* blockframe.c (get_prev_frame_info): If INIT_FRAME_PC_FIRST is
set, run it before INIT_EXTRA_FRAME_INFO.
* stack.c (frame_info): If PRINT_EXTRA_FRAME_INFO defined, call it.
* mips-tdep.c (init_extra_frame_info): Only clobber the `frame'
(FP) value in the frame_info struct if it is zero (as from top of
execution stack).
(setup_arbitrary_frame): Implement FRAME_SPECIFICATION_DYADIC.
* mips-xdep.c (fetch_inferior_registers): ZERO_REGNUM always
comes back as zero. So does FP_REGNUM, as a trigger for
init_extra_frame_info.
* tm-mips.h (INIT_FRAME_PC_FIRST): Kludge, FIXME, defined to get
the program counter set before INIT_EXTRA_FRAME_INFO is run.
(INIT_FRAME_PC): Defined to null.
(PRINT_EXTRA_FRAME_INFO): print frame pointer location via symtab.
(FRAME_SPECIFICATION_DYADIC): Ask for two args in frame command.
Briefly explain MIPS stacks in GDB.
Rich and I believe the "real problem" was that both single_step
and target_resume were issuing PT_CONTINUE calls. This would
cause the second PT_CONTINUE to sometimes fail because the process
was already running.
* infptrace.c (child_resume): Remove AIX_BUGGY_PTRACE_CALL kludge.
* infptrace.c (child_resume): Don't deal with NO_SINGLE_STEP
here; it is dealt with at a gdb-target-independent level.
* rs6000-tdep.c (single_step): Don't call ptrace, we are a
high toned routine. Fix return type to void.
screws nested configures. "unset PWD" is the right answer, except that
Decstation sh doesn't *have* unset.
Note also that the send-pr build will fail if you have PWD in your
environment, because emacs also (foolishly) trusts it.
* configure: just set PWD=`pwd` at the top, since Ultrix sh
doesn't have unset and all success paths (and most error paths)
out set it anyway. (Note: should change all uses of ${PWD=`pwd`}
to just ${PWD} to avoid confusion.)
before printing anything, and again afterward. Fix comments.
(vfprintf_filtered): Eliminate static buffer; use auto buffer,
or alloca() if needed.
* rs6000-xdep.c: Use correct conditional (IBM6000_TARGET) to
detect native versus cross-host.
Tue Sep 15 16:00:54 1992 Ian Lance Taylor (ian@cygnus.com)
* configure: always set $(tooldir) to $(libdir)/$(target_alias),
even for a native compilation.
termination condition.
(full_filename): Don't prepend dirname if filename begins with /.
(breakpoint_notify, kernel_dispatch): Deal with DynamicLoad
messages properly!!
(getpty (NCR)): Lotsa bug fixes...
(kernel_dispatch (case StopRType)): Use kill() instead of
killpg() so we can deal with losing systems...
(wait_for_events): Work around NCR select() lossage. It doesn't
zero out readfds when select() returns zero.
(energize_initialize): Use I_SETSIG to get SIGIO interrupts. Use
getcwd() instead of getwd() (NCR getwd() seg faults).
(energize_wait): Use sigaction() for NCR. Their version of signal()
is unreliable by default.
places where we must notify kernel of where the given pc is.
(cplus_demangle): Put single-quotes around demangled names.
(energize_call_command): Call send_location() after doing up,
down, and frame commands.
* Added WARN_SIGNED_OVERFLOW_WORD define to give an error if any
.word is < -32768 or > 32767. The -J flag causes the error to be
ignored. This is to catch over-sized switches generated by gcc on
systems which don't support the broken .word hack.
as.c (main): permit -J if WARN_SIGNED_OVERFLOW_WORD.
write.c (fixup_segment): check for signed .word overflow if
WARN_SIGNED_OVERFLOW_WORD.
pass at doc for two new remote targets---29K using UDI, and Tandem
ST2000 using STDBUG.
* doc/all.m4, doc/none.m4: new m4 switch, _ST2000__
* doc/all.m4: turn on H8/300 stuff for generic manual
* coffcode.h (coff_renumber_symbols): set the last renumbered
symbol pointer to NULL, as expected by coff_write_linenumbers.
(coff_write_relocs): apparently a non-zero addend is OK for reloc
type R_IHCONST used on the 29k.
