const works different in C vs C++. In C++, a global "const" variable
has internal linkage by default, resulting in link errors like:
...
extension.o: In function `get_ext_lang_defn(extension_language)':
gdb/extension.c:126: undefined reference to `extension_language_guile'
gdb/extension.c:124: undefined reference to `extension_language_guile'
...
The fix is to define exported const objects with "extern const". But
that in C would not be a definition. So we need to #ifdef C vs C++ in
this case.
EXPORTED_CONST comes from include/ansidecl.h, but in the
feature_to_c.sh case I think it's better to leave the script with no
dependencies.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* cp-valprint.c (vtbl_ptr_name): Use EXPORTED_CONST.
* guile/guile.c (extension_language_guile): Use EXPORTED_CONST.
* features/feature_to_c.sh: Tag the generated xml_builtin array
with extern const in C++ mode.
/home/pedro/gdb/mygit/src/gdb/minidebug.c: At global scope:
/home/pedro/gdb/mygit/src/gdb/minidebug.c:55:8: error: using typedef-name ‘lzma_stream’ after ‘struct’
struct lzma_stream
^
In file included from /usr/include/lzma.h:281:0,
from /home/pedro/gdb/mygit/src/gdb/minidebug.c:28:
/usr/include/lzma/base.h:498:3: note: ‘lzma_stream’ has a previous declaration here
} lzma_stream;
^
gdb/ChangeLog:
2015-02-27 Tom Tromey <tromey@redhat.com>
* minidebug.c (struct lzma_stream): Rename to ...
(struct gdb_lzma_stream): ... this.
(lzma_open, lzma_pread, lzma_close, lzma_stat): Adjust.
The enums are value compatible by design, but building in C++ mode trips
on them, like:
...
gdb/mi/mi-cmd-stack.c:363:34: error: cannot convert ‘print_values’ to ‘ext_lang_frame_args’ for argument ‘3’ to ‘ext_lang_bt_status apply_ext_lang_frame_filter(frame_info*, int, ext_lang_frame_args, ui_out*, int, int)’
...
Fix this by adding a helper function.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* mi/mi-cmd-stack.c (mi_apply_ext_lang_frame_filter): New
function.
(mi_cmd_stack_list_locals, mi_cmd_stack_list_args)
(mi_cmd_stack_list_variables): Use it.
In C++ mode, we get:
gdb/gdbserver/linux-x86-low.c: In function ‘void x86_linux_dr_set(ptid_t, int, long unsigned int)’:
gdb/gdbserver/linux-x86-low.c:558:38: error: ‘regnum’ cannot appear in a constant-expression
offsetof (struct user, u_debugreg[regnum]), value);
^
gdb/gdbserver/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* linux-x86-low.c (u_debugreg_offset): New function.
(x86_linux_dr_get, x86_linux_dr_set): Use it.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* x86-linux-nat.c (u_debugreg_offset): New function.
(x86_linux_dr_get, x86_linux_dr_set): Use it.
Can't do that in C++.
2015-02-27 Pedro Alves <palves@redhat.com>
* nat/x86-dregs.h (enum target_hw_bp_type): Remove forward
declaration.
Include break-common.h.
In C++, we can't do arithmetic on enums. This patch fixes build errors like:
src/gdb/i386-tdep.c: In function ‘int i386_stap_parse_special_token(gdbarch*, stap_parse_info*)’:
src/gdb/i386-tdep.c:4309:7: error: no match for ‘operator++’ (operand type is ‘i386_stap_parse_special_token(gdbarch*, stap_parse_info*)::<anonymous enum>’)
++current_state;
^
...
src/gdb/rs6000-tdep.c:4265:18: error: no match for ‘operator++’ (operand type is ‘powerpc_vector_abi’)
src/gdb/arm-tdep.c:9428:71: error: no match for ‘operator++’ (operand type is ‘arm_float_model’)
src/gdb/arm-tdep.c:9465:64: error: no match for ‘operator++’ (operand type is ‘arm_abi_kind’)
...
gdb/ChangeLog:
2015-02-27 Tom Tromey <tromey@redhat.com>
Pedro Alves <palves@redhat.com>
* arm-tdep.c (set_fp_model_sfunc, arm_set_abi): Use 'int' for
local used to iterate over enums.
* completer.c (signal_completer): Likewise.
* i386-tdep.c (i386_stap_parse_special_token): Likewise.
* rs6000-tdep.c (powerpc_set_vector_abi): Likewise.
* tui/tui-data.c (tui_next_win, tui_prev_win): Likewise.
* tui/tui-layout.c (next_layout, prev_layout): Likewise.
* tui/tui-win.c (tui_refresh_all_win, tui_rehighlight_all)
(tui_resize_all, tui_set_focus_command, tui_all_windows_info): Likewise.
* tui-wingeneral.c (tui_refresh_all): Likewise.
Fixes:
src/gdb/target.h:753:10: error: use of enum ‘exec_direction_kind’ without previous declaration
in C++ mode. We can't forward declare enums.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* target.h: Include "infrun.h".
libthread_db.so calls symbols in the client (GDB), through the
proc-service interface. These routines must have extern "C" linkage
so their symbol names are not mangled when GDB is built as a C++
program. On the GDBserver side, we were missing fallback declarations for
all these symbols.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* gdb_proc_service.h: Wrap with EXTERN_C_PUSH/EXTERN_C_POP.
gdb/gdbserver/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* gdb_proc_service.h: Wrap with EXTERN_C_PUSH/EXTERN_C_POP.
[!HAVE_PROC_SERVICE_H] (struct ps_prochandle): Forward declare.
[!HAVE_PROC_SERVICE_H] (ps_pdread, ps_pdwrite, ps_ptread)
ps_ptwrite, ps_lgetregs, ps_lsetregs, ps_lgetfpregs)
(ps_lsetfpregs, ps_getpid)
(ps_get_thread_area, ps_pglobal_lookup, ps_pstop, ps_pcontinue)
(ps_lstop, ps_lcontinue, ps_lgetxregsize, ps_lgetxregs)
(ps_lsetxregs, ps_plog): Declare.
Functions and variables that are exported by the IPA DSO (that
GDBserver needs to look up) should have "C" mangling, thus be declared
with extern "C".
Function and variable declarations need the extern "C" marker, but
variable definitions can't be marked extern, so the patch splits
IP_AGENT_EXPORT into three.
Building in C++ mode revealed that a few variables were missing
IP_AGENT_EXPORT, thus the IPA has been broken when stripped, even in C
mode... So this ends being a bug fix as well.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* common/agent.h (IPA_SYM_EXPORTED_NAME): New.
(IPA_SYM): Use it.
* common/common-defs.h (EXTERN_C_PUSH, EXTERN_C_POP): New macros.
gdb/gdbserver/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* linux-amd64-ipa.c (gdb_agent_get_raw_reg): Use
IP_AGENT_EXPORT_FUNC.
* linux-i386-ipa.c (gdb_agent_get_raw_reg): Use
IP_AGENT_EXPORT_FUNC.
* tracepoint.c (ATTR_USED, ATTR_NOINLINE, ATTR_CONSTRUCTOR)
(IP_AGENT_EXPORT): Delete.
(gdb_tp_heap_buffer, gdb_jump_pad_buffer, gdb_jump_pad_buffer_end)
(gdb_trampoline_buffer, gdb_trampoline_buffer_end)
(gdb_trampoline_buffer_error, collecting, gdb_collect)
(stop_tracing, flush_trace_buffer, about_to_request_buffer_space)
(trace_buffer_is_full, stopping_tracepoint, expr_eval_result)
(error_tracepoint, tracepoints, tracing, trace_buffer_ctrl)
(trace_buffer_ctrl_curr, trace_buffer_lo, trace_buffer_hi)
(traceframe_read_count, traceframe_write_count)
(traceframes_created, trace_state_variables, get_raw_reg)
(get_trace_state_variable_value, set_trace_state_variable_value)
(ust_loaded, helper_thread_id, cmd_buf): Use
IPA_SYM_EXPORTED_NAME.
(stop_tracing, flush_trace_buffer): Use IP_AGENT_EXPORT_FUNC.
(tracepoints) Use IP_AGENT_EXPORT_VAR.
(stopping_tracepoint, trace_buffer_is_full, expr_eval_result): Use
IP_AGENT_EXPORT_VAR and wrap in EXTERN_C_PUSH/EXTERN_C_POP.
(last_tracepoint): Move into !IN_PROCESS_AGENT block.
(error_tracepoint): Use IP_AGENT_EXPORT_VAR and wrap in
EXTERN_C_PUSH/EXTERN_C_POP.
(trace_state_variables): Use IP_AGENT_EXPORT_VAR.
(trace_buffer_lo, trace_buffer_hi): Use IP_AGENT_EXPORT_VAR and
wrap in EXTERN_C_PUSH/EXTERN_C_POP.
(trace_buffer_ctrl, trace_buffer_ctrl_curr)
(traceframe_write_count, traceframe_read_count)
(traceframes_created, tracing): Use IP_AGENT_EXPORT_VAR.
(about_to_request_buffer_space, get_trace_state_variable_value)
(set_trace_state_variable_value): Use IP_AGENT_EXPORT_FUNC.
(collecting): Use IP_AGENT_EXPORT_VAR and wrap in
EXTERN_C_PUSH/EXTERN_C_POP.
(gdb_collect): Use IP_AGENT_EXPORT_FUNC.
(ust_loaded, cmd_buf): Use IP_AGENT_EXPORT_VAR.
(helper_thread_id, gdb_agent_capability): Use IP_AGENT_EXPORT_VAR
and wrap in EXTERN_C_PUSH/EXTERN_C_POP.
(gdb_tp_heap_buffer, gdb_jump_pad_buffer, gdb_jump_pad_buffer_end)
(gdb_trampoline_buffer, gdb_trampoline_buffer_end)
(gdb_trampoline_buffer_error): Use IP_AGENT_EXPORT_VAR.
* tracepoint.h (ATTR_USED, ATTR_NOINLINE, EXPORTED_SYMBOL):
Define.
(IP_AGENT_EXPORT_FUNC, IP_AGENT_EXPORT_VAR)
(IP_AGENT_EXPORT_VAR_DECL): Define.
(tracing): Declare.
(gdb_agent_get_raw_reg): Declare.
These symbols are defined in C code, so in C++ mode we need to use
extern "C" to declare them. As extern "C" can't be used inside a
function's scope, we move the declarations to the global scope at the
same time.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* cli-out.c (_rl_erase_entire_line): Move declaration out of
cli_mld_erase_entire_line, and make it extern "C".
* common/common-defs.h (EXTERN_C): New.
* completer.c (_rl_completion_prefix_display_length)
(_rl_print_completions_horizontally, QSFUNC): Move declarations
out of gdb_display_match_list_1.
(_rl_qsort_string_compare): Move declaration out of
gdb_display_match_list_1, and make it extern "C".
* defs.h (re_comp): Use EXTERN_C.
* maint.c (_mcleanup): Move declaration out of mcleanup_wrapper,
and make it extern "C".
(monstartup): Move declaration out of maintenance_set_profile_cmd,
and make it extern "C".
(main): Move declaration out of maintenance_set_profile_cmd.
* nat/linux-ptrace.c (linux_ptrace_attach_fail_reason_string): Use
EXTERN_C.
Compiling python.c in C++ mode, we get:
...src/gdb/python/python.c: At global scope:
...src/gdb/python/python.c:106:31: error: storage size of ‘GdbMethods’ isn’t known
static PyMethodDef GdbMethods[];
^
Fix it by making the affected array objects extern.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* python/python.c (GdbMethods): Rename to ...
(python_GdbMethods): ... this and make extern.
(GdbModuleDef): Rename to ...
(python_GdbModuleDef): ... this and make extern.
The set_record_btrace_cmdlist and show_record_btrace_cmdlist objects
are declared twice in the file, seemingly a simply copy/paste
oversight. In C, the first time counts as forward declaration, but in
C++, they are all definitions. That results in:
src/gdb/record-btrace.c:80:33: error: redefinition of ‘cmd_list_element* set_record_btrace_cmdlist’
src/gdb/record-btrace.c:61:33: error: ‘cmd_list_element* set_record_btrace_cmdlist’ previously declared here
src/gdb/record-btrace.c:81:33: error: redefinition of ‘cmd_list_element* show_record_btrace_cmdlist’
src/gdb/record-btrace.c:62:33: error: ‘cmd_list_element* show_record_btrace_cmdlist’ previously declared here
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* record-btrace.c (set_record_btrace_cmdlist)
(show_record_btrace_cmdlist): Remove redefinitions.
---
gdb/record-btrace.c | 4 ----
1 file changed, 4 deletions(-)
In C, an enum or structure defined inside other structure has global
scope just like it had been defined outside the struct in the first
place. However, in C++, such a nested structure is given a name that
is nested inside the structure. This patch moves such affected
structures/enums out to global scope, so that code using them works
the same in C++ as it works today in C.
gdb/ChangeLog:
2015-02-27 Tom Tromey <tromey@redhat.com>
Pedro Alves <palves@redhat.com>
* dwarf2-frame.c (enum cfa_how_kind, struct
dwarf2_frame_state_reg_info): Move out of struct
dwarf2_frame_state.
* dwarf2read.c (struct tu_stats): Move out of struct
dwarf2_per_objfile.
(struct file_entry): Move out of struct line_header.
(struct nextfield, struct nextfnfield, struct fnfieldlist, struct
typedef_field_list): Move out of struct field_info.
* gdbtypes.h (enum dynamic_prop_kind, union dynamic_prop_data):
Move out of struct dynamic_prop.
(union type_owner, union field_location, struct field, struct
range_bounds, union type_specific): Move out of struct main_type.
(struct fn_fieldlist, struct fn_field, struct typedef_field)
(VOFFSET_STATIC): Move out of struct cplus_struct_type.
(struct call_site_target, union call_site_parameter_u, struct
call_site_parameter): Move out of struct call_site.
* m32c-tdep.c (enum m32c_prologue_kind): Move out of struct
m32c_prologue.
(enum srcdest_kind): Move out of struct srcdest.
* main.c (enum cmdarg_kind): Move out of struct cmdarg.
* prologue-value.h (enum prologue_value_kind): Move out of struct
prologue_value.
* s390-linux-tdep.c (enum s390_abi_kind): Move out of struct
gdbarch_tdep.
* stabsread.c (struct nextfield, struct next_fnfieldlist): Move
out of struct field_info.
* symfile.h (struct other_sections): Move out of struct
section_addr_info.
* symtab.c (struct symbol_cache_slot): Move out struct
block_symbol_cache.
* target-descriptions.c (enum tdesc_type_kind): Move out of
typedef struct tdesc_type.
* tui/tui-data.h (enum tui_line_or_address_kind): Move out of
struct tui_line_or_address.
* value.c (enum internalvar_kind, union internalvar_data): Move
out of struct internalvar.
* xtensa-tdep.h (struct ctype_cache): Move out of struct
gdbarch_tdep.
This patch renames symbols that happen to have names which are
reserved keywords in C++.
Most of this was generated with Tromey's cxx-conversion.el script.
Some places where later hand massaged a bit, to fix formatting, etc.
And this was rebased several times meanwhile, along with re-running
the script, so re-running the script from scratch probably does not
result in the exact same output. I don't think that matters anyway.
gdb/
2015-02-27 Tom Tromey <tromey@redhat.com>
Pedro Alves <palves@redhat.com>
Rename symbols whose names are reserved C++ keywords throughout.
gdb/gdbserver/
2015-02-27 Tom Tromey <tromey@redhat.com>
Pedro Alves <palves@redhat.com>
Rename symbols whose names are reserved C++ keywords throughout.
This new option, disabled by default for now, allows specifying
whether to build GDB, GDBserver, and friends with a C++ (98/03)
compiler.
The name of the switch should be familiar to those who followed GCC's
own C++ conversion process.
. Adding -fpermissive to COMPILER in C++ mode (see the new
build-with-cxx.m4 file) makes errors like these be warnings instead:
gdb/infrun.c:6597:1: error: initializing argument 1 of ‘void sig_print_info(gdb_signal)’ [-fpermissive]
sig_print_info (enum gdb_signal oursig)
^
gdb/infrun.c: In function ‘void do_restore_infcall_suspend_state_cleanup(void*)’:
gdb/infrun.c:7164:39: error: invalid conversion from ‘void*’ to ‘infcall_suspend_state*’ [-fpermissive]
restore_infcall_suspend_state (state);
^
so that the compiler carries on compiling the file. -Werror still
catches the warnings, so nothing is lost, only our lifes are made
easier by concentrating on getting other more important things out of
the way first.
There's no way to quiet those warnings. Until they're all fixed, when
building in C++ mode, -Werror is disabled by default.
. Adding -Wno-narrowing suppresses thousands of instances of this warning:
gdb/arm-linux-tdep.c:439:1: error: narrowing conversion of ‘-1’ from ‘int’ to ‘ULONGEST {aka long unsigned int}’ inside { } is ill-formed in C++11 [-Werror=narrowing]
gdb/arm-linux-tdep.c:439:1: error: narrowing conversion of ‘-1l’ from ‘LONGEST {aka long int}’ to ‘ULONGEST {aka long unsigned int}’ inside { } is ill-formed in C++11 [-Werror=narrowing]
gdb/arm-linux-tdep.c:450:1: error: narrowing conversion of ‘-1’ from ‘int’ to ‘ULONGEST {aka long unsigned int}’ inside { } is ill-formed in C++11 [-Werror=narrowing]
We can defer handling those until we target C++11.
. Adding -Wno-sign-compare suppresses thousands of instances of this warning:
gdb/linux-record.c:1763:32: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
if (tmpulongest == tdep->fcntl_F_GETLK64)
^
. Adding -Wno-write-strings suppresses thousands of instances of this warning:
gdb/mi/mi-cmd-var.c: In function ‘void mi_cmd_var_show_attributes(char*, char**, int)’:
gdb/mi/mi-cmd-var.c:514:12: warning: deprecated conversion from string constant to ‘char*’ [-Wwrite-strings]
attstr = "editable";
^
gdb/mi/mi-cmd-var.c:516:12: warning: deprecated conversion from string constant to ‘char*’ [-Wwrite-strings]
attstr = "noneditable";
^
For now, it's best to hide these warnings from view until we're
'-fpermissive'-clean, and can thus start building with -Werror.
The C compiler has always managed to build working GDBs with these
issues in the code, so a C++ compiler should too.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* Makefile.in (COMPILER): New, get it from autoconf.
(COMPILE.pre, CC_LD): Use COMPILER.
(CXX): Get from autoconf instead.
(CXX_FOR_TARGET): Default to g++ instead of gcc.
* acinclude.m4: Include build-with-cxx.m4.
* build-with-cxx.m4: New file.
* configure.ac: Call AC_PROG_CXX and GDB_AC_BUILD_WITH_CXX.
Disable -Werror by default if building in C++ mode.
(build_warnings): Add -Wno-sign-compare, -Wno-write-strings and
-Wno-narrowing in C++ mode. Only enable -Wpointer-sign in C mode.
Run supported-warning-flags tests with the C++ compiler.
Save/restore CXXFLAGS too.
* configure: Regenerate.
gdb/gdbserver/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* Makefile.in (COMPILER): New, get it from autoconf.
(CXX): Get from autoconf instead.
(COMPILE.pre): Use COMPILER.
(CC-LD): Rename to ...
(CC_LD): ... this. Use COMPILER.
(gdbserver$(EXEEXT), gdbreplay$(EXEEXT), $(IPA_LIB)): Adjust.
(CXX_FOR_TARGET): Default to g++ instead of gcc.
* acinclude.m4: Include build-with-cxx.m4.
* configure.ac: Call AC_PROG_CXX and GDB_AC_BUILD_WITH_CXX.
Disable -Werror by default if building in C++ mode.
(build_warnings): Add -Wno-sign-compare, -Wno-write-strings and
-Wno-narrowing in C++ mode. Run supported-warning-flags tests with
the C++ compiler. Save/restore CXXFLAGS too.
* configure: Regenerate.
Converting GDB to be a C++ program, I stumbled on 'basename' issues,
like:
src/gdb/../include/ansidecl.h:169:64: error: new declaration ‘char* basename(const char*)’
/usr/include/string.h:597:26: error: ambiguates old declaration ‘const char* basename(const char*)’
which I believe led to this bit in gold's configure.ac:
dnl We have to check these in C, not C++, because autoconf generates
dnl tests which have no type information, and current glibc provides
dnl multiple declarations of functions like basename when compiling
dnl with C++.
AC_CHECK_DECLS([basename, ffs, asprintf, vasprintf, snprintf, vsnprintf, strverscmp])
These checks IIUC intend to generate all the HAVE_DECL_FOO symbols
that libiberty.h and ansidecl.h check.
GDB is missing these checks currently, which results in the conflict
shown above.
This adds an m4 file that both GDB and GDBserver's configury use to
pull in the autoconf checks that libiberty clients needs done in order
to use these libiberty.h/ansidecl.h.
gdb/ChangeLog:
2015-02-27 Pedro Alves <palves@redhat.com>
* libiberty.m4: New file.
* acinclude.m4: Include libiberty.m4.
* configure.ac: Call libiberty_INIT.
* config.in, configure: Regenerate.
gdb/gdbserver/
2015-02-27 Pedro Alves <palves@redhat.com>
* acinclude.m4: Include libiberty.m4.
* configure.ac: Call libiberty_INIT.
* config.in, configure: Regenerate.
In some scenarios, GDB or GDBserver can be spawned with input _not_
connected to a tty, and then tests that rely on stdio fail with
timeouts, because the inferior's stdout and stderr streams end up
fully buffered.
See discussion here:
https://sourceware.org/ml/gdb-patches/2015-02/msg00809.html
We have a hack in place that works around this for Windows testing,
that forces every test program to link with an .o file that does
(lib/set_unbuffered_mode.c):
static int __gdb_set_unbuffered_output (void) __attribute__ ((constructor));
static int
__gdb_set_unbuffered_output (void)
{
setvbuf (stdout, NULL, _IONBF, BUFSIZ);
setvbuf (stderr, NULL, _IONBF, BUFSIZ);
}
That's a bit hacky; it ends up done for _all_ tests.
This patch adds a way to do this unbuffering explicitly from the test
code itself, so it is done only when necessary, and for all
targets/hosts. For starters, it adjusts gdb.base/interrupt.c to use
it.
Tested on x86_64 Fedora 20, native, and against a remote gdbserver
board file that connects to the target with ssh, with and without -t
(create pty).
gdb/testsuite/
2015-02-27 Pedro Alves <palves@redhat.com>
* lib/unbuffer_output.c: New file.
* gdb.base/interrupt.c: Include "../lib/unbuffer_output.c".
(main): Call gdb_unbuffer_output.
As far as I know, "catch syscall" is supported on hppa*-hp-hpux*, but
the test catch-syscall.exp is skipped on this target by mistake. This
patch is to fix it. However, I don't have a hpux machine to test.
gdb/testsuite:
2015-02-27 Yao Qi <yao.qi@linaro.org>
* gdb.base/catch-syscall.exp: Don't skip it on hppa*-hp-hpux*
target.
On 64-bit S390 platforms the "compile" command always failed because
gcc was not invoked correctly. This patch fixes the compiler
invocation.
gdb/ChangeLog:
* s390-linux-tdep.c (s390_gcc_target_options): Not just handle
31-bit targets, but 64-bit targets as well.
(s390_gnu_triplet_regexp): New function.
(s390_gdbarch_init): Set the gcc_target_options gdbarch method for
64-bit targets as well. Set the gnu_triplet_regexp gdbarch
method.
For amd64, CONTEXT_FULL does not contain CONTEXT_SEGMENTS, which seems
to be needed to retrieve all the segment registers. Add it explicitly,
with a little de-cruftification.
The value of the segment registers isn't terribly useful on amd64, but
at least this makes the output of 'info registers' correct.
Before:
(gdb) i r cs ss ds es fs gs
cs 0x33 51
ss 0x2b 43
ds 0x0 0
es 0x0 0
fs 0x0 0
gs 0x0 0
After:
(gdb) i r cs ss ds es fs gs
cs 0x33 51
ss 0x2b 43
ds 0x2b 43
es 0x2b 43
fs 0x53 83
gs 0x2b 43
gdb/ChangeLog
2015-02-27 Jon TURNEY <jon.turney@dronecode.org.uk>
* windows-nat.c (CONTEXT_DEBUGGER): Remove.
(CONTEXT_DEBUGGER_DR): Add CONTEXT_SEGMENTS. Incorporate flags
from CONTEXT_DEBUGGER.
.decr_pc_after_break is never higher than .breakpoint_len, so use
.breakpoint_len directly. Based on idea from Yao here:
https://sourceware.org/ml/gdb-patches/2015-02/msg00689.html
gdb/gdbserver/ChangeLog:
2015-02-26 Pedro Alves <palves@redhat.com>
* linux-low.c (linux_wait_1): When incrementing the PC past a
program breakpoint always use the_low_target.breakpoint_len as
increment, rather than the maximum between that and
the_low_target.decr_pc_after_break.
Fixes building gdb on x86_64-apple-darwin14 with clang, which produces
a number of warnings from -Wformat-nonliteral.
Ref: https://sourceware.org/ml/gdb/2015-02/msg00047.html
gdb/ChangeLog:
2015-02-26 Pedro Alves <palves@redhat.com>
* auto-load.h (file_is_auto_load_safe): Add ATTRIBUTE_PRINTF.
* complaints.c (vcomplaint): Pass argument FMT directly to
printf-like functions instead of complaint->fmt.
* ctf.c (ctf_save_write_metadata): Add ATTRIBUTE_PRINTF.
* darwin-nat.c (inferior_debug): Add ATTRIBUTE_PRINTF.
* compile/compile-loc2c.c (pushf, unary, binary): Add
ATTRIBUTE_PRINTF.
(do_compile_dwarf_expr_to_c): Pass string literal as format string
to pushf.
(BINARY): Pass string literal as format string to 'binary'.
* compile/compile-object-load.c (link_callbacks_einfo): Add
ATTRIBUTE_PRINTF.
* guile/guile-internal.h (gdbscm_printf): Add ATTRIBUTE_PRINTF.
Preparation for using this on all hosts.
Confirmed that --host=x86_64-w64-mingw32 still builds the stub
termcap.
gdb/ChangeLog:
2015-02-26 Pedro Alves <palves@redhat.com>
* windows-termcap.c: Rename to ...
* stub-termcap.c: ... this. Adjust header line.
* Makefile.in (SFILES): Refer to stub-termcap.c instead of
windows-termcap.c.
* configure: Regenerate.
* configure.ac: Refer to stub-termcap.o instead of
windows-termcap.o.
* gdb_curses.h: Mention stub-termcap.c instead of
windows-termcap.c.
One could not call IFUNCs (=indirect functions) from the compiled injected
code. Either it errored with:
gdb command line:1:1: error: function return type cannot be function
or it just called the IFUNC dispatcher in normal way, returning real function
implementation address instead of the function return value (and thus no
function was called).
gdb/ChangeLog
2015-02-26 Jan Kratochvil <jan.kratochvil@redhat.com>
* compile/compile-c-symbols.c (convert_one_symbol, convert_symbol_bmsym)
(gcc_symbol_address): Call gnu_ifunc_resolve_addr.
gdb/testsuite/ChangeLog
2015-02-26 Jan Kratochvil <jan.kratochvil@redhat.com>
* gdb.compile/compile-ifunc.c: New file.
* gdb.compile/compile-ifunc.exp: New file.
When doing finish in a function, if gdb fails to return a value, gdb
also fails at printing the value type if this type is a struct.
For example :
(gdb) fin
....
Value returned has type: . Cannot determine contents
This patch fixes this by calling type_to_string to print the type
so that we can support these types.
This patch returns the following example output :
(gdb) fin
....
Value returned has type: struct test. Cannot determine contents
Also, this patch modifies structs.exp to check that we return the
correct type.
gdb/ChangeLog:
* gdb/infcmd.c (print_return_value): use type_to_string to print type.
gdb/testsuite/ChangeLog:
* gdb.base/structs.exp: Check for correct struct on finish.
On aarch64, we got the following fail:
(gdb) disassemble func
Dump of assembler code for function func:
0x0000000000400730 <+0>: ret
End of assembler dump.^M
(gdb) x/2i func+0^M
0x400730 <func>: ret^M
0x400734 <main>: stp x29, x30, [sp,#-16]!^M
(gdb) FAIL: gdb.dwarf2/dw2-ifort-parameter.exp: x/2i func+0
the pattern in proc function_range expects to match <func+0>, however,
GDB doesn't display the offset when it is zero. This patch is to
adjust the pattern when $func_length is zero.
gdb/testsuite:
2015-02-26 Yao Qi <yao.qi@linaro.org>
* lib/dwarf.exp (function_range): Adjust pattern when $func_length
is zero.
The attached patch fixes the SEGV and lets GDB successfully
load all kernel modules installed by default on RHEL 7.
Valgrind on F-21 x86_64 host has shown me more clear what is the problem:
Reading symbols from /home/jkratoch/t/cordic.ko...Reading symbols from
/home/jkratoch/t/cordic.ko.debug...=================================================================
==22763==ERROR: AddressSanitizer: heap-use-after-free on address 0x6120000461c8 at pc 0x150cdbd bp 0x7fffffffc7e0 sp 0x7fffffffc7d0
READ of size 8 at 0x6120000461c8 thread T0
#0 0x150cdbc in ppc64_elf_get_synthetic_symtab /home/jkratoch/redhat/gdb-test-asan/bfd/elf64-ppc.c:3282
#1 0x8c5274 in elf_read_minimal_symbols /home/jkratoch/redhat/gdb-test-asan/gdb/elfread.c:1205
#2 0x8c55e7 in elf_symfile_read /home/jkratoch/redhat/gdb-test-asan/gdb/elfread.c:1268
[...]
0x6120000461c8 is located 264 bytes inside of 288-byte region [0x6120000460c0,0x6120000461e0)
freed by thread T0 here:
#0 0x7ffff715454f in __interceptor_free (/lib64/libasan.so.1+0x5754f)
#1 0xde9cde in xfree common/common-utils.c:98
#2 0x9a04f7 in do_my_cleanups common/cleanups.c:155
#3 0x9a05d3 in do_cleanups common/cleanups.c:177
#4 0x8c538a in elf_read_minimal_symbols /home/jkratoch/redhat/gdb-test-asan/gdb/elfread.c:1229
#5 0x8c55e7 in elf_symfile_read /home/jkratoch/redhat/gdb-test-asan/gdb/elfread.c:1268
[...]
previously allocated by thread T0 here:
#0 0x7ffff71547c7 in malloc (/lib64/libasan.so.1+0x577c7)
#1 0xde9b95 in xmalloc common/common-utils.c:41
#2 0x8c4da2 in elf_read_minimal_symbols /home/jkratoch/redhat/gdb-test-asan/gdb/elfread.c:1147
#3 0x8c55e7 in elf_symfile_read /home/jkratoch/redhat/gdb-test-asan/gdb/elfread.c:1268
[...]
SUMMARY: AddressSanitizer: heap-use-after-free /home/jkratoch/redhat/gdb-test-asan/bfd/elf64-ppc.c:3282 ppc64_elf_get_synthetic_symtab
[...]
==22763==ABORTING
A similar case a few lines later I have fixed in 2010 by:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=3f1eff0a2c7f0e7078f011f55b8e7f710aae0cc2
My testcase does not always reproduce it but at least a bit:
* GDB without ppc64 target (even as a secondary one) is reported as "untested"
* ASAN-built GDB with ppc64 target always crashes (and PASSes with this fix)
* unpatched non-ASAN-built GDB with ppc64 target crashes from commandline
* unpatched non-ASAN-built GDB with ppc64 target PASSes from runtest (?)
gdb/ChangeLog
2015-02-26 Jan Kratochvil <jan.kratochvil@redhat.com>
* elfread.c (elf_read_minimal_symbols): Use bfd_alloc for
bfd_canonicalize_symtab.
gdb/testsuite/ChangeLog
2015-02-26 Jan Kratochvil <jan.kratochvil@redhat.com>
* gdb.arch/cordic.ko.bz2: New file.
* gdb.arch/cordic.ko.debug.bz2: New file.
* gdb.arch/ppc64-symtab-cordic.exp: New file.
- Use signal frame sniffers that look for the signal trampoline
instruction sequence to detect most signal frames.
- FreeBSD kernels between 9.2 and 10.1 inclusive do not include the
signal trampoline code in process core dumps. To detect signal
frames for core dumps under these kernels, use the
kern.proc.sigtramp.<pid> sysctl to fetch the location of the signal
trampoline in the gdb process and assume that PC values within this
location are signal frames. This depends on that location being
identical for all binaries.
gdb/ChangeLog:
2015-02-25 John Baldwin <jhb@FreeBSD.org>
* amd64fbsd-nat.c: Include sys/user.h.
(_initialize_amd64fbsd_nat): Use the KERN_PROC_SIGTRAMP sysctl
instead of KERN_PS_STRINGS to locate the signal trampoline.
* i386fbsd-nat.c: Include sys/user.h.
(_initialize_i386fbsd_nat): Use the KERN_PROC_SIGTRAMP sysctl
instead of KERN_PS_STRINGS to locate the signal trampoline.
* amd64fbsd-tdep.c (amd64fbsd_sigtramp_code): New.
(amd64fbsd_sigtramp_p): New.
(amd64fbsd_sigtramp_start_addr, amd64fbsd_sigtramp_end_addr): No
longer set default values.
(amd64fbsd_init_abi): Set "sigtramp_p" to "amd64fbsd_sigtramp_p".
* i386fbsd-tdep.c (i386fbsd_sigtramp_start)
(i386fbsd_sigtramp_middle, i386fbsd_sigtramp_end)
(i386fbsd_freebsd4_sigtramp_start)
(i386fbsd_freebsd4_sigtramp_middle)
(i386fbsd_freebsd4_sigtramp_end, i386fbsd_osigtramp_start)
(i386fbsd_osigtramp_middle, i386fbsd_osigtramp_end): New.
(i386fbsd_sigtramp_p): New.
(i386fbsd_sigtramp_start_addr, i386fbsd_sigtramp_end_addr): No
longer set default values.
(i386fbsd_init_abi): Set "sigtramp_p" to "i386fbsd_sigtramp_p".
amd64fbsd_sigcontext_addr is using frame_unwind_register_unsigned to
fetch the stack pointer which results in infinite recursion. This
patch changes it to use get_frame_register to match the
sigcontext_addr methods in the i386-bsd and amd64-linux targets
instead.
gdb/ChangeLog:
2015-02-25 John Baldwin <jhb@freebsd.org>
* amd64fbsd-tdep.c (amd64fbsd_sigcontext_addr): Use
get_frame_register instead of frame_unwind_register_unsigned.
Hi,
I see the following fail in aarch64-linux-gnu testing...
(gdb) set tdesc file /XXX/gdb/testsuite/gdb.xml/single-reg.xml^M
warning: Architecture rejected target-supplied description^M
(gdb) FAIL: gdb.xml/tdesc-regs.exp: set tdesc file single-reg.xml
core-regs isn't set for aarch64 target, and looks it is an oversight
when aarch64 port was added.
gdb/testsuite:
2015-02-25 Yao Qi <yao.qi@linaro.org>
* gdb.xml/tdesc-regs.exp: Set core-regs to aarch64-core.xml for
aarch64*-*-* target.
I'm going to add an alternate mechanism of breakpoint trap
identification to 'check_stopped_by_breakpoint' that does not rely on
checking the instruction at PC. The mechanism currently used to tell
whether we're stepping over a permanent breakpoint doesn't fit in that
new method. This patch redoes the whole logic in a different way that
works with both old and new methods, in essence moving the "stepped
permanent breakpoint" detection "one level up". It makes lower level
check_stopped_by_breakpoint always the adjust the PC, and then has
linux_wait_1 advance the PC past the breakpoint if necessary. This
ends up being better also because this now handles
non-decr_pc_after_break targets too. Before, such targets would get
stuck forever reexecuting the breakpoint instruction.
Tested on x86_64 Fedora 20.
gdb/gdbserver/ChangeLog:
2015-02-23 Pedro Alves <palves@redhat.com>
* linux-low.c (check_stopped_by_breakpoint): Don't check if the
thread was doing a step-over; always adjust the PC if
we stepped over a permanent breakpoint.
(linux_wait_1): If we stepped over breakpoint that was on top of a
permanent breakpoint, manually advance the PC past it.
Because delete_breakpoints uses gdb_expect directly, an internal error
results in slow timeouts instead of quickly bailing out. This patch
rewrites the procedure to use gdb_test_multiple instead, while
preserving the existing general logic ("delete breakpoints" + "info
breakpoints").
gdb/testsuite/
2015-02-23 Pedro Alves <palves@redhat.com>
* lib/gdb.exp (delete_breakpoints): Rewrite using
gdb_test_multiple.
We need to be careful with parsing optional stop reasons that start
with an hex character ("awatch", "core"), as GDBs that aren't aware of
them parse them as real numbers. That's silly of course, given that
there should be a colon after those magic "numbers". So if strtol on
"abbz:" doesn't return "first invalid char" pointing to the colon, we
know that "abbz" isn't really a register number. It must be optional
stop info we don't know about. This adjusts GDB to work that way,
removing the need for the special casing done upfront:
/* If this packet is an awatch packet, don't parse the 'a'
as a register number. */
if (strncmp (p, "awatch", strlen("awatch")) != 0
&& strncmp (p, "core", strlen ("core") != 0))
For as long as we care about compatibility with GDB 7.9, we'll need to
continue to be careful about this, so I added a comment.
Tested on x86_64 Fedora 20, native gdbserver.
gdb/ChangeLog:
2015-02-23 Pedro Alves <palves@redhat.com>
* remote.c (skip_to_semicolon): New function.
(remote_parse_stop_reply) <T stop reply>: Use it. Don't
special case the stop reasons that look like hex numbers
upfront. Instead handle real register numbers after matching
all the known stop reasons.
Fixes:
> gdb compile failed, /gdb/testsuite/gdb.base/info-os.c: In function 'main':
> /gdb/testsuite/gdb.base/info-os.c:65:3: warning: implicit declaration of function 'atexit' [-Wimplicit-function-declaration]
> atexit (ipc_cleanup);
> ^
> FAIL: gdb.base/info-os.exp: cannot compile test program
with recent GCCs.
gdb/testsuite/ChangeLog:
2015-02-23 Pedro Alves <palves@redhat.com>
* gdb.base/info-os.c: Include stdlib.h.
$ make check RUNTESTFLAGS="--target_board=native-gdbserver/-m32 clone-thread_db.exp"
gdb.log shows:
Running target native-gdbserver/-m32
...
clone-thread_db: src/gdb/testsuite/gdb.threads/clone-thread_db.c:57: thread_fn: Assertion `res != -1' failed.
...
(gdb) FAIL: gdb.threads/clone-thread_db.exp: continue to end
That was waitpid returning -1 / EINTR. We don't see that when testing
with unix/-m32 (native debugging). Turns out to be that when
debugging a 32-bit inferior, a 64-bit GDBserver is reading/writing
$orig_eax from/to the wrong ptrace register buffer offset. When
gdbserver is 64-bit, the ptrace register buffer is in 64-bit layout,
so the register is found at "ORIG_EAX * 8", not at "ORIG_EAX * 4".
Fixes these with --target_board=native-gdbserver/-m32 on x86_64 Fedora 20:
-FAIL: gdb.threads/clone-thread_db.exp: continue to end
+PASS: gdb.threads/clone-thread_db.exp: continue to end
-FAIL: gdb.threads/hand-call-in-threads.exp: all dummies popped
+PASS: gdb.threads/hand-call-in-threads.exp: all dummies popped
PASS: gdb.threads/hand-call-in-threads.exp: breakpoint on all_threads_running
PASS: gdb.threads/hand-call-in-threads.exp: breakpoint on hand_call
PASS: gdb.threads/hand-call-in-threads.exp: disable scheduler locking
@@ -29339,15 +29331,15 @@ PASS: gdb.threads/hand-call-in-threads.e
PASS: gdb.threads/hand-call-in-threads.exp: discard hand call, thread 4
PASS: gdb.threads/hand-call-in-threads.exp: discard hand call, thread 5
PASS: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 1
-FAIL: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 2
-FAIL: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 3
-FAIL: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 4
+PASS: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 2
+PASS: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 3
+PASS: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 4
PASS: gdb.threads/hand-call-in-threads.exp: dummy stack frame number, thread 5
PASS: gdb.threads/hand-call-in-threads.exp: enable scheduler locking
PASS: gdb.threads/hand-call-in-threads.exp: hand call, thread 1
-FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 2
-FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 3
-FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 4
+PASS: gdb.threads/hand-call-in-threads.exp: hand call, thread 2
+PASS: gdb.threads/hand-call-in-threads.exp: hand call, thread 3
+PASS: gdb.threads/hand-call-in-threads.exp: hand call, thread 4
PASS: gdb.threads/hand-call-in-threads.exp: hand call, thread 5
PASS: gdb.threads/hand-call-in-threads.exp: prepare to discard hand call, thread 1
PASS: gdb.threads/hand-call-in-threads.exp: prepare to discard hand call, thread 2
gdb/gdbserver/ChangeLog
2015-02-23 Pedro Alves <palves@redhat.com>
* linux-x86-low.c (REGSIZE): Define in both 32-bit and 64-bit
modes.
(x86_fill_gregset, x86_store_gregset): Use it when handling
$orig_eax.
gdb/testsuite/ChangeLog:
PR symtab/17855
* gdb.ada/exec_changed.exp: Add second test where symbol lookup cache
is read after symbols have been re-read.
* gdb.ada/exec_changed/first.adb (First): New procedure Break_Me.
* gdb.ada/exec_changed/second.adb (Second): Ditto.
This patch addresses two issues.
The basic problem is that "(anonymous namespace)" doesn't get entered
into the symbol table because when dwarf2read.c:new_symbol_full is called
the DIE has no name (dwarf2_name returns NULL).
PR 17976: ptype '(anonymous namespace)' should work like any namespace
PR 17821: perf issue looking up (anonymous namespace)
bash$ gdb monster-program
(gdb) mt set per on
(gdb) mt set symbol-cache-size 0
(gdb) break (anonymous namespace)::foo
Before:
Command execution time: 3.266289 (cpu), 6.169030 (wall)
Space used: 811429888 (+12910592 for this command)
After:
Command execution time: 1.264076 (cpu), 4.057408 (wall)
Space used: 798781440 (+0 for this command)
gdb/ChangeLog:
PR c++/17976, symtab/17821
* cp-namespace.c (cp_search_static_and_baseclasses): New parameter
is_in_anonymous. All callers updated.
(find_symbol_in_baseclass): Ditto.
(cp_lookup_nested_symbol_1): Ditto. Don't search all static blocks
for symbols in an anonymous namespace.
* dwarf2read.c (namespace_name): Don't call dwarf2_name, fetch
DW_AT_name directly.
(dwarf2_name): Convert missing namespace name to
CP_ANONYMOUS_NAMESPACE_STR.
gdeb/testsuite/ChangeLog:
* gdb.cp/anon-ns.exp: Add test for ptype '(anonymous namespace)'.
gdb/testsuite/ChangeLog
2015-02-21 Jan Kratochvil <jan.kratochvil@redhat.com>
PR corefiles/17808
* gdb.arch/i386-biarch-core.core.bz2: New file.
* gdb.arch/i386-biarch-core.exp: New file.
The buildbot shows that the new
gdb.threads/multi-create-ns-info-thr.exp test is timing out when
tested with --target=native-extended-remote. The reason is:
No breakpoints or watchpoints.
(gdb) break main
Breakpoint 1 at 0x10000b00: file ../../../binutils-gdb/gdb/testsuite/gdb.threads/multi-create.c, line 72.
(gdb) run
Starting program: /home/gdb-buildbot/fedora-21-ppc64be-1/fedora-ppc64be-native-extended-gdbserver/build/gdb/testsuite/outputs/gdb.threads/multi-create-ns-info-thr/multi-cre
ate-ns-info-thr
Process /home/gdb-buildbot/fedora-21-ppc64be-1/fedora-ppc64be-native-extended-gdbserver/build/gdb/testsuite/outputs/gdb.threads/multi-create-ns-info-thr/multi-create-ns-inf
o-thr created; pid = 16266
Unexpected vCont reply in non-stop mode: T0501:00003fffffffd190;40:00000080560fe290;thread:p3f8a.3f8a;core:0;
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(gdb) break multi-create.c:45
Breakpoint 2 at 0x10000994: file ../../../binutils-gdb/gdb/testsuite/gdb.threads/multi-create.c, line 45.
(gdb) commands
Type commands for breakpoint(s) 2, one per line.
Non-stop tests don't really work with the
--target_board=native-extended-remote board, because tests toggle
non-stop on after GDB is already connected to gdbserver, while
Currently, non-stop must be enabled before connecting.
This adjusts the test to bail if running to main fails, like all other
non-stop tests.
Note non-stop tests do work with --target_board=native-gdbserver.
gdb/testsuite/ChangeLog:
2015-02-21 Pedro Alves <palves@redhat.com>
* gdb.threads/multi-create-ns-info-thr.exp: Return early if
runto_main fails.
Commit 6f9b8491 (Adapt `info probes' to support printing probes of
different types.) added a new type column to "info probes". That
caused a solib-corrupted.exp regression:
~~~~~~~~~~~~~~~~~~~~~
Running /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.base/solib-corrupted.exp ...
FAIL: gdb.base/solib-corrupted.exp: corrupted list
=== gdb Summary ===
# of expected passes 2
# of unexpected failures 1
~~~~~~~~~~~~~~~~~~~~~
Tested on x86_64 Fedora 20.
gdb/testsuite/ChangeLog:
2015-02-20 Pedro Alves <palves@redhat.com>
* gdb.base/solib-corrupted.exp: Expect "stap" as first column of
info probes.
TL;DR - GDB can hang if something refreshes the thread list out of the
target while the target is running. GDB hangs inside td_ta_thr_iter.
The fix is to not use that libthread_db function anymore.
Long version:
Running the testsuite against my all-stop-on-top-of-non-stop series is
still exposing latent non-stop bugs.
I was originally seeing this with the multi-create.exp test, back when
we were still using libthread_db thread event breakpoints. The
all-stop-on-top-of-non-stop series forces a thread list refresh each
time GDB needs to start stepping over a breakpoint (to pause all
threads). That test hits the thread event breakpoint often, resulting
in a bunch of step-over operations, thus a bunch of thread list
refreshes while some threads in the target are running.
The commit adds a real non-stop mode test that triggers the issue,
based on multi-create.exp, that does an explicit "info threads" when a
breakpoint is hit. IOW, it does the same things the as-ns series was
doing when testing multi-create.exp.
The bug is a race, so it unfortunately takes several runs for the test
to trigger it. In fact, even when setting the test running in a loop,
it sometimes takes several minutes for it to trigger for me.
The race is related to libthread_db's td_ta_thr_iter. This is
libthread_db's entry point for walking the thread list of the
inferior.
Sometimes, when GDB refreshes the thread list from the target,
libthread_db's td_ta_thr_iter can somehow see glibc's thread list as a
cycle, and get stuck in an infinite loop.
The issue is that when a thread exits, its thread control structure in
glibc is moved from a "used" list to a "cache" list. These lists are
simply circular linked lists where the "next/prev" pointers are
embedded in the thread control structure itself. The "next" pointer
of the last element of the list points back to the list's sentinel
"head". There's only one set of "next/prev" pointers for both lists;
thus a thread can only be in one of the lists at a time, not in both
simultaneously.
So when thread C exits, simplifying, the following happens. A-C are
threads. stack_used and stack_cache are the list's heads.
Before:
stack_used -> A -> B -> C -> (&stack_used)
stack_cache -> (&stack_cache)
After:
stack_used -> A -> B -> (&stack_used)
stack_cache -> C -> (&stack_cache)
td_ta_thr_iter starts by iterating at the list's head's next, and
iterates until it sees a thread whose next pointer points to the
list's head again. Thus in the before case above, C's next points to
stack_used, indicating end of list. In the same case, the stack_cache
list is empty.
For each thread being iterated, td_ta_thr_iter reads the whole thread
object out of the inferior. This includes the thread's "next"
pointer.
In the scenario above, it may happen that td_ta_thr_iter is iterating
thread B and has already read B's thread structure just before thread
C exits and its control structure moves to the cached list.
Now, recall that td_ta_thr_iter is running in the context of GDB, and
there's no locking between GDB and the inferior. From it's local copy
of B, td_ta_thr_iter believes that the next thread after B is thread
C, so it happilly continues iterating to C, a thread that has already
exited, and is now in the stack cache list.
After iterating C, td_ta_thr_iter finds the stack_cache head, which
because it is not stack_used, td_ta_thr_iter assumes it's just another
thread. After this, unless the reverse race triggers, GDB gets stuck
in td_ta_thr_iter forever walking the stack_cache list, as no thread
in thatlist has a next pointer that points back to stack_used (the
terminating condition).
Before fully understanding the issue, I tried adding cycle detection
to GDB's td_ta_thr_iter callback. However, td_ta_thr_iter skips
calling the callback in some cases, which means that it's possible
that the callback isn't called at all, making it impossible for GDB to
break the loop. I did manage to get GDB stuck in that state more than
once.
Fortunately, we can avoid the issue altogether. We don't really need
td_ta_thr_iter for live debugging nowadays, given PTRACE_EVENT_CLONE.
We already know how to map and lwp id to a thread id without iterating
(thread_from_lwp), so use that more.
gdb/ChangeLog:
2015-02-20 Pedro Alves <palves@redhat.com>
* linux-nat.c (linux_handle_extended_wait): Call
thread_db_notice_clone whenever a new clone LWP is detected.
(linux_stop_and_wait_all_lwps, linux_unstop_all_lwps): New
functions.
* linux-nat.h (thread_db_attach_lwp): Delete declaration.
(thread_db_notice_clone, linux_stop_and_wait_all_lwps)
(linux_unstop_all_lwps): Declare.
* linux-thread-db.c (struct thread_get_info_inout): Delete.
(thread_get_info_callback): Delete.
(thread_from_lwp): Use td_thr_get_info and record_thread.
(thread_db_attach_lwp): Delete.
(thread_db_notice_clone): New function.
(try_thread_db_load_1): If /proc is mounted and shows the
process'es task list, walk over all LWPs and call thread_from_lwp
instead of relying on td_ta_thr_iter.
(attach_thread): Don't call check_thread_signals here. Split the
tail part of the function (which adds the thread to the core GDB
thread list) to ...
(record_thread): ... this function. Call check_thread_signals
here.
(thread_db_wait): Don't call thread_db_find_new_threads_1. Always
call thread_from_lwp.
(thread_db_update_thread_list): Rename to ...
(thread_db_update_thread_list_org): ... this.
(thread_db_update_thread_list): New function.
(thread_db_find_thread_from_tid): Delete.
(thread_db_get_ada_task_ptid): Simplify.
* nat/linux-procfs.c: Include <sys/stat.h>.
(linux_proc_task_list_dir_exists): New function.
* nat/linux-procfs.h (linux_proc_task_list_dir_exists): Declare.
gdb/gdbserver/ChangeLog:
2015-02-20 Pedro Alves <palves@redhat.com>
* thread-db.c: Include "nat/linux-procfs.h".
(thread_db_init): Skip listing new threads if the kernel supports
PTRACE_EVENT_CLONE and /proc/PID/task/ is accessible.
gdb/testsuite/ChangeLog:
2015-02-20 Pedro Alves <palves@redhat.com>
* gdb.threads/multi-create-ns-info-thr.exp: New file.
This function has a few latent bugs that are triggered by a non-stop
mode test that will be added in a subsequent patch.
First, as described in the function's intro comment, the function is
supposed to return 1 if we're already auto attached to the thread, but
haven't processed the PTRACE_EVENT_CLONE event of its parent thread
yet.
Then, we may find that we're trying to attach to a clone child that
hasn't yet stopped for its initial stop, and therefore 'waitpid(...,
WNOHANG)' returns 0. In that case, we're currently adding the LWP to
the stopped_pids list, which results in linux_handle_extended_wait
skipping the waitpid call on the child, and thus confusing things
later on when the child eventually reports the stop.
Then, the tail end of lin_lwp_attach_lwp always sets the
last_resume_kind of the LWP to resume_stop, which is wrong given that
the user may be doing "info threads" while some threads are running.
And then, the else branch of lin_lwp_attach_lwp always sets the
stopped flag of the LWP. This branch is reached if the LWP is the
main LWP, which may well be running at this point (to it's wrong to
set its 'stopped' flag).
AFAICS, there's no reason anymore for special-casing the main/leader
LWP here:
- For the "attach" case, linux_nat_attach already adds the main LWP to
the lwp list, and sets its 'stopped' flag.
- For the "run" case, after linux_nat_create_inferior, end up in
linux_nat_wait_1 here:
/* The first time we get here after starting a new inferior, we may
not have added it to the LWP list yet - this is the earliest
moment at which we know its PID. */
if (ptid_is_pid (inferior_ptid))
{
/* Upgrade the main thread's ptid. */
thread_change_ptid (inferior_ptid,
ptid_build (ptid_get_pid (inferior_ptid),
ptid_get_pid (inferior_ptid), 0));
lp = add_initial_lwp (inferior_ptid);
lp->resumed = 1;
}
... which adds the LWP to the LWP list already, before
lin_lwp_attach_lwp can ever be reached.
gdb/ChangeLog:
2015-02-20 Pedro Alves <palves@redhat.com>
* linux-nat.c (lin_lwp_attach_lwp): No longer special case the
main LWP. Handle the case of waitpid returning 0 if we're already
attached to the LWP. Don't set the LWP's last_resume_kind to
resume_stop if we already knew about the LWP.
(linux_nat_filter_event): Add debug logs.
The definition was removed a year ago, but the declaration managed to
stay behind.
gdb/ChangeLog
2015-02-20 Pedro Alves <palves@redhat.com>
* target.h (forward_target_decr_pc_after_break): Delete
declaration.
