an error, regardless of force_ellipsis. In the non-EIO case,
just print the error message rather than calling error(). Don't
access *(bufptr-1) if bufptr points to the start of the buffer.
When looking for '\0', don't increment bufptr and addr if bufptr
started out already at limit. If an error happens on fetching the
first character, don't print the string.
FRAME_CHAIN, FRAMELESS_FUNCTION_INVOCATION, FRAME_SAVED_PC):
Define to make backtracing through the various sigtramp handlers
work.
* i386-tdep.c (i386v4_sigtramp_saved_pc): New routine to fetch
the saved pc from ucontext on the stack for SVR4 signal handling.
target.h, remote-es.c, remote-nindy.c: Don't set baud rate if
baud_rate is -1. Remove sr_get_baud_rate and sr_set_baud_rate;
just use the global variable itself. When printing baud rate,
don't print a baud rate if baud_rate is -1.
next over recursive call. Revise the tests from that point on
(which had been skipped) to match GDB's actual (correct) output.
* gdb.t07/watchpoint.exp (test_stepping): Remove xfail for bug
with next over a breakpoint whose condition is false.
(bpstat_what_main_action): New code BPSTAT_WHAT_THROUGH_SIGTRAMP.
* breakpoint.c (bpstat_what): Return BPSTAT_WHAT_THROUGH_SIGTRAMP
if we hit a bp_through_sigtramp breakpoint. Remove kludge which
ignored bs->stop for a bp_step_resume breakpoint.
* infrun.c (wait_for_inferior): Make a through_sigtramp_breakpoint
which performs one (the check_sigtramp2 one) of the functions
which had been handled by the step_resume_breakpoint. For each
use of the step_resume_breakpoint, make it still use the
step_resume_breakpoint, use the through_sigtramp_breakpoint, or
operate on both.
Deal with BPSTAT_WHAT_THROUGH_SIGTRAMP return from bpstat_what.
When setting the frame address of the step resume breakpoint, set
it to the address for frame *before* the call instruction is
executed, not after.
print_scalar_formatted rather than duplicating all the
CC_HAS_LONG_LONG and so on.
(mips_push_dummy_frame): Use read_register_gen rather than using
read_register and then putting it back in target format with
store_unsigned_integer. If registers are more than 4 bytes, give
an error rather than have some registers overwrite other
registers.
#if 0 unused include of opcode/mips.h.
* salib.c: Use macros instead of constants for I/O addresses to
make 931 support easier.
* sparclite.h: Change constraint for LOC to "rJ" to force use of
register in sta/lda instructions.
well as for ELF and SOM. Unify the descriptions of ELF and SOM
as "stabs in sections" rather than just saying "ELF and SOM".
Also make that stuff apply to COFF.
error() on TYPE_CODE_FLT arguments whose size is greater than 8,
swap all other TYPE_CODE_FLT arguments as mips_push_arguments
ensures that floats are promoted to doubles before they are pushed
on the stack.
them in the pst.
* mdebugread.c (handle_psymbol_enumerators): New function to enter
the enumerators of an ecoff enum into the partial symbol table.
* mdebugread.c (parse_partial_symbols): Call it.
* symfile.c (reread_symbols): Initialize objfile->*_psymbols.next.
* symmisc.c (dump_psymtab): Fix typo, clean up output of section
offsets. Cast psymtab->read_symtab to PTR before passing it to
gdb_print_address.
* i386-tdep.c (i386_skip_prologue): Skip over instructions that
set up the global offset table pointer in pic compiled code.
* config/mips/tm-mips.h (FIX_CALL_DUMMY): For big endian targets,
error() on TYPE_CODE_FLT arguments whose size is greater than 8,
swap all other TYPE_CODE_FLT arguments as mips_push_arguments
ensures that floats are promoted to doubles before they are pushed
on the stack.
Fixup cache_on and flush_i_cache so that they work for both the
930 and 932 processors. Rewrite most low level funcs (uart
access & cache stuff) to use new ASI access macros in sparclite.h.
Also make it easy to access second serial port.
ourselves instead of worrying about INSTALL_XFORM. This enables
users to override INSTALL_PROGRAM in the standard way.
* Makefile.in (c-exp.tab.o, ch-exp.tab.o, m2-exp.tab.o): Don't
depend on Makefile.in.
gone now.
* c-exp.y, ch-exp.y, core.c, corelow.c, eval.c, fork-child.c,
m2-exp.y, minsyms.c, nlmread.c, parse.c, putenv.c, regex.c
remote-utils.c, stabsread.c: Include <string.h>.
* regex.c: Include "defs.h", change re_comp argument to const char *.
* infptrace.c (fetch_register, store_inferior_registers): Change
regaddr to type CORE_ADDR.
* config/alpha/alpha-nw.mt, config/alpha/alpha-osf1.mt (MT_CFLAGS):
Remove, no longer necessary now that we use bfd_vma for a CORE_ADDR.
are fixed.
* gdb.t02/whatis.exp: Remove xfails for v_char_array and
v_signed_char_array; the test is looser than when the xfails were
added.
* gdb.t00/{help.exp,default.exp}: Don't test "target remote" on UDI.
* gdb.t00/help.exp: Don't expect remote to be in "help target" output.
printing the program, rather than passing NULL to printf.
(udi_detach): Set udi_session_id to -1 so that udi_close doesn't
try to call UDIDisconnect again. Print better message.
(udi_kill): Just call UDIDisconnect ourselves, rather than doing
it via udi_close.
(udi_open): Unpush UDI target after target_preopen, now that
udi_kill doesn't pop the target.
(udi_create_inferior): If udi_session_id is negative, open a new
TIP rather than giving an error.
config/ns32k/ns32km3.mh: Define NAT_FILE.
* config/nm-m3.h: Change guard from _OS_MACH3_H_ and _OS_MACH3_H
(it was inconsistent and namespace-wrong) to NM_M3_H.
* m3-nat.c (mach_really_wait): Change parameter name to ourstatus.
(m3_open): New function.
(m3_ops): Use it.
* TODO: Update Mach section.
* Makefile.in: Remove "rapp" stuff; it is superseded by gdbserver.
New functions.
(mips_store_word): Change calling convention to return errors, and
to provide old contents if the caller wants it.
(mips_xfer_memory): Deal with errors from mips_store_word.
* config/mips/tm-idt.h, config/mips/tm-idtl.h: Remove BREAKPOINT
define now that remote-mips.c doesn't use BREAKPOINT.
properly quoting regexps. This makes the second argument to
demangle a fixed string, which matches the callers, instead of a
shell-style pattern which is what the previous implementation
used. Update some of the callers.
declaration because some compilers don't like arguments whose
names are the same as types.
* remote.c (remote_xfer_memory): Cast "myaddr" to unsigned char *
before passing it to remote_*_bytes.
* h8500-tdep.c (saved_pc_after_call): The size of the
pc is memory model dependent. (segmented_command,
unsegmented_command, _initialize_h8500_tdep): New commands to
change memory model.
* remote-7000.c (initialize_remote_e7000): Change name of snoop
command.
* remote-hms.c (hms_load): Remove breakpoints when loaded.
(hms_wait): Use new status structure
(hms_open): Push the target here. (hms_before_main_loop): Not
here. (supply_val, hms_fetch_register, hms_store_register): Cope
with H8/500 names too.
* sh-tdep.c (show_regs, initialize_sh_tdep): New command to print
all registers in a compact way.
only in the names (I probably got most of them). I was even
generous with some which differed in underscores in names, and
probably missed a few, and I *still* was able to remove almost a
third of the file with almost no impact on testsuite coverage.
versus functions without line numbers, also mention stepping into
them as well as "step" when you are in them. Tell the user how to
deal with the situation. Add comment about "debugging information".
(add_set_cmd): Use it instead of not_just_help_class_command.
(not_just_help_class_command): Change calling convention back to
what it was before yesterday's change.
gdb.t03/{ptype.exp,gdbme.c} from gdb.stabs/{weird.def,weird.exp}.
Reverts the change of 26 May 93. Per today's change to
stabs.texinfo, the behavior that weird.exp was expecting was
unreasonable; what is important is that the compiler+gdb get
things right, which happens with a recent gcc. Also fix the test
to deal with native compilers which put out the stab gdb can't
deal with.
an address to 'w'. Make it 'g' or 'h', depending on TARGET_PTR_BIT.
* defs.h: Just typedef CORE_ADDR to bfd_vma. Include bfd.h.
If BFD64, make a LONGEST a BFD_HOST_64_BIT.
* defs.h (longest_to_int): Don't depend on CC_HAS_LONG_LONG; instead
always just check against INT_MIN and INT_MAX (this also fixes things
if sizeof (long) > sizeof (int), e.g. Alpha).
* config/pa/tm-hppa.h, config/i386/sun386.h, config/rs6000/tm-rs6000.h:
Don't define LONGEST or BUILTIN_TYPE_LONGEST.
* gdbtypes.h: Remove BUILTIN_TYPE_LONGEST and
BUILTIN_TYPE_UNSIGNED_LONGEST.
* language.h, c-lang.c, ch-lang.c, m2-lang.c: Remove longest_int
and longest_unsigned_int.
* value.h (struct value): Just align to LONGEST, rather than worrying
about CC_HAS_LONG_LONG.
* valarith.c (value_binop, _initialize_valarith): Figure out type
ourself based on sizeof (LONGEST) rather than relying on
BUILTIN_TYPE_LONGEST. The point is that we don't depend on
CC_HAS_LONG_LONG anymore.
* valprint.c (val_print_type_code_int): Just call
extract_unsigned_integer directly, rather than going through
unpack_long.
* printcmd.c (decode_format): Remove code which would sometimes
change 'g' size to 'w' for integers. print_scalar_formatted handles
printing huge integers well enough, thank you.
to make this the sfunc, not cfunc, since that is how we call it.
* command.h: Comment difference between sfunc and cfunc.
* demangle.c (set_demangling_command): Add third arg since that
is how it is called.
(_initialize_demangler): Use sfunc, not cfunc, for
set_demangling_command, since that is how it is called.
Remove show_demangling_command; it has no effect.
Change comments regarding TYPE_CODE_BOOL.
* language.c (boolean_type): Always return 1 for TYPE_CODE_BOOL,
regardless of the language.
(value_true): Just call value_logical_not regardless of language.
* coffread.c (coff_read_enum_type), stabsread.c (read_enum_type):
Remove #if 0'd code which makes some enums TYPE_CODE_BOOL.
* language.h: Improve comment for la_builtin_type_vector.
* m2-lang.c (_initialize_m2_language): Don't add any fields to
builtin_type_m2_bool.
init_extra_frame_info and put it in new function mips_find_saved_regs.
(READ_FRAME_REG): Remove macro and replace uses with the expansion.
* mips-tdep.c, config/mips/tm-mips.h: When examining ->saved_regs,
check if it is NULL and call mips_find_saved_regs if so.
alpha and irix. The _sigtramp case has to be handled properly
in the tdep files if we have no ecoff debugging info.
* alpha-tdep.c (alpha_frame_saved_pc, alpha_frame_chain),
mips-tdep.c (mips_frame_saved_pc): Handle signal handler frames
without PC_REGNUM kludge.
* mdebugread.c (fixup_sigtramp), mips-tdep.c (read_next_frame_reg):
Clean up handling of mips sigtramp frames, improve comments.
sparc-sun-solaris2.*, these tests now pass with latest gcc and gdb.
* gdb.t20/inherit.exp (test_ptype_si): Update warning message
about known bug in gcc to include up through 2.5.8.
* gdb.t20/inherit.exp (test_ptype_vi): Accept "_vb." as well as
well as "_vb$".
* gdb.t21/cplusfuncs.exp (test_lookup_operator_functions):
Remove setup_xfails for sparc-sun-solaris2.*, these tests no pass
with latest gcc and gdb.
* gdb.t22/virtfunc.exp (test_ptype_of_classes): Accept "_vb." as
well as "_vb$".
* paread.c (pa_symtab_read): Update the "check_strange_names"
filter to match GCC's current output. Filter out section symbols
(which the HP linker sometimes puts in the wrong place).
* Makefile.in: Remove all references to sparcly-nat.c.
* Makefile.in (HFILES_NO_SRCDIR): Include dcache.h remote-utils.h
remote-sim.h directly, rather than via $(remote_utils_h). This avoids
duplicating serial.h and target.h.
* Makefile.in: Don't set M_INSTALL and M_UNINSTALL. These variables
are not used anywhere (a 5 Oct 1993 change removed the uses).
If the timeout is too big to fit in c_cc[VTIME], then do multiple reads
to achieve the desired timeout.
* serial.h (serial_t): Add field timeout_remaining.
with the old tm-m68k-em.h:
(GDBINIT_FILENAME, DEFAULT_PROMPT): Remove.
(HAVE_68881): Don't undefine; this #define is obsolete.
(REGISTER_NAMES): Don't muck with it; what tm-m68k.h has is fine.
Add FIXME regarding GET_LONGJMP_TARGET.
* remote-utils.c (usage): Fix message to be accurate and conform
more closely to normal conventions.
* remote-utils.c (gr_files_info): Have the exec_bfd test control
whether to show information about exec_bfd, and not control whether
to show information about device and speed.
* remote-utils.c (gr_open): If sr_get_device returns NULL, give
usage message, don't dump core.
* remote-bug.c (bug_write_memory): Use alloca, not GCC extension
for variable size array.
(bug_fetch_register, bug_store_register): Rename "value" to
"fpreg_buf" because some compilers don't like variables whose
names are the same as types.
(bug_store_register): Use a cast when converting char * to
unsigned char *.
full details on structure elements without names. This partially
reverts the changes of 1 Jul 1993 and 31 Aug 1993; I think this aspect
of those changes was accidental.
various miscellaneous things discovered in the process:
* printcmd.c, defs.h (print_address_numeric): New function.
* c-valprint.c (c_val_print), ch-valprint.c (chill_val_print)
breakpoint.c (describe_other_breakpoints, breakpoint_1, mention),
cp-valprint.c (cplus_print_value), infcmd.c (jump_command),
printcmd.c, stack.c, symfile.c, symmisc.c, valprint.c:
Use it.
* utils.c, defs.h (gdb_print_address): New function.
* expprint (dump_expression), gdbtypes.h: Use it.
* breakpoint.c (describe_other_breakpoints),
symmisc.c (dump_symtab, print_symbol):
Use filtered not unfiltered I/O.
(remove_breakpoints): Remove BREAKPOINT_DEBUG code. Might as well
just run gdb under a debugger for this (and it had problems with
printing addresses, how to print b->shadow, etc.).
* buildsym.c (make_blockvector), core.c (memory_error),
exec.c (print_section_info), maint.c (print_section_table),
mdebugread.c (parse_procedure), solib.c, source.c, symfile.c,
symmisc.c, symtab.c, valops.c, valprint.c, xcoffexec.c:
Add comments saying code is broken. Marked with "FIXME-32x64".
* dbxread.c (process_one_symbol), partial-stab.h (default),
remote-vx.c (vx_run_files_info):
Don't cast int being passed to local_hex_string.
* symmisc.c (print_symbol): Don't cast long being passed to %lx.
* symtab.h (general_symbol_info): Add comment about SYMBOL_VALUE
only being a long.
* symmisc.c (print_symbol): Print "offset" in message for LOC_ARG
and LOC_LOCAL.
* printcmd.c (print_address): Remove #if 0 code with ADDR_BITS_REMOVE.
* source.c: Include <sys/types.h> regardless of USG.
various miscellaneous things discovered in the process:
* printcmd.c, defs.h (print_address_numeric): New function.
* c-valprint.c (c_val_print), ch-valprint.c (chill_val_print)
breakpoint.c (describe_other_breakpoints, breakpoint_1, mention),
cp-valprint.c (cplus_print_value), infcmd.c (jump_command),
printcmd.c, stack.c, symfile.c, symmisc.c, valprint.c:
Use it.
* utils.c, defs.h (gdb_print_address): New function.
* expprint (dump_expression), gdbtypes.h: Use it.
* breakpoint.c (describe_other_breakpoints),
symmisc.c (dump_symtab, print_symbol):
Use filtered not unfiltered I/O.
(remove_breakpoints): Remove BREAKPOINT_DEBUG code. Might as well
just run gdb under a debugger for this (and it had problems with
printing addresses, how to print b->shadow, etc.).
* buildsym.c (make_blockvector), core.c (memory_error),
exec.c (print_section_info), maint.c (print_section_table),
mdebugread.c (parse_procedure), solib.c, source.c, symfile.c,
symmisc.c, symtab.c, valops.c, valprint.c, xcoffexec.c:
Add comments saying code is broken. Marked with "FIXME-32x64".
* dbxread.c (process_one_symbol), partial-stab.h (default),
remote-vx.c (vx_run_files_info):
Don't cast int being passed to local_hex_string.
* symmisc.c (print_symbol): Don't cast long being passed to %lx.
* symtab.h (general_symbol_info): Add comment about SYMBOL_VALUE
only being a long.
* symmisc.c (print_symbol): Print "offset" in message for LOC_ARG
and LOC_LOCAL.
* printcmd.c (print_address): Remove #if 0 code with ADDR_BITS_REMOVE.
* source.c: Include <sys/types.h> regardless of USG.
not OBJ_UNWIND_INFO. Assigning to a cast is a GCC-ism which
the HP compiler in ANSI mode doesn't like.
* main.c: When defaulting HAVE_SIGSETMASK based on USG, just do it
based on USG, rather than defining HAVE_SIGSETMASK to an expression
containing defined. This is for the HP compiler in ANSI mode.
passing them to print_387_status.
(print_387_status): Don't subtract top from 7 before using it.
* i387-tdep.c: Remove comment about AIX wanting "top" subtracted
from 7; the above explains it.
cleanup chain, not directly.
* ser-unix.c (wait_for) [HAVE_TERMIO, HAVE_TERMIOS]: Make a timeout
of -1 mean forever, like in the HAVE_SGTTY case. Warn if we are
munging the timeout due to the limited range of c_cc[VTIME].
* procfs.c (procfs_create_inferior), inftarg.c (child_create_inferior),
m3-nat.c (m3_create_inferior): Pass it.
* procfs.c: Remove ptrace function. It was declared in a way which
conflicted with the prototype in unistd.h on Solaris.
step_resume_breakpoint for IN_SIGTRAMP cases.
* infrun.c (wait_for_inferior), breakpoint.h (struct bpstat_what),
breakpoint.c (bpstat_what): Move step_resume from its own field of
the struct bpstat_what into the main_action. Make it override
other breakpoints. This is a conservative change in the sense
that before the step resume breakpoint was a breakpoint.c
breakpoint, hitting the step resume breakpoint overrode even
calling bpstat_stop_status.
in case execution was stopped in the called function.
* stack.c (print_frame_info, frame_info): If backtracing through
a call dummy, handle the starting source line number on a line
boundary like backtracing through sigtramp.
* sparc-tdep.c (sparc_frame_find_saved_regs): Get frame address
for call dummy frame right. Remove old test for dummy frame,
it has been unused at least since gdb-3.5.
* sparc-tdep.c (sparc_push_dummy_frame): Set return address register
of the dummy frame.
Enable "diag" instruction. Delete copr from fpu_misc_tests, it's
in the copr tests now.
* gdb.disasm/hppa.exp: Run new sfu and copr tests. Do not expect
copr in the fpu_misc tests anymore. fpu_misc tests are no longer
expected to fail.
reserved fields to hold a stub unwind entry type. Fix typo.
(stub_unwind_entry): New structure for raw stub unwind entries.
(stub_unwind_types): The types of stubs we may encounter.
(UNWIND_ENTRY_SIZE, STUB_UNWIND_ENTRY_SIZE): New defines.
* hppa-tdep.c (rp_saved): Use additional information provided
by linker stub unwind descriptors.
(frameless_function_invocation): Likewise.
(frame_chain_valid): Likewise.
* paread.c (compare_unwind_entries): New function for sorting
unwind table entries.
(read_unwind_info): Rewrite to remove dependency on host endianness.
Read in data from the $UNWIND_END$ subspace which contains linker
stub unwind descriptors. Merge that data into the basic unwind
table.
Revert Rob changes which consisted of un-doing (accidentally,
presumably) the changes which have happened since 4.12 branched off.
In lib/gdb.exp remove the comment which went with the "catch" which
Rob removed.
checking. Fix overflow check to use unsigned LONGEST, not
unsigned int.
* c-exp.y (parse_number): Make it so that integer constants are
builtin_type_long_long if builtin_type_long isn't big enough or if
an "LL" suffix is used. Properly handle "UL" or "LU" suffixes.
is on, call find_pc_function rather than relying just on the minimal
symbols (probably only matters for symbol readers which don't put
statics in the minimal symbols, but changing this strikes me as
not conservative enough).
Initialize name_location in all cases.
If no symbol and no msymbol, don't print anything symbolic.
symtabs for variables as well as functions. Add `set print
fast-symbolic-addr' and default it to fast (the old way).
Print line numbers for data items as well as functions.
* symtab.c (find_addr_symbol): Return the symtab and the symbol
address, if a symbol is found (take two more args pointing to
where to store these results).
* symtab.h (find_addr_symbol): Add prototype.
compiler generated tag names.
* mdebugread.c (parse_type): Handle cross references to qualified
aggregate types.
* valops.c (value_struct_elt): Improve error message if the
address of a method is requested from an object instance.
* valops.c (search_struct_method): Make name_matched non-static
to get it initialized correctly.
partial_symbol): Shrink the storage sizes of symbols, by making
enums into 1-byte bitfields when compiled __GNUC__, moving all the
enums and small ints to the end of each struct to improve
alignment, and switching the section number from int to unsigned short.
to [0,0] rather than [0, end of first source file]. This avoids
problems with other parts of GDB looking for linetables in the
_globals_ symtab. Eliminate variables num_object_files and
first_object_file_end.
Fixes PR gdb/4072.
(pop_frame): Restore PC2 and LR0 from dummy frames.
(push_dummy_frame): Save PC2 and LR0 into dummy frames.
(setup_arbitrary_frame): Handle 3 args and set up real frames.
* config/a29k/tm-a29k.h (FRAME_NUM_ARGS): Update comments.
(DUMMY_FRAME_RSIZE): Add 2 longwords for PC2 and LR0.
(SETUP_ARBITRARY_FRAME): Define.
there.]
* sparc-tdep.c (sparc_pop_frame): Pop the fsr and csr (float and
coprocessor status regs) when popping a frame. This fixes
float exceptions that occur after calling inferior functions.
* sparc-nat.c (fetch_inferior_registers, store_inferior_registers):
Read and write the fsr (float status register) to/from the child
process along with the float regs. Remove Peter Schauer's change
of May 24 '93, which has higher overhead and doesn't solve the
real problem (which was that FSR wasn't being set).
Read and write the fsr (float status register) to/from the child
process. This avoids random float exceptions when running under
GDB, and probably obsoletes Peter Schauer's change of May 24 '93
(which remains for safety).
stack overflow check isn't right after the register stack
adjustment instruction. Metaware R2.3u compiler moves other
things in front of it. This fix isn't perfect but is what's
running.
