Symbol symbol defined by an assignment in a linker script has type
bfd_link_hash_new. elf_i386_convert_load and elf_x86_64_convert_load
should check bfd_link_hash_new to see if a symbol is defined by a linker
script.
bfd/
PR ld/19175
* elf32-i386.c (elf_i386_convert_load): Check bfd_link_hash_new
instead of calling bfd_link_get_defined_symbol.
* elf64-x86-64.c (elf_x86_64_convert_load): Likewise. Skip
relocation overflow for bfd_link_hash_new.
* linker.c (bfd_link_get_defined_symbol): Removed.
* bfd-in2.h: Regenerated.
ld/testsuite/
PR ld/19175
* ld-i386/i386.exp: Run pr19175.
* ld-x86-64/x86-64.exp: Likewise.
* ld-i386/pr19175.d: New file.
* ld-i386/pr19175.s: Likewise.
* ld-i386/pr19175.t: Likewise.
* ld-x86-64/pr19175.d: Likewise.
* ld-x86-64/pr19175.s: Likewise.
* ld-x86-64/pr19175.t: Likewise.
gdb/ChangeLog:
PR python/18938
* cli/cli-cmds (source_script_fron_sctream): New arg file_to_open.
All callers updated.
gdb/testsuite/ChangeLog:
* gdb.python/python.exp: Add test for symlink from .py file to .notpy
file.
gdb/ChangeLog:
* nat/linux-nat.h (__SIGRTMIN): Move here from gdbserver/linux-low.c.
gdb/gdbserver/ChangeLog:
* linux-low.c (__SIGRTMIN): Move to nat/linux-nat.h.
gdb/ChangeLog:
* common/gdb_wait.h (W_STOPCODE): Define, moved here from
gdbserver/linux-low.c.
(WSETSTOP): Simplify.
gdb/gdbserver/ChangeLog:
* linux-low.c (W_STOPCODE): Moved to common/gdb_wait.h.
gdb/ChangeLog:
* linux-thread-db.c (find_new_threads_callback): Cast ti.ti_tid to
unsigned long for debug_printf.
(thread_db_pid_to_str): Ditto.
gdb/gdbserver/ChangeLog:
* thread-db.c (find_one_thread): Cast ti.ti_tid to unsigned long
for debug_printf.
(attach_thread, find_new_threads_callback): Ditto.
As pointed out by Pedro, it's clearer to do it this way. We can trust
that scm_mode_bits won't try to modify our string, even though it takes
a non-const char *.
gdb/ChangeLog:
* guile/scm-ports.c (ioscm_make_gdb_stdio_port): Do not pass a
local char array to scm_mode_bits, use a cast instead.
I stumbled upon this while doing some cxx-conversion work. Since the
x-family alloc functions throw on failure, it is useless to test their
result for failure. The else branch of != NULL is basically dead code.
I changed the type of element_block_ptr to struct tui_win_element, which
seems obvious (this is actually what raised the flag, casting the result
of xmalloc to struct tui_win_element* wouldn't work).
gdb/ChangeLog:
* tui/tui-data.c (tui_alloc_content): Don't check xmalloc
result. Change type of element_block_ptr. Change allocation to
use XNEWVEC.
Since __start_XXX and __stop_XXX symbols aren't defined when address
load is being converted, we need to check if there is an XXX output
section to get their section and value. This patch adds a new function,
bfd_link_get_defined_symbol, to search for the XXX output section to
check if __start_XXX and __stop_XXX symbols are defined.
bfd/
PR ld/19171
* elf32-i386.c (elf_i386_convert_load): Call
bfd_link_get_defined_symbol to check if a symbol is defined.
* elf64-x86-64.c (elf_x86_64_convert_load): Call
bfd_link_get_defined_symbol to get defined symbol section and
value.
* linker.c (bfd_link_get_defined_symbol): New function.
* bfd-in2.h: Regenerated.
ld/testsuite/
PR ld/19171
* ld-i386/lea1.s: Add tests for address load of __start_XXX
and __stop_XXX.
* ld-i386/mov1.s: Likewise.
* ld-x86-64/lea1.s: Likewise.
* ld-x86-64/mov1.s: Likewise.
* ld-i386/lea1a.d: Updated.
* ld-i386/lea1b.d: Likewise.
* ld-i386/lea1c.d: Likewise.
* ld-i386/mov1a.d: Likewise.
* ld-i386/mov1b.d: Likewise.
* ld-x86-64/lea1a.d: Likewise.
* ld-x86-64/lea1b.d: Likewise.
* ld-x86-64/lea1c.d: Likewise.
* ld-x86-64/lea1d.d: Likewise.
* ld-x86-64/lea1e.d: Likewise.
* ld-x86-64/lea1f.d: Likewise.
* ld-x86-64/mov1a.d: Likewise.
* ld-x86-64/mov1b.d: Likewise.
* ld-x86-64/mov1c.d: Likewise.
* ld-x86-64/mov1d.d: Likewise.
I caught a segmentation fault while running gdb.reverse/sigall-reverse.exp,
in a mingw32 GDB, in this code path. It boils down to the code trying to
strlen () a NULL pointer. I tracked things down and it looks like
record_full_message_wrapper_safe is the only offender.
gdb/ChangeLog:
2015-10-26 Luis Machado <lgustavo@codesourcery.com>
* record-full.c (record_full_message_wrapper_safe): Pass empty string to
catch_errors call instead of NULL.
ioscm_make_gdb_stdio_port passes const char pointers (literal strings) to
scm_mode_bits, which takes a non-const char pointer. Ideally, we would
change scm_mode_bits to take a const char pointer, but it's not part of
an API we control.
Instead, it's easy enough to build the string to pass to scm_mode_bits in
a (non-const) char array and pass that.
gdb/ChangeLog:
* guile/scm-ports.c (ioscm_make_gdb_stdio_port): Pass non-const
char pointer to scm_mode_bits.
By having a local variable of type (const gdb_byte *), we can avoid adding
two casts.
gdb/ChangeLog:
* guile/scm-ports.c (gdbscm_memory_port_write): Declare new
"data" local variable and use it.
We currently pass integers as domain_enums to lookup_symbol. The
most obvious fix is to add casts there.
I first thought of changing the type of the domain variables to
domain_enum. However, because we pass a pointer to them to
gdbscm_parse_function_args, which expects them to be integers (because
of the format string), I don't think it would be correct. If the enum
does not have the same size as an int, gdbscm_parse_function_args could
write past the memory of domain, overwriting something else on the
stack.
gdb/ChangeLog:
* guile/scm-symbol.c (gdbscm_lookup_global_symbol): Add
domain_enum cast.
(gdbscm_lookup_symbol): Likewise.
The (void *) casts make the build fail in C++ mode and are unnecessary.
gdb/ChangeLog:
* tui/tui-data.c (tui_add_to_source_windows): Remove void *
cast.
(tui_add_content_elements): Likewise.
A cast here is necessary, just as it's necessary in ps_pdwrite just
below. The type of buf can't be changed, since it's fixed in the ps_pd*
API.
gdb/ChangeLog:
* proc-service.c (ps_pdread): Add cast.
This:
valbuf = memcpy (buf, valbuf, len);
causes a build failure in C++, because memcpy returns the value of
"buf" as a void *. Instead of adding a cast, we can just do the
assignment separately.
gdb/ChangeLog:
* sparc64-tdep.c (sparc64_store_arguments): Split assignment of
valbuf.
Remove these (void *) casts, which cause a build failure in C++ mode.
gdb/ChangeLog:
* ia64-tdep.c (ia64_pseudo_register_write): Remove cast.
(ia64_push_dummy_call): Remove cast and change type of "to" to
array of gdb_byte.
This patch fixes the build that was broken by :
https://sourceware.org/ml/gdb-patches/2015-10/msg00369.html
It implements the sw_breakpoint_from_kind operation on these targets and removes
the calls to set_breakpoint_data.
Compiliation tested on win32.
Not tested : nto, spu.
gdb/gdbserver/ChangeLog:
* nto-low.c (nto_sw_breakpoint_from_kind): New function.
(struct target_ops) <sw_breakpoint_from_kind>: Initialize.
(initialize_low): Remove set_breakpoint_data call.
* spu-low.c (spu_sw_breakpoint_from_kind): New function.
(struct target_ops) <sw_breakpoint_from_kind>: Iniitalize.
(initialize_low): Remove set_breakpoint_data call.
* win32-low.c (win32_sw_breakpoint_from_kind): New function.
(struct target_ops) <sw_breakpoint_from_kind>: Initialize.
(initialize_low): Remove set_breakpoint_data call.
This patch moves default_breakpoint_kind_from_pc to target.c and creates a macro
so that all targets can easily use it.
This allows the breakpoint_kind_from_pc operation to be left unimplemented in
targets that do not need it.
This is preparation to fix the win32/nto/spu build that was broken by this
patch: https://sourceware.org/ml/gdb-patches/2015-10/msg00369.html
No regression on Ubuntu 14.04 x86-64 with gdbserver-{native-extended}
gdb/gdbserver/ChangeLog:
* linux-low.c (default_breakpoint_kind_from_pc): Move to target.c.
* mem-break.c (set_breakpoint_at): Use target_breakpoint_kind_from_pc.
* target.c (default_breakpoint_kind_from_pc): Moved from linux-low.c
* target.h (target_breakpoint_kind_from_pc): New macro.
This patch initialize dsd.insn_count, otherwise, it triggers the assert
below on testings we did recently.
gdb:
2015-10-23 Yao Qi <yao.qi@linaro.org>
* aarch64-tdep.c (aarch64_displaced_step_copy_insn): Set
dsd.insn_count to zero.
Revert bba037e0, it breaks the gc-sections model. Instead fix the
underlying problem which is that _bfd_elf_gc_mark_hook is too late to
be setting SEC_KEEP.
PR ld/11133
PR ld/19161
* elflink.c (elf_gc_sweep): Revert last patch.
(_bfd_elf_gc_mark_hook): Don't set SEC_KEEP here.
SEC_KEEP check in elf_gc_sweep was missing in commit:
commit bde6f3eb6d
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Fri Jan 8 01:43:23 2010 +0000
Set SEC_KEEP on section XXX for undefined __start_XXX/__stop_XXX
bfd/
2010-01-07 H.J. Lu <hongjiu.lu@intel.com>
PR ld/11133
* elflink.c (_bfd_elf_gc_mark_hook): Check section XXX for
undefined __start_XXX/__stop_XXX in all input files and set
SEC_KEEP.
This patch adds SEC_KEEP check to elf_gc_sweep.
bfd/
PR ld/19161
* elflink.c (elf_gc_sweep): Always keep sections marked with
SEC_KEEP.
ld/testsuite/
PR ld/19161
* ld-gc/gc.exp: Run pr19161 test.
* ld-gc/pr19161-1.c: New file.
* ld-gc/pr19161-2.c: Likewise.
* ld-gc/pr19161.d: Likewise.
This patch fixes a regression introduced by :
https://sourceware.org/ml/gdb-patches/2015-10/msg00369.html
Tests : gdb.trace/trace-break.exp and gdb.trace/trace-mt.exp would fail on x86
with gdbserver-{native,extended}.
Before this patch, the breakpoint kind set by GDB with a Z packet and the one
set in the case of a tracepoint would be inconsistent on targets that did not
implement breakpoint_kind_from_pc. On x86 for example a breakpoint set by GDB
would have a kind of 1 but a breakpoint set by a tracepoint would have a kind of
0.
This created a missmatch when trying to insert a tracepoint and a breakpoint at
the same location. One of the two breakpoints would be removed with debug
message : "Inconsistent breakpoint kind".
This patch fixes the issue by changing the default 0 breakpoint kind to be
the size of the breakpoint according to sw_breakpoint_from_kind.
The default breakpoint kind must be the breakpoint length to keep consistency
between breakpoints set via GDB and the ones set internally by GDBServer.
No regression on Ubuntu 14.04 x86-64 with gdbserver-{native-extended}
gdb/gdbserver/ChangeLog:
* linux-low.c (default_breakpoint_kind_from_pc): New function.
(linux_breakpoint_kind_from_pc): Use default_breakpoint_kind_from_pc for
the default breakpoint kind.
Explation below based on what Joel wrote at:
https://sourceware.org/ml/gdb-patches/2015-10/msg00274.html
The merge async/sync code paths patch broke attaching on Windows.
This is what we observe, after attaching to any process. At first, it
seems like everything worked fine, since the process stops, and we get
the prompt back:
(gdb) att 3156
Attaching to program `C:\[...]\foo.exe', process 3156
[New Thread 3156.0xcd8]
[New Thread 3156.0xfe4]
0x7770000d in ntdll!DbgBreakPoint () from C:\Windows\SysWOW64\ntdll.dll
(gdb)
However, enter any commands at all, and GDB appears to be hanging.
For instance:
(gdb) set lang ada
[nothing happens]
Despite appearances, GDB is not reading from the prompt. It is
blocked waiting for an event from the inferior. And since our
inferior is stopped, there aren't going to be any events to read.
In chronological order, what happens is that windows_attach calls
do_initial_windows_stuff, which performs the inferior creation,
and repeatedly waits until we get the first SIGTRAP:
while (1)
{
stop_after_trap = 1;
wait_for_inferior ();
tp = inferior_thread ();
if (tp->suspend.stop_signal != GDB_SIGNAL_TRAP)
resume (tp->suspend.stop_signal);
else
break;
}
The call to wait_for_inferior triggers a call to do_target_wait to get
the event, followed by handle_inferior_event to process it. However,
because the first couple of events are "spurious" events, GDB resumes
the execution, and prepares the inferior to wait again:
case TARGET_WAITKIND_SPURIOUS:
[...]
resume (GDB_SIGNAL_0);
prepare_to_wait (ecs);
And prepare_to_wait just does...
ecs->wait_some_more = 1;
if (!target_is_async_p ())
mark_infrun_async_event_handler ();
... which as a result sets the infrun_async_event_handler "ready"
flag to 1.
We get a couple of spurious events before we get the initial SIGTRAP,
at which point we exit the "while (1)" loop above, after which we
reach the end of the attach_command, followed by the normal
end-of-command processing (normal_stop, bp handling, printing the GDB
prompt), back finally to the root of the event loop.
Notice that, at this point, nothing has unset the "ready" flag for the
infrun_async_event_handler. So, when another cycle of
gdb_do_one_event from the event loop, we eventually call
check_async_event_handlers, which finds that the infrun async event
handler is "ready", and therefore calls it's associated "proc"
callback, which does...
inferior_event_handler (INF_REG_EVENT, NULL);
... triggering a blocking call to target_wait, thus hanging forever.
The fix is to use windows_wait and windows_resume directly, similarly
to gdbserver. This will also allow getting rid of 'stop_after_trap'.
gdb/ChangeLog:
2015-10-22 Pedro Alves <palves@redhat.com>
* windows-nat.c (do_initial_windows_stuff): Rewrite loop using
windows_wait and windows_resume directly instead of
wait_for_inferior and resume.
bfd:
* mmo.c (struct mmo_data_struct): New members
symbol_consistency_override_calculated and ignore_symbol_consistency.
(mmo_section_has_contents, mmo_ignore_symbol_consistency): New
functions.
(mmo_create_symbol): Check with mmo_ignore_symbol_consistency if to
report an error for unexpected value of Main.
(mmo_write_symbols_and_terminator): Similar.
Simon Marchi
GDB Administrator
H.J. Lu
Doug Evans
Simon Marchi
Luis Machado
Jan Kratochvil
Iain Buclaw
Antoine Tremblay
Alan Modra
Yao Qi
Pedro Alves
Nick Clifton
Hans-Peter Nilsson