* gdbarch.sh: Add FP0_REGNUM to list of multi-arched register numbers.
* gdbarch.c, gdbarch.h: Regenerate.
* core-regset.c: Change FP0_REGNUM to a runtime test.
* procfs.c: Ditto.
* sun3-nat.c: Ditto.
* sparc-tdep.c: Ditto.
* i386mach-nat.c: Remove unnecessary ifdef for FP0_REGNUM.
* ptx4-nat.c: Ditto.
* sol-thread.c (sol_thread_create_inferior): only add the thread
to the thread list if it is not already in there.
(prototypes for thread_db func ptrs): pretty up formatting.
* procfs.c: Cleanup of procfs tracing. Move defines and
prototypes to proc-utils.h
* proc-utils.h: Define tracing macros. Declare trace functions.
* proc-api.c: Make procfs tracing a runtime option.
(prepare_to_trace): New function, abstracted out of several
places. Open a trace file if one is required.
(ioctl_with_trace, write_with_trace, open_with_trace,
close_with_trace, wait_with_trace, lseek_with_trace):
Report errno if an error occurs in a system call.
(write_with_trace): Make 2nd arg void *, to agree with write.
* procfs.c (many functions): change int cmd; to long cmd;
Solaris /proc API calls this parameter a long, and requires it
for 64-bit hosts (where sizeof(long) != sizeof(int)). UnixWare
calls it an int, but on existing UnixWare hosts, int and long
are the same size. So long it must be. If a future UnixWare
version has problems with this, we'll have to use autoconf.
* proc-api.c (write_with_trace): use long instead of int.
* procfs.c (comments): Eliminate "???" in comments, which GCC
wants to interpret as a trigraph. (fill_gregset, supply_gregset,
fill_fpregset, supply_fpregset): declare.
(procfs_wait): Sysargs is a long, change printf format to match.
(test-mapping, mapping_test, test_mapping_cmd) Remove (test only).
init only nonzero fields, leave to_require_attach and
to_require_detach empty, not needed for /proc systems yet.
(_initialize_procfs): Call init_procfs_ops.
* infrun.c (set/show scheduler-locking) New command. Set a
mode bit that will control how GDB attempts to control thread
scheduling for step, continue, etc. (resume): make use of
the schedule-locking mode.
* target.h (struct target_ops): new field to_has_thread_control.
* sol-thread.c: initialize target_ops to_has_thread_control.
* procfs.c: ditto.
* target.c: ditto.
* m3-nat.c: ditto.
* remote.c: ditto.
* hpux-thread.c: ditto.
* procfs.c: Added replacement macros for LWP stuff. Fixed support
for UnixWare / SVR4.2MP targets and any targets which use
multi-file /proc entries. Fixed support for hardware watchpoints.
* solib.c: SCO needs some of the same code as SunOS. Change
preprocessor conditionals.
* config/i386/i386sco5.mt: New file.
* config/i386/tm-i386sco5.h: New file.
* config/i386/i386sco5.mh (NATDEPFILES): add i386v-nat.o.
* config/i386/nm-i386v42mp.h
(TARGET_HAS_HARDWARE_WATCHPOINTS): define.
Add other macros for hardware assisted watchpoints.
* config/i386/nm-i386sco5.h: Correct attributions.
(TARGET_HAS_HARDWARE_WATCHPOINTS): define.
* config/i386/nm-linux.h (target_remote_watchpoint): Pass
'type' through to i386_insert_watchpoint.
Mon Apr 20 14:12:30 1998 Peter Schauer (pes@regent.e-technik.tu-muenchen.de)
* infrun.c (wait_for_inferior): Don't add signalled processes
as new threads.
* procfs.c (wait_fd): Note if LWP has exited.
(procfs_wait): use GETPID to get process ID.
NOTE: I didn't commit Kean's changes to configure.host/configure.tgt
yet because they are kind of bogus and I'm punting back to him to let
him figure out what he's trying to do. So configuring for UnixWare
may not work just yet.
Update support for x86 Solaris 2.
* config/i386/tm-i386sol2.h, nm-i386sol2.h: New configuration
files for x86 Solaris 2.
* config/i386/i386sol2.mt, i386sol2.mh: Use them.
* config/sparc/tm-sun4sol2.h (PROCFS_GET_CARRY): New macro, extract
carry flag from a given regset.
(IS_STATIC_TRANSFORM_NAME): New macro, check if a symbol name
is a SunPro transformed name.
* i386-tdep.c (sunpro_static_transform_name): New function to
extract the source name from a SunPro transformed name.
* inferior.h (procfs_first_available, procfs_get_pid_fd):
Add prototypes.
* infrun.c (wait_for_inferior): Handle breakpoint hit in
signal handler without intervening stop in sigtramp.
* procfs.c (procfs_lwp_creation_handler): Use PROCFS_GET_CARRY
instead of direct access to the status register.
(procfs_get_pid_fd): New function, returns procfs fd for a given pid.
* sol-thread.c (ps_lgetLDT): New function, returns LDT for a given
lwpid.
(sol_find_new_threads): Handle failed libthread_db initialization
gracefully.
* stabsread.c (define_symbol): Use IS_STATIC_TRANSFORM_NAME
to check for a SunPro transformed symbol name.
unconditionally_kill_inferior): Undo Oct 26 1996 and Apr 26 1997
changes to trace PRFS_STOPTERM and handle PR_DEAD.
These changes tried to work around a problem with an early DU 4.0
release, but they trigger subtle timing dependent kernel bugs
in older OSF/1 releases.
* procfs.c (init_procinfo): new function, abstracts some code
shared by create_procinfo and do_attach;
(procfs_set_inferior_syscall_traps): new function, abstracts
some code needed by procfs_init_inferior, do_attach, and
procfs_lwp_creation_handler; (procfs_first_available): new
function, find any LWP that's runnable; (procfs_thread_alive):
replace stub function with real implementation;
(procfs_lwp_creation_handler): fix bug starting new child
threads; (info_proc): bug fixes and enhancements for the
"INFO PROCESSES" command; (close_procinfo_file): call new
function "delete_thread" to cleanup GDB's thread database;
(proc_init_failed): add new argument "kill", to control whether
process is killed (so this function can be shared by
create_procinfo and do_attach); (procfs_exit_handler): handle
exit from an attached process, and cleanup procinfo handles
when the process exits; (procfs_resume, procfs_wait): cleanup
after a thread when it exits; (do_attach, do_detach): handle
attached processes with multiple threads; plus some general
improvements in the diagnostic output.
* sol-thread.c (sol_thread_alive): replace stub with real
implementation; (thread_to_lwp, lwp_to_thread): enhance to
handle threads that may have exited; (sol_thread_attach): add
startup setup stuff; (sol_thread_detach): add unpush_target
call; (sol_thread_mourn_inferior): add unpush_target call;
(sol_thread_wait, sol_thread_resume): enhance to deal with
thread exit cleanly; (sol_thread_new_objfile,
sol_thread_pid_to_str): detect unsuccessful startup and
don't crash; plus some general cleanup.
* thread.c (delete_thread): new function, allows targets to
notify gdb when a thread is no longer valid.
* infrun.c (wait_for_inferior): don't try to detect a new
thread on receiving a TARGET_EXITED event.
* procfs.c: Minor changes to make pre-ANSI compilers happy.
(procfs_notice_signals): Copy traced signal set back to
pi->prrun.pr_trace.
(unconditionally_kill_inferior): If PR_DEAD is defined,
rerun inferior after killing it.
* procfs.c: Substantial (but incomplete) changes to support
sysv4.2mp procfs as implemented in UnixWare 2.1. The procinfo
struct now has substructs like struct flt_ctl instead of
just a fltset_t and has a ctl_fd, status_fd, as_fd, and
map_fd instead of a single fd. Non-sysv4.2mp procfs models
still have the structs and multiple fds, but don't use the
entire struct and the four fds all point to the same thing.
We use PROCFS_USE_READ_WRITE to decide whether to talk to
procfs with reads/writes or use ioctl instead. We use
HAVE_MULTIPLE_PROC_FDS to determine whether procfs really has
multiple fds or not. PROC_NAME_FMT is split out into
CTL_PROC_NAME_FMT, AS_PROC_NAME_FMT, MAP_PROC_NAME_FMT,
STATUS_PROC_NAME_FMT.
(procfs_notice_signals): now a necessary wrapper around
(notice_signals): which are the new guts for noticing signals
(open_proc_file): gets a new flag arg used in sysv4.2mp to
determine whether or not to attempt to open the ctl_fd.
(procfs_read_status): new local function, reads procfs status
(procfs_write_pcwstop): new local function, writes a PCWSTOP
(procfs_write_pckill): new local function, writes a PCKILL
(unconditionally_kill_inferior): remove signo since we now
just call procfs_write_pckill().
(procfs_xfer_memory): call lseek with SEEK_SET rather than 0
(proc_iterate_over_mappings): the whole function is ifdefed
on UNIXWARE to keep things readable.
Expanded the syscall_table to include new potential sysv4.2mp
members. Note that all ifdefs of UNIXWARE should be eliminated
if possible or renamed to describe what's being selected for a
bit better. Sysv4.2mp and IRIX both have SYS_sproc so the
IRIX specific code now also checks it's not UNIXWARE.
* config/i386/tm-i386v42mp.h: also define HAVE_PSTATUS_T,
HAVE_NO_PRRUN_T, PROCFS_USE_READ_WRITE, and UNIXWARE
* config/mips/nm-irix4.h: set CTL_PROC_NAME_FMT et al to
"/debug/%d" as PROC_NAME_FMT used to be
* procfs.c (wait_fd): Handle EINTR error return from poll
by restarting the poll.
* defs.h (PIDGET): Define a default version that just
returns its argument unchanged.
* inflow.c (terminal_init_inferior): Eliminate #ifdef
of PIDGET and fold both alternatives into common code.
(pass_signal): Use PIDGET for pid passed to kill().
the entry to the exit system call to detect termination of the
inferior stopped working. Trace termination of the inferior via
PRFS_STOPTERM instead.
(procfs_init_inferior): Do not trace entry to exit system call
if PIOCSSPCACT is defined.
(procfs_wait): Handle PR_DEAD event, which signals the termination
of the inferior if PRFS_STOPTERM is set.
* mdebugread.c (parse_partial_symbols): Ignore stNil section
start address symbols.
* sparc-tdep.c (get_saved_register): Get saved PC from the
frame info if not in innermost frame.
* (procfs_pid_to_str): New routine to print out thread id's in an
intelligible manner.
* sol-thread.c (sol_thread_fetch_registers): Re-order manner in
which supply_register is called to fix bug with writing writing
individual regs.
* config/sparc/tm-sun4sol2.h: Define default for
target_pid_to_str in case host lacks libthread_db.
config/sparc/nm-sun4sol2.h config/sparc/sun4sol2.mh
config/sparc/tm-sun4sol2.h: Use autoconf to config Solaris thread
and pthread support, since pre-2.5 systems don't come with
libthread_db.so.1.
* procfs.c (info_proc): Use int instead of id_t. Old versions of
Irix don't see to define this.
infrun.c, mac-nat.c, procfs.c, remote.c, sol-thread.c, thread.c,
win32-nat.c, config/nm-lynx.h: Rename thread.h to gdbthread.h to
avoid conflict with Solaris /usr/include/thread.h.
returns a possibly modified pid.
* inftarg.c (ptrace_him): Now returns pid;
* m3-nat.c (m3_trace_him): Now returns pid;
* infcmd.c (run_command): Minor cleanup.
* infrun.c (wait_for_inferior): Add another check for one_stepped
near where we read the pc to avoid erroneously setting
random_signal for multi-threaded support.
* procfs.c: Add support for Solaris LWPs. Remove def of
LOSING_POLL. Many cleanups... Several workarounds for Solaris
lossage. System call entry and exit are now handled by
dynamically registered handlers.
* (syscallname): Don't barf when handed an unknown syscall
number.
* (info_proc_syscalls): Ditto.
* sol-thread.c: New file. Implements Solaris thread support.
* symfile.c (symbol_file_add): Add call to target_new_objfile to
notify target-dependent code about new symbol tables.
* (clear_symtab_users): Call target_new_objfile to notify it of
the removal of all symbol tables.
* target.c (push_target): Make sure that to_close is non-zero
before calling it.
* target.h (target_new_objfile): Provide default.
* config/alpha/nm-osf2.h: Define LOSING_POLL because this version
of OSF can't hack using poll with /proc.
* config/sparc/nm-sun4sol2.h (target_new_objfile): Define to be
sol-thread-new-objfile.
* config/sparc/sun4sol2.mh: Add sol-thread.o to NATDEFFILES, and
add libthread_db.so.1 to NAT_CLIBS.
* config/sparc/tm-sun4sol2.h: Define PIDGET, TIDGET, and
target_pid_to_str.
AT_name tag is missing.
* procfs.c (procfs_stopped_by_watchpoint): Fix logic when
FLTWATCH and FLTKWATCH are defined.
* remote.c (remote_read_bytes): Advance memaddr for transfers,
return number of bytes transferred for partial reads.
* top.c (init_signals): Reset SIGTRAP to SIG_DFL.
Andrew Cagney