Fix from maj@lucid.com.
* remote-udi.c (udi_open): Reset vars so that user can re-run
programs without leaving GDB.
* (many routines): Slightly improve error handling.
* (download): Zero out BSS by longs instead of bytes to avoid
timeouts in real hardware.
* 29k-share/udi/udip2soc.c (UDIDisconnect, UDIKill): Indicate
that connection is no longer in use after shutdown() of socket.
This keeps GDB from dying of a SIGPIPE when you issue multiple
`target udi' commands.
local formatting information.
* language.h (language_defn): Replace individual format info
with language_format_info structs.
* language.h (local_*_format, local_*_format_prefix,
local_*_format_specifier, local_*_format_suffix): New macros
for binary/octal/decimal/hex formats to access info elements.
* c-exp.y (c_language_defn): Update for new format handling.
* m2-exp.y (m2_language_defn): Update for new format handling.
* dbxread.c (language.h): Include for partial-stab.h use.
* defs.h (local_hex_format, local_hex_format_custom,
local_hex_string, local_hex_string_custom): Move to language.h.
* language.c (local_hex_format_custom, local_hex_string,
local_hex_string_custom, local_octal_format_custom): Use new
format handling.
* language.c (unknown_language_defn, auto_language_defn,
local_language_defn): Update for new format handling.
* printcmd.c (print_scalar_formatted): Use new macros
to access decimal and binary format info for printing.
**** start-sanitize-chill ****
* c-exp.y (chill_language_defn): Update for new format handling.
* ch-exp.y (CHARACTER_LITERAL): Add support to yylex.
* ch-exp.y (match_integer_literal): Add function.
* ch-exp.y (builtin_type_chill_char): Add definition.
* gdbtypes.h (builtin_type_chill_char): Add declaration.
**** end-sanitize-chill ****
* tm-rs6000.h (BELIEVE_PCC_PROMOTION): Define, since AIX cc gets
it right.
(aix_framedata): added nosavedpc field.
(SAVED_PC_AFTER_CALL): Don't try to optimize; just call
read_register.
(FRAMELESS_FUNCTION_INVOCATION): Pass second argument of 0.
(FRAME_SAVED_PC): If PC not saved, use SAVED_PC_AFTER_CALL.
* rs6000-tdep.c (skip_prologue): Handle gcc generated stfd
instructions as function_frame_info does. Expand special case of
st r31,-4(r1) to be st r31,NUM(r1), since gcc can generate offsets
other than -4.
(pop_frame): Add 4 rather than sizeof (int) to avoid host
dependence.
(function_frame_info): Set frameless if the function has no frame,
and set nosavedpc if the PC was not saved. Handle gcc generated
stfd 31,-4(31); st 31, -12(31) correctly.
(frameless_function_invocation): New second argument pcsaved; if 0
return whether the function has a frame, if 1 return whether the
function saved PC.
(frame_initial_stack_address): Correct typo: cache registers for
callee_fi, not for fi, (avoids reading garbage memory locations).
* symtab.c (find_pc_psymtab): Clean up a bit.
* (find_pc_symtab): Fix comments, clean up code.
* (find_pc_line): General cleanups, efficiency improvements.
Also, don't return garbage when some line info exists, but there
was no good match.
29k-share/dfe/mini2udi.o.
* config/a29k-udi.mt: Remove yank.o and mini2udi.o.
* alldeps.mak, depend: Update to deal with removal of
aforementioned files.
* remote-udi.c: Major cleanups. Clean up udi_open and drop
requirement for useless 'program' argument. Clean up
udi_create_inferior, and udi_load to call common download()
routine. Create download routine to load remote hosts directly
(with the help of BFD) so that we don't need yank.c and
mini2udi.c.
Fix udi_detach to call UDIDisconnect with
the right arguments. Clean up udi_resume, don't assign tip_error
twice. Clean up udi_wait, straighten out status codes. Make
udi_kill really work.
removed redundant parse rules. (YYSTYPE): added typed_val, removed
UINT and CHAR rules. (parse_number): work out the targetwise type
of a number based upon it's size and qualifiers. (yylex): chars
are now treated the same way as ints.
* source.c (indentify_source_line): don't core dump if wanted
line is larger than number of lines in source. (can happen when
coff gets confused about #included source).
when debugging a program with 100 shared libraries.
* solib.c (solib_map_sections): Always close the BFD we open.
Free all malloc'd storage we allocate, too, including error cases.
(struct so_list): Remove unused so_bfd member.
(clear_solib): Don't bother closing so_bfd.
* symfile.c (symfile_bfd_open): Mark newly opened BFD as
cacheable. `So many symbol files, so little file descriptors.'
Stubs should stand alone as much as possible.
* source.c (show_directories): Avoid printf_filtered length prob
pointed out by Jonathan Stone.
* i960-pinsn.c (MEM_MAX, MEM_SIZ): Set upper limit properly to
avoid accesses beyond end of table. Fix by Lee W. Cooprider,
<Lee_Cooprider@vos.stratus.com>.
* go32-nat.c: Remove, there is no native go32 support.
* go32-xdep.c: Remove unused fork, fvork, wait, execlp, kill_inferior.
* infcmd.c (run_command): Avoid long calls to printf_filtered.
Bug fix courtesy of Alexander Klaiber.
new routines to deal with sending SIGINTs to attached processes
when the user interrupts the controlling GDB.
* inftarg.c (child_wait), procfs.c (procfs_wait): Add calls to
the aforementioned routines when waiting for the attached process.
* minsyms.c (lookup_minimal_symbol_by_pc): subtract 1, not 2, from
minimal_symbol_count, because the NULL symbol is not included in
the count. This prevented this function from finding the last
symbol in the table.
* dbxread.c (process_one_symbol): if not defined
(BLOCK_ADDRESS_FUNCTION_RELATIVE), set function_start_offset at
the start of the function, not just after N_FUN.
remap all malloc's to xmalloc's and all realloc's to xrealloc's.
* c-exp.y, m2-exp.y: Add comment about how malloc/realloc are
remapped to xmalloc/xrealloc, use only malloc/realloc in grammer
file. Remove preprocessor defines that previously did remapping.
Wind River stole them from the Sun Unix distributions.
We can't distribute them in an FSF GDB release.
We should even rewrite them for our own releases, though that is not
as urgent.
Until these two files are rewritten, VxWorks support in GDB will not compile.
The i960 support can be gained by just rewriting the ptrace.h file.
Only small amounts of the reg.h file need to be rebuilt -- compile
remote-vx.c to see what is needed.
John
is null.
* tm-sun4sol2.h (PROLOGUE_FIRSTLINE_OVERLAP): Remove -- it
causes problems in setting breakpoint in the right place in
functions with `float' args which are passed as doubles.
* xm-vaxbsd.h (MEM_FNS_DECLARED): Avoid erroneous redecl's.
* config/rs6000.mh (NATDEPFILES): Add corelow.o.
* dwarfread.c: include <sys/types.h> for SCO.
* infptrace.c: Don't include ptrace.h under SCO.
* config/i386sco.mh: Use -D_POSIX_SOURCE instead of -posix for
gcc.
* config/i386v.mt: Add exec.o to TDEPFILES.
* irix4-nat.c: Remove externs of registers[], include inferior.h
instead.
* mips-nat.c: Explicitly initialize zerobuf to 0!
* mips-tdep.c (init_extra_frame_info): Undo John's last change.
Always setup fci->frame, even if it's non-zero. Too many places
depend upon this behavior (and I have to get a release out the
door)!
* mipsread.c (parse_partial_symbols, psymtab_to_symtab_1): Set
processing_gcc_compilation if we find the embedded stabs marker.
This fixes several bugs with finding the location of short or char
function parameters passed on the stack.
coredep.c, but zero out FP_REGNUM & ZERO_REGNUM so that stack
backtraces from core files work.
* config/decstation.mh, config/irix3.mh: Remove coredep.o.
Functions are now defined in mips-nat.c.
* doc/gdbint.texinfo: Update GET_LONGJMP_TARGET, L_SET doc.
* irix4-nat.c, mips-nat.c (JB_ELEMENT_SIZE, get_longjmp_target):
Move from mips-tdep.c and tm-{irix3,mips}.h.
* mips-nat.c: Remove a bunch of code that was ifdef'd out of
native MIPS ports.
* nm-irix3.h, nm-mips.h (GET_LONGJMP_TARGET): Move from tm-irix3.h
and tm-mips.h.
* ultra3-nat.c (register_addr): Move from ultra3-xdep.c.
(fetch_core_registers): Fix bfd_seek arguments.
(OBS): Put version.o first, so Makefile rebuild happens early.
* command.h: Publicize prototype for not_just_help_class_command.
* command.c: Remove proto.
* maint.c: Mark "mt" as an abbrev, to avoid duplicated help
output. Move "maint info" from class info to class maintenance,
and improve text.
* infrun.c: Move "stop" to class_obscure, and give it a
function so it will not be seen as a global help topic. FIXME,
it should be possible to set these attributes independently.
* core.c (core_command): Make nicer error message for no core support.
* dbxread.c (read_dbx_symtab): Use L_SET as bfd_seek arg.
(elfstab_build_psymtabs): staboffset and stabstroffset args are
file_ptr's.
* dwarfread.c (struct dwfinfo): Convert dbfoff, lnfoff to file_ptr.
(scan_compilation_units): Punt unused filename arg. dbfoff,
lnoffset, and curlnoffset are file_ptr's now.
(dwarf_build_psymtabs): Drop desc and filename args; use
objfile. dbfoff and lnoffset are file_ptr's now.
(read_ofile_symtab): foffset is now file_ptr. Use L_SET in bfd_seek.
* elfread.c (struct elfinfo): dboffset and lnoffset are file_ptr's.
(elf_symfile_read): Skip desc and filename args to
dwarf_build_psymtabs. Pass file_ptr's to elfstab_build_psymtabs.
* gdb-stabs.h: Use file_ptr rather than off_t.
* mipsread.c (fixup_symtab): f_ptr is a file_ptr. Re-enable
compile-time debug check that someone turned off as "unused".
(read_the_mips_symtab): st_filptr is a file_ptr. Fix bfd_seek call.
* symfile.h: Update dwarf_build_psymtabs and
elfstab_build_psymtabs prototypes.
* xcoffread.c (init_stringtab, init_lineno, xcoff_symfile_read):
Use file_ptr offsets. bfd_seeks use L_SET.
symbol. #define MIPS_EFI_SYMBOL_NAME instead. Use different
value so that demangler won't be invoked. This greatly speeds up
stepping.
* mips-tdep.c (mips_pop_frame): Rewrite handling of
linked_proc_info so that it properly deallocates the appropriate
item after it is done with it instead of before.
* Don't pass bogus frame pointer to create_new_frame(). Just
leave it as zero so that lower level code will figure out the
correct value.
* xm-hppah.h: if __STDC__ is not defined, define
HPPA_COMPILER_BUG.
symtab.c (decode_line_1): avoid a bug in the HP9000/700 native
compiler; see the comment in the file.
Here's the comment from the file:
/* FIXME: The native HP 9000/700 compiler has a bug which appears
when optimizing this file with target i960-vxworks. I haven't
been able to construct a simple test case. The problem is that
in the second call to SKIP_PROLOGUE below, the compiler somehow
does not realize that the statement val = find_pc_line (...) will
change the values of the fields of val. It extracts the elements
into registers at the top of the block, and does not update the
registers after the call to find_pc_line. You can check this by
inserting a printf at the end of find_pc_line to show what values
it is returning for val.pc and val.end and another printf after
the call to see what values the function actually got (remember,
this is compiling with cc -O, with this patch removed). You can
also examine the assembly listing: search for the second call to
skip_prologue; the LDO statement before the next call to
find_pc_line loads the address of the structure which
find_pc_line will return; if there is a LDW just before the LDO,
which fetches an element of the structure, then the compiler
still has the bug. */
* tm-i386bsd.h, xm-i386bsd.h: New config files.
* i386b-nat.c: New native support file.
* configure.in: Add host and target for i[34]86-*-bsd*.
* config/i386bsd.mh, config/i386bsd.mt: New config files.
* nindy-share/nindy.c (coffstrip): Use that pathname, instead of searching for
a "bfd_strip" program. Also, fixed up arguments passed to that program.
* tm-nindy960.h (ADDITIONAL_OPTIONS): Use "-ser" rather than "-r", which is now
used for something else. Rewrite description of associated parameters to match
how gdb does it now.
(ADDITIONAL_OPTION_HELP): Fix message accordingly.
* m68k-pinsn.c (print_insn_arg): Handle new "`" operand type.
in the ser-*.c files. Needed for some systems, such as GO32, which don't
have stuff like <termios.h>. Makefile.in now uses $(REMOTE), which defaults
to remote.o, instead of remote.o.
Specifically, the calling of inferiors methods is improved.
* value.h: New macros METHOD_PTR_IS_VIRTUAL,
METHOD_PTR_FROM_VOFFSET, METHOD_PTR_TO_VOFFSET to partially
hide the implementation details of pointer-to-method objects.
How to tell if the pointer points to a virtual method is
still very dependent on the particular compiler, but this
should make it easier to find the places to change.
* eval.c (evaluate_subexp [case OP_FUNCALL]), valprint.c
(val_print [case TYPE_CODE_PTR]): Use the new METHOD_PTR_*
macros, instead of a hard-wired-in code that incorrectly
assumed a no-longerused representation of pointer-to-method
values. And otherwise fix the relevant bit-rotted code.
* valprint.c (type_print_base [case TYPE_CODE_STRUCT]):
If there are both fields and methods, put a space between.
* stabsread.c (read_struct_type): Fix bug in handling of
GNU C++ anonymous type (indicated by CPLUS_MARKER followed
by '_'). (It used to prematurely exit the loop reading in
the fields, so it would think it should start reading
methods while still in the fields. This could crash gdb
given a gcc that can emit nested type information.)
* valops.c (search_struct_method): Pass 'this' value by
reference instead of by value. This provides a more
consistent interface through a recursive search where the
"bottom" functions may need to adjust offsets (due to multiple
inheritance).
* valops.c, value.h, values.c: Pass extra parameters to
value_fn_field and value_virtual_fn_field so we can
correctly adjust offset for multiple inheritance.
* eval.c (evaluate_subexp [case OP_FUNCALL]): Simplify
virtual function calls by using value_virtual_fn_field().
* values.c: New function baseclass_offset, derived from
baseclass_addr (which perhaps can be made obsolete?).
It returns an offset rather than an address. This is a
cleaner interface since it doesn't mess around allocating
new values.
* valops.c (search_struct_method): Use baseclass_offset
rather than baseclass_addr.
* remote-udi.c (udi_resume): Clean up.
* (udi_wait): Rewrite, leave out bugs.
* Add debugging code to print out all register fetches and stores.
* Straighten out target_ops.
* tm-29k.h (DUMMY_FRAME_RSIZE): Pad out to doubleword.
* remote-udi.c (udi_create_inferior): run with no arguments should
not pass the program name as an argument, since the UDI code
already handles that.
(udi_load): get the symbols from prog_name, not arg_string (the
load command should really use arg_string as the program name, not
prog_name, but at least the run command works now).
* am29k-tdep.c (_initialize_29k): Make it possible for the
user to set and query the address where function calls into the
inferior write a small scratch routine. `set call_scratch_address'
* inferior.h (PC_IN_CALL_DUMMY): Fix fencepost error.
* remote.c (remote_prepare_to_store): Only fetch regs if they are
not already cached validly.
* infptrace.c: remove #ifdef USG from around include ptrace.h.
machines without this header should not be compiling this file.
* nm-rs6000.h, rs6000-nat.c: new files for native support.
* rs6000-tdep.c: do not include sys/ptrace.h or sys/reg.h.
* rs6000-xdep.c: removed. all code now in rs6000-nat.c.
* xm-rs6000.h: do not include ptrace.h.
(ATTACH_DETACH, FETCH_INFERIOR_REGISTERS): moved to nm-rs6000.h.
* config/rs6000.mh (XDEPFILES): removed rs6000-xdep.o.
infptrace.o and inftarg.o move to NATDEPFIES.
(NAT_FILE, NATDEPFILES): new macro for native support.
* config/sun3os3.mh, config/sun3os4.mh (NAT_FILE, NATDEPFILES):
new macros for native support.
(XDEPFILES): moved infptrace.o and inftarg.o to NATDEPFILES,
removed sun3-xdep.o.
* xm-sun3.h (ATTACH_DETACH, FETCH_INFERIOR_REGISTERS): moved to
nm-sun3.h.
* sun3-xdep.c: removed. All code is now in sun3-nat.c.
* sun3-nat.c, nm-sun3.h: new files for native support.
* dwarfread.c (dwarf_build_psymtabs): Rename dbsize parameter to
dbfsize and use it to initialize new local file scope dbsize.
* dwarfread.c (read_ofile_symtab): Initialize dbsize and use it.
* dwarfread.c (basicdieinfo): Use dbsize to check for oversize
DIEs as well as the current check for undersize DIEs. This helps
to gracefully detect and reject corrupted DIE information.
