (p_words = 0, parm_num = 0; parm_num < nargs; parm_num++)
p_words +=
store_parm (p_words, left_parm_addr, value_arg_coerce (args[parm_num]));
/* Now that we are all done storing the parameter values into the stack, we
must go back and load up the parameter registers with the values from the
corresponding stack slots. Note that in the two cases of (a) gaps in the
parameter word sequence causes by (otherwise) misaligned doubles, and (b)
slots correcponding to structs or unions, the work we do here in loading
some parameter registers may be unnecessary, but who cares? */
for (p_words = 0; p_words < 8; p_words++)
{
write_register (FIRST_PARM_REGNUM + p_words,
read_memory_integer (left_parm_addr + (p_words * 4), 4));
}
}
void
pop_frame ()
{
error ("Feature not implemented for the m88k yet.");
return;
}
void
collect_returned_value (rval, value_type, struct_return, nargs, args)
value *rval;
struct type *value_type;
int struct_return;
int nargs;
value *args;
{
char retbuf[REGISTER_BYTES];
bcopy (registers, retbuf, REGISTER_BYTES);
*rval = value_being_returned (value_type, retbuf, struct_return);
return;
}
#if 0
/* Now handled in a machine independent way with CALL_DUMMY_LOCATION. */
/* Stuff a breakpoint instruction onto the stack (or elsewhere if the stack
is not a good place for it). Return the address at which the instruction
got stuffed, or zero if we were unable to stuff it anywhere. */
CORE_ADDR
push_breakpoint ()
{
static char breakpoint_insn[] = BREAKPOINT;
extern CORE_ADDR text_end; /* of inferior */
static char readback_buffer[] = BREAKPOINT;
int i;
/* With a little bit of luck, we can just stash the breakpoint instruction
in the word just beyond the end of normal text space. For systems on
which the hardware will not allow us to execute out of the stack segment,
we have to hope that we *are* at least allowed to effectively extend the
text segment by one word. If the actual end of user's the text segment
happens to fall right at a page boundary this trick may fail. Note that
we check for this by reading after writing, and comparing in order to
be sure that the write worked. */
write_memory (text_end, &breakpoint_insn, 4);
/* Fill the readback buffer with some garbage which is certain to be
unequal to the breakpoint insn. That way we can tell if the
following read doesn't actually succeed. */
for (i = 0; i < sizeof (readback_buffer); i++)
readback_buffer[i] = ~ readback_buffer[i]; /* Invert the bits */
/* Now check that the breakpoint insn was successfully installed. */
read_memory (text_end, readback_buffer, sizeof (readback_buffer));
for (i = 0; i < sizeof (readback_buffer); i++)
if (readback_buffer[i] != breakpoint_insn[i])
return 0; /* Failed to install! */
return text_end;
}
#endif