summaryrefslogtreecommitdiff
path: root/gdb/target.h
diff options
context:
space:
mode:
Diffstat (limited to 'gdb/target.h')
-rw-r--r--gdb/target.h1264
1 files changed, 1264 insertions, 0 deletions
diff --git a/gdb/target.h b/gdb/target.h
new file mode 100644
index 00000000000..21c3babddba
--- /dev/null
+++ b/gdb/target.h
@@ -0,0 +1,1264 @@
+/* Interface between GDB and target environments, including files and processes
+ Copyright 1990, 91, 92, 93, 94, 1999 Free Software Foundation, Inc.
+ Contributed by Cygnus Support. Written by John Gilmore.
+
+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. */
+
+#if !defined (TARGET_H)
+#define TARGET_H
+
+/* This include file defines the interface between the main part
+ of the debugger, and the part which is target-specific, or
+ specific to the communications interface between us and the
+ target.
+
+ A TARGET is an interface between the debugger and a particular
+ kind of file or process. Targets can be STACKED in STRATA,
+ so that more than one target can potentially respond to a request.
+ In particular, memory accesses will walk down the stack of targets
+ until they find a target that is interested in handling that particular
+ address. STRATA are artificial boundaries on the stack, within
+ which particular kinds of targets live. Strata exist so that
+ people don't get confused by pushing e.g. a process target and then
+ a file target, and wondering why they can't see the current values
+ of variables any more (the file target is handling them and they
+ never get to the process target). So when you push a file target,
+ it goes into the file stratum, which is always below the process
+ stratum. */
+
+#include "bfd.h"
+#include "symtab.h"
+
+enum strata {
+ dummy_stratum, /* The lowest of the low */
+ file_stratum, /* Executable files, etc */
+ core_stratum, /* Core dump files */
+ download_stratum, /* Downloading of remote targets */
+ process_stratum /* Executing processes */
+};
+
+enum thread_control_capabilities {
+ tc_none = 0, /* Default: can't control thread execution. */
+ tc_schedlock = 1, /* Can lock the thread scheduler. */
+ tc_switch = 2 /* Can switch the running thread on demand. */
+};
+
+/* Stuff for target_wait. */
+
+/* Generally, what has the program done? */
+enum target_waitkind {
+ /* The program has exited. The exit status is in value.integer. */
+ TARGET_WAITKIND_EXITED,
+
+ /* The program has stopped with a signal. Which signal is in value.sig. */
+ TARGET_WAITKIND_STOPPED,
+
+ /* The program has terminated with a signal. Which signal is in
+ value.sig. */
+ TARGET_WAITKIND_SIGNALLED,
+
+ /* The program is letting us know that it dynamically loaded something
+ (e.g. it called load(2) on AIX). */
+ TARGET_WAITKIND_LOADED,
+
+ /* The program has forked. A "related" process' ID is in value.related_pid.
+ I.e., if the child forks, value.related_pid is the parent's ID.
+ */
+ TARGET_WAITKIND_FORKED,
+
+ /* The program has vforked. A "related" process's ID is in value.related_pid.
+ */
+ TARGET_WAITKIND_VFORKED,
+
+ /* The program has exec'ed a new executable file. The new file's pathname
+ is pointed to by value.execd_pathname.
+ */
+ TARGET_WAITKIND_EXECD,
+
+ /* The program has entered or returned from a system call. On HP-UX, this
+ is used in the hardware watchpoint implementation. The syscall's unique
+ integer ID number is in value.syscall_id;
+ */
+ TARGET_WAITKIND_SYSCALL_ENTRY,
+ TARGET_WAITKIND_SYSCALL_RETURN,
+
+ /* Nothing happened, but we stopped anyway. This perhaps should be handled
+ within target_wait, but I'm not sure target_wait should be resuming the
+ inferior. */
+ TARGET_WAITKIND_SPURIOUS
+ };
+
+/* The numbering of these signals is chosen to match traditional unix
+ signals (insofar as various unices use the same numbers, anyway).
+ It is also the numbering of the GDB remote protocol. Other remote
+ protocols, if they use a different numbering, should make sure to
+ translate appropriately. */
+
+/* This is based strongly on Unix/POSIX signals for several reasons:
+ (1) This set of signals represents a widely-accepted attempt to
+ represent events of this sort in a portable fashion, (2) we want a
+ signal to make it from wait to child_wait to the user intact, (3) many
+ remote protocols use a similar encoding. However, it is
+ recognized that this set of signals has limitations (such as not
+ distinguishing between various kinds of SIGSEGV, or not
+ distinguishing hitting a breakpoint from finishing a single step).
+ So in the future we may get around this either by adding additional
+ signals for breakpoint, single-step, etc., or by adding signal
+ codes; the latter seems more in the spirit of what BSD, System V,
+ etc. are doing to address these issues. */
+
+/* For an explanation of what each signal means, see
+ target_signal_to_string. */
+
+enum target_signal {
+ /* Used some places (e.g. stop_signal) to record the concept that
+ there is no signal. */
+ TARGET_SIGNAL_0 = 0,
+ TARGET_SIGNAL_FIRST = 0,
+ TARGET_SIGNAL_HUP = 1,
+ TARGET_SIGNAL_INT = 2,
+ TARGET_SIGNAL_QUIT = 3,
+ TARGET_SIGNAL_ILL = 4,
+ TARGET_SIGNAL_TRAP = 5,
+ TARGET_SIGNAL_ABRT = 6,
+ TARGET_SIGNAL_EMT = 7,
+ TARGET_SIGNAL_FPE = 8,
+ TARGET_SIGNAL_KILL = 9,
+ TARGET_SIGNAL_BUS = 10,
+ TARGET_SIGNAL_SEGV = 11,
+ TARGET_SIGNAL_SYS = 12,
+ TARGET_SIGNAL_PIPE = 13,
+ TARGET_SIGNAL_ALRM = 14,
+ TARGET_SIGNAL_TERM = 15,
+ TARGET_SIGNAL_URG = 16,
+ TARGET_SIGNAL_STOP = 17,
+ TARGET_SIGNAL_TSTP = 18,
+ TARGET_SIGNAL_CONT = 19,
+ TARGET_SIGNAL_CHLD = 20,
+ TARGET_SIGNAL_TTIN = 21,
+ TARGET_SIGNAL_TTOU = 22,
+ TARGET_SIGNAL_IO = 23,
+ TARGET_SIGNAL_XCPU = 24,
+ TARGET_SIGNAL_XFSZ = 25,
+ TARGET_SIGNAL_VTALRM = 26,
+ TARGET_SIGNAL_PROF = 27,
+ TARGET_SIGNAL_WINCH = 28,
+ TARGET_SIGNAL_LOST = 29,
+ TARGET_SIGNAL_USR1 = 30,
+ TARGET_SIGNAL_USR2 = 31,
+ TARGET_SIGNAL_PWR = 32,
+ /* Similar to SIGIO. Perhaps they should have the same number. */
+ TARGET_SIGNAL_POLL = 33,
+ TARGET_SIGNAL_WIND = 34,
+ TARGET_SIGNAL_PHONE = 35,
+ TARGET_SIGNAL_WAITING = 36,
+ TARGET_SIGNAL_LWP = 37,
+ TARGET_SIGNAL_DANGER = 38,
+ TARGET_SIGNAL_GRANT = 39,
+ TARGET_SIGNAL_RETRACT = 40,
+ TARGET_SIGNAL_MSG = 41,
+ TARGET_SIGNAL_SOUND = 42,
+ TARGET_SIGNAL_SAK = 43,
+ TARGET_SIGNAL_PRIO = 44,
+ TARGET_SIGNAL_REALTIME_33 = 45,
+ TARGET_SIGNAL_REALTIME_34 = 46,
+ TARGET_SIGNAL_REALTIME_35 = 47,
+ TARGET_SIGNAL_REALTIME_36 = 48,
+ TARGET_SIGNAL_REALTIME_37 = 49,
+ TARGET_SIGNAL_REALTIME_38 = 50,
+ TARGET_SIGNAL_REALTIME_39 = 51,
+ TARGET_SIGNAL_REALTIME_40 = 52,
+ TARGET_SIGNAL_REALTIME_41 = 53,
+ TARGET_SIGNAL_REALTIME_42 = 54,
+ TARGET_SIGNAL_REALTIME_43 = 55,
+ TARGET_SIGNAL_REALTIME_44 = 56,
+ TARGET_SIGNAL_REALTIME_45 = 57,
+ TARGET_SIGNAL_REALTIME_46 = 58,
+ TARGET_SIGNAL_REALTIME_47 = 59,
+ TARGET_SIGNAL_REALTIME_48 = 60,
+ TARGET_SIGNAL_REALTIME_49 = 61,
+ TARGET_SIGNAL_REALTIME_50 = 62,
+ TARGET_SIGNAL_REALTIME_51 = 63,
+ TARGET_SIGNAL_REALTIME_52 = 64,
+ TARGET_SIGNAL_REALTIME_53 = 65,
+ TARGET_SIGNAL_REALTIME_54 = 66,
+ TARGET_SIGNAL_REALTIME_55 = 67,
+ TARGET_SIGNAL_REALTIME_56 = 68,
+ TARGET_SIGNAL_REALTIME_57 = 69,
+ TARGET_SIGNAL_REALTIME_58 = 70,
+ TARGET_SIGNAL_REALTIME_59 = 71,
+ TARGET_SIGNAL_REALTIME_60 = 72,
+ TARGET_SIGNAL_REALTIME_61 = 73,
+ TARGET_SIGNAL_REALTIME_62 = 74,
+ TARGET_SIGNAL_REALTIME_63 = 75,
+#if defined(MACH) || defined(__MACH__)
+ /* Mach exceptions */
+ TARGET_EXC_BAD_ACCESS = 76,
+ TARGET_EXC_BAD_INSTRUCTION = 77,
+ TARGET_EXC_ARITHMETIC = 78,
+ TARGET_EXC_EMULATION = 79,
+ TARGET_EXC_SOFTWARE = 80,
+ TARGET_EXC_BREAKPOINT = 81,
+#endif
+ /* Some signal we don't know about. */
+ TARGET_SIGNAL_UNKNOWN,
+
+ /* Use whatever signal we use when one is not specifically specified
+ (for passing to proceed and so on). */
+ TARGET_SIGNAL_DEFAULT,
+
+ /* Last and unused enum value, for sizing arrays, etc. */
+ TARGET_SIGNAL_LAST
+};
+
+struct target_waitstatus {
+ enum target_waitkind kind;
+
+ /* Forked child pid, execd pathname, exit status or signal number. */
+ union {
+ int integer;
+ enum target_signal sig;
+ int related_pid;
+ char * execd_pathname;
+ int syscall_id;
+ } value;
+};
+
+/* Return the string for a signal. */
+extern char *target_signal_to_string PARAMS ((enum target_signal));
+
+/* Return the name (SIGHUP, etc.) for a signal. */
+extern char *target_signal_to_name PARAMS ((enum target_signal));
+
+/* Given a name (SIGHUP, etc.), return its signal. */
+enum target_signal target_signal_from_name PARAMS ((char *));
+
+
+/* If certain kinds of activity happen, target_wait should perform
+ callbacks. */
+/* Right now we just call (*TARGET_ACTIVITY_FUNCTION) if I/O is possible
+ on TARGET_ACTIVITY_FD. */
+extern int target_activity_fd;
+/* Returns zero to leave the inferior alone, one to interrupt it. */
+extern int (*target_activity_function) PARAMS ((void));
+
+struct target_ops
+{
+ char *to_shortname; /* Name this target type */
+ char *to_longname; /* Name for printing */
+ char *to_doc; /* Documentation. Does not include trailing
+ newline, and starts with a one-line descrip-
+ tion (probably similar to to_longname). */
+ void (*to_open) PARAMS ((char *, int));
+ void (*to_close) PARAMS ((int));
+ void (*to_attach) PARAMS ((char *, int));
+ void (*to_post_attach) PARAMS ((int));
+ void (*to_require_attach) PARAMS ((char *, int));
+ void (*to_detach) PARAMS ((char *, int));
+ void (*to_require_detach) PARAMS ((int, char *, int));
+ void (*to_resume) PARAMS ((int, int, enum target_signal));
+ int (*to_wait) PARAMS ((int, struct target_waitstatus *));
+ void (*to_post_wait) PARAMS ((int, int));
+ void (*to_fetch_registers) PARAMS ((int));
+ void (*to_store_registers) PARAMS ((int));
+ void (*to_prepare_to_store) PARAMS ((void));
+
+ /* Transfer LEN bytes of memory between GDB address MYADDR and
+ target address MEMADDR. If WRITE, transfer them to the target, else
+ transfer them from the target. TARGET is the target from which we
+ get this function.
+
+ Return value, N, is one of the following:
+
+ 0 means that we can't handle this. If errno has been set, it is the
+ error which prevented us from doing it (FIXME: What about bfd_error?).
+
+ positive (call it N) means that we have transferred N bytes
+ starting at MEMADDR. We might be able to handle more bytes
+ beyond this length, but no promises.
+
+ negative (call its absolute value N) means that we cannot
+ transfer right at MEMADDR, but we could transfer at least
+ something at MEMADDR + N. */
+
+ int (*to_xfer_memory) PARAMS ((CORE_ADDR memaddr, char *myaddr,
+ int len, int write,
+ struct target_ops * target));
+
+#if 0
+ /* Enable this after 4.12. */
+
+ /* Search target memory. Start at STARTADDR and take LEN bytes of
+ target memory, and them with MASK, and compare to DATA. If they
+ match, set *ADDR_FOUND to the address we found it at, store the data
+ we found at LEN bytes starting at DATA_FOUND, and return. If
+ not, add INCREMENT to the search address and keep trying until
+ the search address is outside of the range [LORANGE,HIRANGE).
+
+ If we don't find anything, set *ADDR_FOUND to (CORE_ADDR)0 and return. */
+ void (*to_search) PARAMS ((int len, char *data, char *mask,
+ CORE_ADDR startaddr, int increment,
+ CORE_ADDR lorange, CORE_ADDR hirange,
+ CORE_ADDR *addr_found, char *data_found));
+
+#define target_search(len, data, mask, startaddr, increment, lorange, hirange, addr_found, data_found) \
+ (*current_target.to_search) (len, data, mask, startaddr, increment, \
+ lorange, hirange, addr_found, data_found)
+#endif /* 0 */
+
+ void (*to_files_info) PARAMS ((struct target_ops *));
+ int (*to_insert_breakpoint) PARAMS ((CORE_ADDR, char *));
+ int (*to_remove_breakpoint) PARAMS ((CORE_ADDR, char *));
+ void (*to_terminal_init) PARAMS ((void));
+ void (*to_terminal_inferior) PARAMS ((void));
+ void (*to_terminal_ours_for_output) PARAMS ((void));
+ void (*to_terminal_ours) PARAMS ((void));
+ void (*to_terminal_info) PARAMS ((char *, int));
+ void (*to_kill) PARAMS ((void));
+ void (*to_load) PARAMS ((char *, int));
+ int (*to_lookup_symbol) PARAMS ((char *, CORE_ADDR *));
+ void (*to_create_inferior) PARAMS ((char *, char *, char **));
+ void (*to_post_startup_inferior) PARAMS ((int));
+ void (*to_acknowledge_created_inferior) PARAMS ((int));
+ void (*to_clone_and_follow_inferior) PARAMS ((int, int *));
+ void (*to_post_follow_inferior_by_clone) PARAMS ((void));
+ int (*to_insert_fork_catchpoint) PARAMS ((int));
+ int (*to_remove_fork_catchpoint) PARAMS ((int));
+ int (*to_insert_vfork_catchpoint) PARAMS ((int));
+ int (*to_remove_vfork_catchpoint) PARAMS ((int));
+ int (*to_has_forked) PARAMS ((int, int *));
+ int (*to_has_vforked) PARAMS ((int, int *));
+ int (*to_can_follow_vfork_prior_to_exec) PARAMS ((void));
+ void (*to_post_follow_vfork) PARAMS ((int, int, int, int));
+ int (*to_insert_exec_catchpoint) PARAMS ((int));
+ int (*to_remove_exec_catchpoint) PARAMS ((int));
+ int (*to_has_execd) PARAMS ((int, char **));
+ int (*to_reported_exec_events_per_exec_call) PARAMS ((void));
+ int (*to_has_syscall_event) PARAMS ((int, enum target_waitkind *, int *));
+ int (*to_has_exited) PARAMS ((int, int, int *));
+ void (*to_mourn_inferior) PARAMS ((void));
+ int (*to_can_run) PARAMS ((void));
+ void (*to_notice_signals) PARAMS ((int pid));
+ int (*to_thread_alive) PARAMS ((int pid));
+ void (*to_stop) PARAMS ((void));
+ int (*to_query) PARAMS ((int/*char*/, char *, char *, int *));
+ struct symtab_and_line * (*to_enable_exception_callback) PARAMS ((enum exception_event_kind, int));
+ struct exception_event_record * (*to_get_current_exception_event) PARAMS ((void));
+ char * (*to_pid_to_exec_file) PARAMS ((int pid));
+ char * (*to_core_file_to_sym_file) PARAMS ((char *));
+ enum strata to_stratum;
+ struct target_ops
+ *DONT_USE; /* formerly to_next */
+ int to_has_all_memory;
+ int to_has_memory;
+ int to_has_stack;
+ int to_has_registers;
+ int to_has_execution;
+ int to_has_thread_control; /* control thread execution */
+ struct section_table
+ *to_sections;
+ struct section_table
+ *to_sections_end;
+ int to_magic;
+ /* Need sub-structure for target machine related rather than comm related? */
+};
+
+/* Magic number for checking ops size. If a struct doesn't end with this
+ number, somebody changed the declaration but didn't change all the
+ places that initialize one. */
+
+#define OPS_MAGIC 3840
+
+/* The ops structure for our "current" target process. This should
+ never be NULL. If there is no target, it points to the dummy_target. */
+
+extern struct target_ops current_target;
+
+/* An item on the target stack. */
+
+struct target_stack_item
+{
+ struct target_stack_item *next;
+ struct target_ops *target_ops;
+};
+
+/* The target stack. */
+
+extern struct target_stack_item *target_stack;
+
+/* Define easy words for doing these operations on our current target. */
+
+#define target_shortname (current_target.to_shortname)
+#define target_longname (current_target.to_longname)
+
+/* The open routine takes the rest of the parameters from the command,
+ and (if successful) pushes a new target onto the stack.
+ Targets should supply this routine, if only to provide an error message. */
+#define target_open(name, from_tty) \
+ (*current_target.to_open) (name, from_tty)
+
+/* Does whatever cleanup is required for a target that we are no longer
+ going to be calling. Argument says whether we are quitting gdb and
+ should not get hung in case of errors, or whether we want a clean
+ termination even if it takes a while. This routine is automatically
+ always called just before a routine is popped off the target stack.
+ Closing file descriptors and freeing memory are typical things it should
+ do. */
+
+#define target_close(quitting) \
+ (*current_target.to_close) (quitting)
+
+/* Attaches to a process on the target side. Arguments are as passed
+ to the `attach' command by the user. This routine can be called
+ when the target is not on the target-stack, if the target_can_run
+ routine returns 1; in that case, it must push itself onto the stack.
+ Upon exit, the target should be ready for normal operations, and
+ should be ready to deliver the status of the process immediately
+ (without waiting) to an upcoming target_wait call. */
+
+#define target_attach(args, from_tty) \
+ (*current_target.to_attach) (args, from_tty)
+
+/* The target_attach operation places a process under debugger control,
+ and stops the process.
+
+ This operation provides a target-specific hook that allows the
+ necessary bookkeeping to be performed after an attach completes.
+ */
+#define target_post_attach(pid) \
+ (*current_target.to_post_attach) (pid)
+
+/* Attaches to a process on the target side, if not already attached.
+ (If already attached, takes no action.)
+
+ This operation can be used to follow the child process of a fork.
+ On some targets, such child processes of an original inferior process
+ are automatically under debugger control, and thus do not require an
+ actual attach operation. */
+
+#define target_require_attach(args, from_tty) \
+ (*current_target.to_require_attach) (args, from_tty)
+
+/* Takes a program previously attached to and detaches it.
+ The program may resume execution (some targets do, some don't) and will
+ no longer stop on signals, etc. We better not have left any breakpoints
+ in the program or it'll die when it hits one. ARGS is arguments
+ typed by the user (e.g. a signal to send the process). FROM_TTY
+ says whether to be verbose or not. */
+
+extern void
+target_detach PARAMS ((char *, int));
+
+/* Detaches from a process on the target side, if not already dettached.
+ (If already detached, takes no action.)
+
+ This operation can be used to follow the parent process of a fork.
+ On some targets, such child processes of an original inferior process
+ are automatically under debugger control, and thus do require an actual
+ detach operation.
+
+ PID is the process id of the child to detach from.
+ ARGS is arguments typed by the user (e.g. a signal to send the process).
+ FROM_TTY says whether to be verbose or not. */
+
+#define target_require_detach(pid, args, from_tty) \
+ (*current_target.to_require_detach) (pid, args, from_tty)
+
+/* Resume execution of the target process PID. STEP says whether to
+ single-step or to run free; SIGGNAL is the signal to be given to
+ the target, or TARGET_SIGNAL_0 for no signal. The caller may not
+ pass TARGET_SIGNAL_DEFAULT. */
+
+#define target_resume(pid, step, siggnal) \
+ (*current_target.to_resume) (pid, step, siggnal)
+
+/* Wait for process pid to do something. Pid = -1 to wait for any pid
+ to do something. Return pid of child, or -1 in case of error;
+ store status through argument pointer STATUS. Note that it is
+ *not* OK to return_to_top_level out of target_wait without popping
+ the debugging target from the stack; GDB isn't prepared to get back
+ to the prompt with a debugging target but without the frame cache,
+ stop_pc, etc., set up. */
+
+#define target_wait(pid, status) \
+ (*current_target.to_wait) (pid, status)
+
+/* The target_wait operation waits for a process event to occur, and
+ thereby stop the process.
+
+ On some targets, certain events may happen in sequences. gdb's
+ correct response to any single event of such a sequence may require
+ knowledge of what earlier events in the sequence have been seen.
+
+ This operation provides a target-specific hook that allows the
+ necessary bookkeeping to be performed to track such sequences.
+ */
+
+#define target_post_wait(pid, status) \
+ (*current_target.to_post_wait) (pid, status)
+
+/* Fetch register REGNO, or all regs if regno == -1. No result. */
+
+#define target_fetch_registers(regno) \
+ (*current_target.to_fetch_registers) (regno)
+
+/* Store at least register REGNO, or all regs if REGNO == -1.
+ It can store as many registers as it wants to, so target_prepare_to_store
+ must have been previously called. Calls error() if there are problems. */
+
+#define target_store_registers(regs) \
+ (*current_target.to_store_registers) (regs)
+
+/* Get ready to modify the registers array. On machines which store
+ individual registers, this doesn't need to do anything. On machines
+ which store all the registers in one fell swoop, this makes sure
+ that REGISTERS contains all the registers from the program being
+ debugged. */
+
+#define target_prepare_to_store() \
+ (*current_target.to_prepare_to_store) ()
+
+extern int target_read_string PARAMS ((CORE_ADDR, char **, int, int *));
+
+extern int
+target_read_memory PARAMS ((CORE_ADDR memaddr, char *myaddr, int len));
+
+extern int
+target_read_memory_section PARAMS ((CORE_ADDR memaddr, char *myaddr, int len,
+ asection *bfd_section));
+
+extern int
+target_read_memory_partial PARAMS ((CORE_ADDR, char *, int, int *));
+
+extern int
+target_write_memory PARAMS ((CORE_ADDR, char *, int));
+
+extern int
+xfer_memory PARAMS ((CORE_ADDR, char *, int, int, struct target_ops *));
+
+extern int
+child_xfer_memory PARAMS ((CORE_ADDR, char *, int, int, struct target_ops *));
+
+extern char *
+child_pid_to_exec_file PARAMS ((int));
+
+extern char *
+child_core_file_to_sym_file PARAMS ((char *));
+
+#if defined(CHILD_POST_ATTACH)
+extern void
+child_post_attach PARAMS ((int));
+#endif
+
+extern void
+child_post_wait PARAMS ((int, int));
+
+extern void
+child_post_startup_inferior PARAMS ((int));
+
+extern void
+child_acknowledge_created_inferior PARAMS ((int));
+
+extern void
+child_clone_and_follow_inferior PARAMS ((int, int *));
+
+extern void
+child_post_follow_inferior_by_clone PARAMS ((void));
+
+extern int
+child_insert_fork_catchpoint PARAMS ((int));
+
+extern int
+child_remove_fork_catchpoint PARAMS ((int));
+
+extern int
+child_insert_vfork_catchpoint PARAMS ((int));
+
+extern int
+child_remove_vfork_catchpoint PARAMS ((int));
+
+extern int
+child_has_forked PARAMS ((int, int *));
+
+extern int
+child_has_vforked PARAMS ((int, int *));
+
+extern void
+child_acknowledge_created_inferior PARAMS ((int));
+
+extern int
+child_can_follow_vfork_prior_to_exec PARAMS ((void));
+
+extern void
+child_post_follow_vfork PARAMS ((int, int, int, int));
+
+extern int
+child_insert_exec_catchpoint PARAMS ((int));
+
+extern int
+child_remove_exec_catchpoint PARAMS ((int));
+
+extern int
+child_has_execd PARAMS ((int, char **));
+
+extern int
+child_reported_exec_events_per_exec_call PARAMS ((void));
+
+extern int
+child_has_syscall_event PARAMS ((int, enum target_waitkind *, int *));
+
+extern int
+child_has_exited PARAMS ((int, int, int *));
+
+extern int
+child_thread_alive PARAMS ((int));
+
+/* From exec.c */
+
+extern void
+print_section_info PARAMS ((struct target_ops *, bfd *));
+
+/* Print a line about the current target. */
+
+#define target_files_info() \
+ (*current_target.to_files_info) (&current_target)
+
+/* Insert a breakpoint at address ADDR in the target machine.
+ SAVE is a pointer to memory allocated for saving the
+ target contents. It is guaranteed by the caller to be long enough
+ to save "sizeof BREAKPOINT" bytes. Result is 0 for success, or
+ an errno value. */
+
+#define target_insert_breakpoint(addr, save) \
+ (*current_target.to_insert_breakpoint) (addr, save)
+
+/* Remove a breakpoint at address ADDR in the target machine.
+ SAVE is a pointer to the same save area
+ that was previously passed to target_insert_breakpoint.
+ Result is 0 for success, or an errno value. */
+
+#define target_remove_breakpoint(addr, save) \
+ (*current_target.to_remove_breakpoint) (addr, save)
+
+/* Initialize the terminal settings we record for the inferior,
+ before we actually run the inferior. */
+
+#define target_terminal_init() \
+ (*current_target.to_terminal_init) ()
+
+/* Put the inferior's terminal settings into effect.
+ This is preparation for starting or resuming the inferior. */
+
+#define target_terminal_inferior() \
+ (*current_target.to_terminal_inferior) ()
+
+/* Put some of our terminal settings into effect,
+ enough to get proper results from our output,
+ but do not change into or out of RAW mode
+ so that no input is discarded.
+
+ After doing this, either terminal_ours or terminal_inferior
+ should be called to get back to a normal state of affairs. */
+
+#define target_terminal_ours_for_output() \
+ (*current_target.to_terminal_ours_for_output) ()
+
+/* Put our terminal settings into effect.
+ First record the inferior's terminal settings
+ so they can be restored properly later. */
+
+#define target_terminal_ours() \
+ (*current_target.to_terminal_ours) ()
+
+/* Print useful information about our terminal status, if such a thing
+ exists. */
+
+#define target_terminal_info(arg, from_tty) \
+ (*current_target.to_terminal_info) (arg, from_tty)
+
+/* Kill the inferior process. Make it go away. */
+
+#define target_kill() \
+ (*current_target.to_kill) ()
+
+/* Load an executable file into the target process. This is expected to
+ not only bring new code into the target process, but also to update
+ GDB's symbol tables to match. */
+
+#define target_load(arg, from_tty) \
+ (*current_target.to_load) (arg, from_tty)
+
+/* Look up a symbol in the target's symbol table. NAME is the symbol
+ name. ADDRP is a CORE_ADDR * pointing to where the value of the symbol
+ should be returned. The result is 0 if successful, nonzero if the
+ symbol does not exist in the target environment. This function should
+ not call error() if communication with the target is interrupted, since
+ it is called from symbol reading, but should return nonzero, possibly
+ doing a complain(). */
+
+#define target_lookup_symbol(name, addrp) \
+ (*current_target.to_lookup_symbol) (name, addrp)
+
+/* Start an inferior process and set inferior_pid to its pid.
+ EXEC_FILE is the file to run.
+ ALLARGS is a string containing the arguments to the program.
+ ENV is the environment vector to pass. Errors reported with error().
+ On VxWorks and various standalone systems, we ignore exec_file. */
+
+#define target_create_inferior(exec_file, args, env) \
+ (*current_target.to_create_inferior) (exec_file, args, env)
+
+
+/* Some targets (such as ttrace-based HPUX) don't allow us to request
+ notification of inferior events such as fork and vork immediately
+ after the inferior is created. (This because of how gdb gets an
+ inferior created via invoking a shell to do it. In such a scenario,
+ if the shell init file has commands in it, the shell will fork and
+ exec for each of those commands, and we will see each such fork
+ event. Very bad.)
+
+ Such targets will supply an appropriate definition for this function.
+ */
+#define target_post_startup_inferior(pid) \
+ (*current_target.to_post_startup_inferior) (pid)
+
+/* On some targets, the sequence of starting up an inferior requires
+ some synchronization between gdb and the new inferior process, PID.
+ */
+#define target_acknowledge_created_inferior(pid) \
+ (*current_target.to_acknowledge_created_inferior) (pid)
+
+/* An inferior process has been created via a fork() or similar
+ system call. This function will clone the debugger, then ensure
+ that CHILD_PID is attached to by that debugger.
+
+ FOLLOWED_CHILD is set TRUE on return *for the clone debugger only*,
+ and FALSE otherwise. (The original and clone debuggers can use this
+ to determine which they are, if need be.)
+
+ (This is not a terribly useful feature without a GUI to prevent
+ the two debuggers from competing for shell input.)
+ */
+#define target_clone_and_follow_inferior(child_pid,followed_child) \
+ (*current_target.to_clone_and_follow_inferior) (child_pid, followed_child)
+
+/* This operation is intended to be used as the last in a sequence of
+ steps taken when following both parent and child of a fork. This
+ is used by a clone of the debugger, which will follow the child.
+
+ The original debugger has detached from this process, and the
+ clone has attached to it.
+
+ On some targets, this requires a bit of cleanup to make it work
+ correctly.
+ */
+#define target_post_follow_inferior_by_clone() \
+ (*current_target.to_post_follow_inferior_by_clone) ()
+
+/* On some targets, we can catch an inferior fork or vfork event when it
+ occurs. These functions insert/remove an already-created catchpoint for
+ such events.
+ */
+#define target_insert_fork_catchpoint(pid) \
+ (*current_target.to_insert_fork_catchpoint) (pid)
+
+#define target_remove_fork_catchpoint(pid) \
+ (*current_target.to_remove_fork_catchpoint) (pid)
+
+#define target_insert_vfork_catchpoint(pid) \
+ (*current_target.to_insert_vfork_catchpoint) (pid)
+
+#define target_remove_vfork_catchpoint(pid) \
+ (*current_target.to_remove_vfork_catchpoint) (pid)
+
+/* Returns TRUE if PID has invoked the fork() system call. And,
+ also sets CHILD_PID to the process id of the other ("child")
+ inferior process that was created by that call.
+ */
+#define target_has_forked(pid,child_pid) \
+ (*current_target.to_has_forked) (pid,child_pid)
+
+/* Returns TRUE if PID has invoked the vfork() system call. And,
+ also sets CHILD_PID to the process id of the other ("child")
+ inferior process that was created by that call.
+ */
+#define target_has_vforked(pid,child_pid) \
+ (*current_target.to_has_vforked) (pid,child_pid)
+
+/* Some platforms (such as pre-10.20 HP-UX) don't allow us to do
+ anything to a vforked child before it subsequently calls exec().
+ On such platforms, we say that the debugger cannot "follow" the
+ child until it has vforked.
+
+ This function should be defined to return 1 by those targets
+ which can allow the debugger to immediately follow a vforked
+ child, and 0 if they cannot.
+ */
+#define target_can_follow_vfork_prior_to_exec() \
+ (*current_target.to_can_follow_vfork_prior_to_exec) ()
+
+/* An inferior process has been created via a vfork() system call.
+ The debugger has followed the parent, the child, or both. The
+ process of setting up for that follow may have required some
+ target-specific trickery to track the sequence of reported events.
+ If so, this function should be defined by those targets that
+ require the debugger to perform cleanup or initialization after
+ the vfork follow.
+ */
+#define target_post_follow_vfork(parent_pid,followed_parent,child_pid,followed_child) \
+ (*current_target.to_post_follow_vfork) (parent_pid,followed_parent,child_pid,followed_child)
+
+/* On some targets, we can catch an inferior exec event when it
+ occurs. These functions insert/remove an already-created catchpoint
+ for such events.
+ */
+#define target_insert_exec_catchpoint(pid) \
+ (*current_target.to_insert_exec_catchpoint) (pid)
+
+#define target_remove_exec_catchpoint(pid) \
+ (*current_target.to_remove_exec_catchpoint) (pid)
+
+/* Returns TRUE if PID has invoked a flavor of the exec() system call.
+ And, also sets EXECD_PATHNAME to the pathname of the executable file
+ that was passed to exec(), and is now being executed.
+ */
+#define target_has_execd(pid,execd_pathname) \
+ (*current_target.to_has_execd) (pid,execd_pathname)
+
+/* Returns the number of exec events that are reported when a process
+ invokes a flavor of the exec() system call on this target, if exec
+ events are being reported.
+ */
+#define target_reported_exec_events_per_exec_call() \
+ (*current_target.to_reported_exec_events_per_exec_call) ()
+
+/* Returns TRUE if PID has reported a syscall event. And, also sets
+ KIND to the appropriate TARGET_WAITKIND_, and sets SYSCALL_ID to
+ the unique integer ID of the syscall.
+ */
+#define target_has_syscall_event(pid,kind,syscall_id) \
+ (*current_target.to_has_syscall_event) (pid,kind,syscall_id)
+
+/* Returns TRUE if PID has exited. And, also sets EXIT_STATUS to the
+ exit code of PID, if any.
+ */
+#define target_has_exited(pid,wait_status,exit_status) \
+ (*current_target.to_has_exited) (pid,wait_status,exit_status)
+
+/* The debugger has completed a blocking wait() call. There is now
+ some process event that must be processed. This function should
+ be defined by those targets that require the debugger to perform
+ cleanup or internal state changes in response to the process event.
+ */
+
+/* The inferior process has died. Do what is right. */
+
+#define target_mourn_inferior() \
+ (*current_target.to_mourn_inferior) ()
+
+/* Does target have enough data to do a run or attach command? */
+
+#define target_can_run(t) \
+ ((t)->to_can_run) ()
+
+/* post process changes to signal handling in the inferior. */
+
+#define target_notice_signals(pid) \
+ (*current_target.to_notice_signals) (pid)
+
+/* Check to see if a thread is still alive. */
+
+#define target_thread_alive(pid) \
+ (*current_target.to_thread_alive) (pid)
+
+/* Make target stop in a continuable fashion. (For instance, under Unix, this
+ should act like SIGSTOP). This function is normally used by GUIs to
+ implement a stop button. */
+
+#define target_stop current_target.to_stop
+
+/* Queries the target side for some information. The first argument is a
+ letter specifying the type of the query, which is used to determine who
+ should process it. The second argument is a string that specifies which
+ information is desired and the third is a buffer that carries back the
+ response from the target side. The fourth parameter is the size of the
+ output buffer supplied. */
+
+#define target_query(query_type, query, resp_buffer, bufffer_size) \
+ (*current_target.to_query) (query_type, query, resp_buffer, bufffer_size)
+
+/* Get the symbol information for a breakpointable routine called when
+ an exception event occurs.
+ Intended mainly for C++, and for those
+ platforms/implementations where such a callback mechanism is available,
+ e.g. HP-UX with ANSI C++ (aCC). Some compilers (e.g. g++) support
+ different mechanisms for debugging exceptions. */
+
+#define target_enable_exception_callback(kind, enable) \
+ (*current_target.to_enable_exception_callback) (kind, enable)
+
+/* Get the current exception event kind -- throw or catch, etc. */
+
+#define target_get_current_exception_event() \
+ (*current_target.to_get_current_exception_event) ()
+
+/* Pointer to next target in the chain, e.g. a core file and an exec file. */
+
+#define target_next \
+ (current_target.to_next)
+
+/* Does the target include all of memory, or only part of it? This
+ determines whether we look up the target chain for other parts of
+ memory if this target can't satisfy a request. */
+
+#define target_has_all_memory \
+ (current_target.to_has_all_memory)
+
+/* Does the target include memory? (Dummy targets don't.) */
+
+#define target_has_memory \
+ (current_target.to_has_memory)
+
+/* Does the target have a stack? (Exec files don't, VxWorks doesn't, until
+ we start a process.) */
+
+#define target_has_stack \
+ (current_target.to_has_stack)
+
+/* Does the target have registers? (Exec files don't.) */
+
+#define target_has_registers \
+ (current_target.to_has_registers)
+
+/* Does the target have execution? Can we make it jump (through
+ hoops), or pop its stack a few times? FIXME: If this is to work that
+ way, it needs to check whether an inferior actually exists.
+ remote-udi.c and probably other targets can be the current target
+ when the inferior doesn't actually exist at the moment. Right now
+ this just tells us whether this target is *capable* of execution. */
+
+#define target_has_execution \
+ (current_target.to_has_execution)
+
+/* Can the target support the debugger control of thread execution?
+ a) Can it lock the thread scheduler?
+ b) Can it switch the currently running thread? */
+
+#define target_can_lock_scheduler \
+ (current_target.to_has_thread_control & tc_schedlock)
+
+#define target_can_switch_threads \
+ (current_target.to_has_thread_control & tc_switch)
+
+extern void target_link PARAMS ((char *, CORE_ADDR *));
+
+/* Converts a process id to a string. Usually, the string just contains
+ `process xyz', but on some systems it may contain
+ `process xyz thread abc'. */
+
+#ifndef target_pid_to_str
+#define target_pid_to_str(PID) \
+ normal_pid_to_str (PID)
+extern char *normal_pid_to_str PARAMS ((int pid));
+#endif
+
+#ifndef target_tid_to_str
+#define target_tid_to_str(PID) \
+ normal_pid_to_str (PID)
+extern char *normal_pid_to_str PARAMS ((int pid));
+#endif
+
+
+#ifndef target_new_objfile
+#define target_new_objfile(OBJFILE)
+#endif
+
+#ifndef target_pid_or_tid_to_str
+#define target_pid_or_tid_to_str(ID) \
+ normal_pid_to_str (ID)
+#endif
+
+/* Attempts to find the pathname of the executable file
+ that was run to create a specified process.
+
+ The process PID must be stopped when this operation is used.
+
+ If the executable file cannot be determined, NULL is returned.
+
+ Else, a pointer to a character string containing the pathname
+ is returned. This string should be copied into a buffer by
+ the client if the string will not be immediately used, or if
+ it must persist.
+ */
+
+#define target_pid_to_exec_file(pid) \
+ (current_target.to_pid_to_exec_file) (pid)
+
+/* Hook to call target-dependant code after reading in a new symbol table. */
+
+#ifndef TARGET_SYMFILE_POSTREAD
+#define TARGET_SYMFILE_POSTREAD(OBJFILE)
+#endif
+
+/* Hook to call target dependant code just after inferior target process has
+ started. */
+
+#ifndef TARGET_CREATE_INFERIOR_HOOK
+#define TARGET_CREATE_INFERIOR_HOOK(PID)
+#endif
+
+/* Hardware watchpoint interfaces. */
+
+/* Returns non-zero if we were stopped by a hardware watchpoint (memory read or
+ write). */
+
+#ifndef STOPPED_BY_WATCHPOINT
+#define STOPPED_BY_WATCHPOINT(w) 0
+#endif
+
+/* HP-UX supplies these operations, which respectively disable and enable
+ the memory page-protections that are used to implement hardware watchpoints
+ on that platform. See wait_for_inferior's use of these.
+ */
+#if !defined(TARGET_DISABLE_HW_WATCHPOINTS)
+#define TARGET_DISABLE_HW_WATCHPOINTS(pid)
+#endif
+
+#if !defined(TARGET_ENABLE_HW_WATCHPOINTS)
+#define TARGET_ENABLE_HW_WATCHPOINTS(pid)
+#endif
+
+/* Provide defaults for systems that don't support hardware watchpoints. */
+
+#ifndef TARGET_HAS_HARDWARE_WATCHPOINTS
+
+/* Returns non-zero if we can set a hardware watchpoint of type TYPE. TYPE is
+ one of bp_hardware_watchpoint, bp_read_watchpoint, bp_write_watchpoint, or
+ bp_hardware_breakpoint. CNT is the number of such watchpoints used so far
+ (including this one?). OTHERTYPE is who knows what... */
+
+#define TARGET_CAN_USE_HARDWARE_WATCHPOINT(TYPE,CNT,OTHERTYPE) 0
+
+#if !defined(TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT)
+#define TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT(byte_count) \
+ (LONGEST)(byte_count) <= REGISTER_SIZE
+#endif
+
+/* However, some addresses may not be profitable to use hardware to watch,
+ or may be difficult to understand when the addressed object is out of
+ scope, and hence should be unwatched. On some targets, this may have
+ severe performance penalties, such that we might as well use regular
+ watchpoints, and save (possibly precious) hardware watchpoints for other
+ locations.
+ */
+#if !defined(TARGET_RANGE_PROFITABLE_FOR_HW_WATCHPOINT)
+#define TARGET_RANGE_PROFITABLE_FOR_HW_WATCHPOINT(pid,start,len) 0
+#endif
+
+
+/* Set/clear a hardware watchpoint starting at ADDR, for LEN bytes. TYPE is 0
+ for write, 1 for read, and 2 for read/write accesses. Returns 0 for
+ success, non-zero for failure. */
+
+#define target_remove_watchpoint(ADDR,LEN,TYPE) -1
+#define target_insert_watchpoint(ADDR,LEN,TYPE) -1
+
+#endif /* TARGET_HAS_HARDWARE_WATCHPOINTS */
+
+#ifndef target_insert_hw_breakpoint
+#define target_remove_hw_breakpoint(ADDR,SHADOW) -1
+#define target_insert_hw_breakpoint(ADDR,SHADOW) -1
+#endif
+
+#ifndef target_stopped_data_address
+#define target_stopped_data_address() 0
+#endif
+
+/* If defined, then we need to decr pc by this much after a hardware break-
+ point. Presumably this overrides DECR_PC_AFTER_BREAK... */
+
+#ifndef DECR_PC_AFTER_HW_BREAK
+#define DECR_PC_AFTER_HW_BREAK 0
+#endif
+
+/* Sometimes gdb may pick up what appears to be a valid target address
+ from a minimal symbol, but the value really means, essentially,
+ "This is an index into a table which is populated when the inferior
+ is run. Therefore, do not attempt to use this as a PC."
+ */
+#if !defined(PC_REQUIRES_RUN_BEFORE_USE)
+#define PC_REQUIRES_RUN_BEFORE_USE(pc) (0)
+#endif
+
+/* This will only be defined by a target that supports catching vfork events,
+ such as HP-UX.
+
+ On some targets (such as HP-UX 10.20 and earlier), resuming a newly vforked
+ child process after it has exec'd, causes the parent process to resume as
+ well. To prevent the parent from running spontaneously, such targets should
+ define this to a function that prevents that from happening.
+ */
+#if !defined(ENSURE_VFORKING_PARENT_REMAINS_STOPPED)
+#define ENSURE_VFORKING_PARENT_REMAINS_STOPPED(PID) (0)
+#endif
+
+/* This will only be defined by a target that supports catching vfork events,
+ such as HP-UX.
+
+ On some targets (such as HP-UX 10.20 and earlier), a newly vforked child
+ process must be resumed when it delivers its exec event, before the parent
+ vfork event will be delivered to us.
+ */
+#if !defined(RESUME_EXECD_VFORKING_CHILD_TO_GET_PARENT_VFORK)
+#define RESUME_EXECD_VFORKING_CHILD_TO_GET_PARENT_VFORK() (0)
+#endif
+
+/* Routines for maintenance of the target structures...
+
+ add_target: Add a target to the list of all possible targets.
+
+ push_target: Make this target the top of the stack of currently used
+ targets, within its particular stratum of the stack. Result
+ is 0 if now atop the stack, nonzero if not on top (maybe
+ should warn user).
+
+ unpush_target: Remove this from the stack of currently used targets,
+ no matter where it is on the list. Returns 0 if no
+ change, 1 if removed from stack.
+
+ pop_target: Remove the top thing on the stack of current targets. */
+
+extern void
+add_target PARAMS ((struct target_ops *));
+
+extern int
+push_target PARAMS ((struct target_ops *));
+
+extern int
+unpush_target PARAMS ((struct target_ops *));
+
+extern void
+target_preopen PARAMS ((int));
+
+extern void
+pop_target PARAMS ((void));
+
+/* Struct section_table maps address ranges to file sections. It is
+ mostly used with BFD files, but can be used without (e.g. for handling
+ raw disks, or files not in formats handled by BFD). */
+
+struct section_table {
+ CORE_ADDR addr; /* Lowest address in section */
+ CORE_ADDR endaddr; /* 1+highest address in section */
+
+ sec_ptr the_bfd_section;
+
+ bfd *bfd; /* BFD file pointer */
+};
+
+/* Builds a section table, given args BFD, SECTABLE_PTR, SECEND_PTR.
+ Returns 0 if OK, 1 on error. */
+
+extern int
+build_section_table PARAMS ((bfd *, struct section_table **,
+ struct section_table **));
+
+/* From mem-break.c */
+
+extern int memory_remove_breakpoint PARAMS ((CORE_ADDR, char *));
+
+extern int memory_insert_breakpoint PARAMS ((CORE_ADDR, char *));
+
+extern breakpoint_from_pc_fn memory_breakpoint_from_pc;
+#ifndef BREAKPOINT_FROM_PC
+#define BREAKPOINT_FROM_PC(pcptr, lenptr) memory_breakpoint_from_pc (pcptr, lenptr)
+#endif
+
+
+/* From target.c */
+
+extern void
+initialize_targets PARAMS ((void));
+
+extern void
+noprocess PARAMS ((void));
+
+extern void
+find_default_attach PARAMS ((char *, int));
+
+void
+find_default_require_attach PARAMS ((char *, int));
+
+void
+find_default_require_detach PARAMS ((int, char *, int));
+
+extern void
+find_default_create_inferior PARAMS ((char *, char *, char **));
+
+void
+find_default_clone_and_follow_inferior PARAMS ((int, int *));
+
+extern struct target_ops *
+find_core_target PARAMS ((void));
+
+/* Stuff that should be shared among the various remote targets. */
+
+/* Debugging level. 0 is off, and non-zero values mean to print some debug
+ information (higher values, more information). */
+extern int remote_debug;
+
+/* Speed in bits per second, or -1 which means don't mess with the speed. */
+extern int baud_rate;
+/* Timeout limit for response from target. */
+extern int remote_timeout;
+
+extern asection *target_memory_bfd_section;
+
+/* Functions for helping to write a native target. */
+
+/* This is for native targets which use a unix/POSIX-style waitstatus. */
+extern void store_waitstatus PARAMS ((struct target_waitstatus *, int));
+
+/* Convert between host signal numbers and enum target_signal's. */
+extern enum target_signal target_signal_from_host PARAMS ((int));
+extern int target_signal_to_host PARAMS ((enum target_signal));
+
+/* Convert from a number used in a GDB command to an enum target_signal. */
+extern enum target_signal target_signal_from_command PARAMS ((int));
+
+/* Any target can call this to switch to remote protocol (in remote.c). */
+extern void push_remote_target PARAMS ((char *name, int from_tty));
+
+/* Imported from machine dependent code */
+
+#ifndef SOFTWARE_SINGLE_STEP_P
+#define SOFTWARE_SINGLE_STEP_P 0
+#define SOFTWARE_SINGLE_STEP(sig,bp_p) abort ()
+#endif /* SOFTWARE_SINGLE_STEP_P */
+
+/* Blank target vector entries are initialized to target_ignore. */
+void target_ignore PARAMS ((void));
+
+/* Macro for getting target's idea of a frame pointer.
+ FIXME: GDB's whole scheme for dealing with "frames" and
+ "frame pointers" needs a serious shakedown. */
+#ifndef TARGET_VIRTUAL_FRAME_POINTER
+#define TARGET_VIRTUAL_FRAME_POINTER(ADDR, REGP, OFFP) \
+ do { *(REGP) = FP_REGNUM; *(OFFP) = 0; } while (0)
+#endif /* TARGET_VIRTUAL_FRAME_POINTER */
+
+#endif /* !defined (TARGET_H) */