binutils-gdb.git
4 years agogdb: Handle malformed ELF, symbols in non-allocatable sections
Andrew Burgess [Fri, 6 Dec 2019 23:12:29 +0000 (23:12 +0000)]
gdb: Handle malformed ELF, symbols in non-allocatable sections

I ended up debugging a malformed ELF where a section containing
executable code was not correctly marked as allocatable.  Before
realising the ELF was corrupted I tried to place a breakpoint on a
symbol in the non-allocatable, executable section, and GDB crashed.

Though trying to debug such an ELF clearly isn't going to go well I
would prefer, as far as possible, that any input, no matter how
corrupted, not crash GDB.

The crash occurs when trying to set a breakpoint on the name of a
function from the corrupted section.  GDB converts the symbol to a
symtab_and_line, and looks up a suitable section for this.

The problem is that the section is actually an obj_section, which is
stored in the table within the objfile, and we only initialise this
table for allocatable sections (see add_to_objfile_sections_full in
objfiles.c).  So, if the symbol is in a non-allocatable section then
we end up referencing an uninitialised obj_section.

Later we call get_sal_arch on the symtab_and_line, which calls
get_objfile_arch, which uses the objfile from the uninitialised
obj_section, which will be nullptr, at which point GDB crashes.

The fix I propose here is that when we setup the section references on
msymbols, we should check if the bfd_section being referenced is
allocatable or not.  If it is not then we should set the section
reference back to the default 0 section (see how MSYMBOL_OBJ_SECTION
and SYMBOL_OBJ_SECTION treat the 0 section index).

With this fix in place GDB no longer crashes.  Instead GDB creates the
breakpoint at the non-allocated address, and then fails, with an
error, when it tries to insert the breakpoint.

gdb/ChangeLog:

* elfread.c (record_minimal_symbol): Set section index to 0 for
non-allocatable sections.

gdb/testsuite/ChangeLog:

* gdb.dwarf2/dw2-bad-elf-other.S: New file.
* gdb.dwarf2/dw2-bad-elf.c: New file.
* gdb.dwarf2/dw2-bad-elf.exp: New file.

Change-Id: Ie05436ab4c6a71440304d20ee639dfb021223f8b

4 years agogdb/testsuite: Allow DWARF assembler to create multiple line tables
Andrew Burgess [Fri, 6 Dec 2019 21:35:18 +0000 (21:35 +0000)]
gdb/testsuite: Allow DWARF assembler to create multiple line tables

Fixes a bug in the DWARF assembler that prevents multiple line tables
from being created in a test.  We currently don't initialise a couple
of flags, as a result we will only ever generate one end of file list,
and one end of header, in the first line table.  Any additional line
tables will be missing these parts, and will therefore be corrupt.

This fix will be required for a later commit.  There should be no
change in the testsuite after this commit.

gdb/testsuite/ChangeLog:

* lib/dwarf.exp (Dwarf::lines): Reset _line_saw_program and
_line_saw_file.

Change-Id: Id7123f217a036f26ee32d608db3064dd43164596

4 years agoDwarf 5: Handle debug_str_offsets and indexed attributes that have base offsets.
Ali Tamur [Tue, 24 Dec 2019 03:31:24 +0000 (19:31 -0800)]
Dwarf 5: Handle debug_str_offsets and indexed attributes that have base offsets.

* Process debug_str_offsets section. Handle DW_AT_str_offsets_base attribute and
keep the value in dwarf2_cu.

* Make addr_base field in dwarf2_cu optional to disambiguate 0 value
(absent or present and 0).

* During parsing, there is no guarantee that DW_AT_str_offsets_base and
DW_AT_rnglists_base fields will be processed before the attributes that need
those values for correct computation. So make two passes, on the first one mark
the attributes that depend on *_base attributes and process only the others.
On the second pass, only process the attributes that are marked on the first
pass.

* For string attributes, differentiate between addresses that directly point to
a string and those that point to an offset in debug_str_offsets section.

* There are now two attributes, DW_AT_addr_base and DW_AT_GNU_addr_base to read
address offset base. Likewise, there are two attributes, DW_AT_rnglists_base
and DW_AT_GNU_ranges_base to read ranges base. Since there is no guarantee which
ones the compiler will generate, create helper functions to handle all cases.

Tested with CC=/usr/bin/gcc (version 8.3.0) against master branch (also with
-gsplit-dwarf and -gdwarf-4 flags) and there was no increase in the set of
tests that fails. (gdb still cannot debug a 'hello world' program with DWARF 5,
so for the time being, this is all we care about).

This is part of an effort to support DWARF-5 in gdb.

4 years agogdb: use gdb::byte_vector instead of std::vector<char> in core_target::get_core_regis...
Simon Marchi [Mon, 13 Jan 2020 23:12:08 +0000 (18:12 -0500)]
gdb: use gdb::byte_vector instead of std::vector<char> in core_target::get_core_register_section

Since the data held by the `contents` variable is arbitrary binary data,
it should have gdb_byte elements, not char elements.  Also, using
gdb::byte_vector is preferable, since it doesn't unnecessarily
zero-initialize the values.

Instead of adding a cast in the call to m_core_vec->core_read_registers,
I have changed core_read_registers' argument to be a gdb_byte* instead
of a char*.

gdb/ChangeLog:

* gdbcore.h (struct core_fns) <core_read_registers>: Change
core_reg_sect type to gdb_byte *.
* arm-nbsd-nat.c (fetch_elfcore_registers): Likewise.
* cris-tdep.c (fetch_core_registers): Likewise.
* corelow.c (core_target::get_core_register_section): Change
type of `contents` to gdb::byte_vector.

4 years agogdb/tui: Place window titles in the center of the border
Andrew Burgess [Fri, 10 Jan 2020 00:04:25 +0000 (00:04 +0000)]
gdb/tui: Place window titles in the center of the border

In tui-wingeneral.c:box_win () a comment suggest we should display
titles like this:

  +-WINDOW TITLE GOES HERE-+

However, we actually display them like this:

  +--WINDOW TITLE GOES HERE+

The former seems nicer to me, so that's what this commit does.  Short
titles will appear as:

  +-SHORT TITLE------------+

We previously didn't test the horizontal windows borders in the test
suite, however, I've updated things so that we do now check for the
'+-' and '-+' on the upper border, this will give us some protection.

gdb/ChangeLog:

* tui/tui-wingeneral.c (box_win): Position the title in the center
of the border.

gdb/testsuite/ChangeLog:

* lib/tuiterm.exp (Term::_check_box): Check some parts of the top
border.

Change-Id: Iead6910e3b4e68bdf6871f861f23d2efd699faf0

4 years agogdb: use std::vector instead of alloca in core_target::get_core_register_section
Simon Marchi [Mon, 13 Jan 2020 19:32:52 +0000 (14:32 -0500)]
gdb: use std::vector instead of alloca in core_target::get_core_register_section

As I was trying to compile gdb for an m68k host, I got this error:

  CXX    corelow.o
In file included from /binutils-gdb/gdb/gdbsupport/common-defs.h:120,
                 from /binutils-gdb/gdb/defs.h:28,
                 from /binutils-gdb/gdb/corelow.c:20:
/binutils-gdb/gdb/corelow.c: In member function 'void core_target::get_core_register_section(regcache*, const regset*, const char*, int, int, const char*, bool)':
/binutils-gdb/gdb/../include/libiberty.h:727:36: error: 'alloca' bound is unknown [-Werror=alloca-larger-than=]
  727 | # define alloca(x) __builtin_alloca(x)
      |                    ~~~~~~~~~~~~~~~~^~~
/binutils-gdb/gdb/corelow.c:625:23: note: in expansion of macro 'alloca'
  625 |   contents = (char *) alloca (size);
      |                       ^~~~~~

We are using alloca to hold the contents of a the core register
sections.  These sections are typically fairly small, but there is no
realy guarantee, so I think it would be more reasonable to just use
dynamic allocation here.

gdb/ChangeLog:

* corelow.c (core_target::get_core_register_section): Use
  std::vector instead of alloca.

4 years agogdbserver: remove rule for files from regformats/i386
Simon Marchi [Mon, 13 Jan 2020 19:19:05 +0000 (14:19 -0500)]
gdbserver: remove rule for files from regformats/i386

The dat files in regformats/i386 were removed a while ago, this rule is
no longer necessary.

gdb/gdbserver/ChangeLog:

* Makefile.in (%-generated.c): Remove rule for files from
regformats/i386.

4 years agoEnable -Wmissing-declarations diagnostic
Simon Marchi [Mon, 13 Jan 2020 19:05:44 +0000 (14:05 -0500)]
Enable -Wmissing-declarations diagnostic

Now that most warnings of this kind are fixed, we can enable
-Wmissing-declarations.  I say "most", because it is likely that there
are some more in some configurations I am not able to build, but they
should be pretty easy to fix.

gdb/ChangeLog:

* warning.m4: Add -Wmissing-declarations to build_warnings.
* configure: Re-generate.

gdb/gdbserver/ChangeLog:

* configure: Re-generate.

Change-Id: Iae9b59f22eb5dd1965d09f34c5c9e212cddf67ba

4 years agogdbserver: set IP_AGENT_EXPORT_FUNC to static when not building IPA, add declarations
Simon Marchi [Mon, 13 Jan 2020 19:05:32 +0000 (14:05 -0500)]
gdbserver: set IP_AGENT_EXPORT_FUNC to static when not building IPA, add declarations

Fixing the -Wmissing-declarations errors in gdbserver's tracepoint.c is
a bit tricky, because some functions are compiled for both gdbserver, in
which case they should be static, since they are only used in that file,
and for libinproctrace.so, in which case they should be externally
visible, since they need to be looked up.  In the case where they are
externally visible, -Wmissing-declarations requires that a declaration
exists (that's the point of the warning).

I've reused the IP_AGENT_EXPORT_FUNC macro to mark the functions as
static when compiled for gdbserver.  Some seemingly unnecessary
declarations are added for when compiling libinproctrace.so (thanks to
Tom for the suggestion).

gdb/gdbserver/ChangeLog:

* tracepoint.h (IP_AGENT_EXPORT_FUNC) [!IN_PROCESS_AGENT]:
Define to static.
* tracepoint.c (stop_tracing, flush_trace_buffer,
about_to_request_buffer_space, get_trace_state_variable_value,
set_trace_state_variable_value, gdb_collect): Add declaration.

Change-Id: If9c66151bd00c3b9c5caa27a7c21c5a3a952de2a

4 years agogdbserver: make some functions static in linux-x86-low.c
Simon Marchi [Mon, 13 Jan 2020 19:04:53 +0000 (14:04 -0500)]
gdbserver: make some functions static in linux-x86-low.c

These functions are only used in this file, so should be static.

gdb/gdbserver/ChangeLog:

* linux-x86-low.c (x86_linux_regs_info, amd64_emit_eq_goto,
amd64_emit_ne_goto, amd64_emit_lt_goto, amd64_emit_le_goto,
amd64_emit_gt_goto, amd64_emit_ge_goto, amd64_emit_ge_goto,
i386_emit_eq_goto, i386_emit_ne_goto, i386_emit_lt_goto,
i386_emit_le_goto, i386_emit_gt_goto, i386_emit_ge_goto): Make
static.

Change-Id: I703da41867735aefadd49140e80cd60f6ab9ad39

4 years agogdbserver: include gdbsupport/common-inferior.h in inferiors.c
Simon Marchi [Mon, 13 Jan 2020 19:03:18 +0000 (14:03 -0500)]
gdbserver: include gdbsupport/common-inferior.h in inferiors.c

So that the definitions of get_inferior_cwd/set_inferior_cwd see their
declarations.

      CXX    inferiors.o
    /home/smarchi/src/binutils-gdb/gdb/gdbserver/inferiors.c: In function ‘const char* get_inferior_cwd()’:
    /home/smarchi/src/binutils-gdb/gdb/gdbserver/inferiors.c:228:1: error: no previous declaration for ‘const char* get_inferior_cwd()’ [-Werror=missing-declarations]
     get_inferior_cwd ()
     ^~~~~~~~~~~~~~~~
    /home/smarchi/src/binutils-gdb/gdb/gdbserver/inferiors.c: In function ‘void set_inferior_cwd(const char*)’:
    /home/smarchi/src/binutils-gdb/gdb/gdbserver/inferiors.c:236:1: error: no previous declaration for ‘void set_inferior_cwd(const char*)’ [-Werror=missing-declarations]
     set_inferior_cwd (const char *cwd)
     ^~~~~~~~~~~~~~~~

gdb/gdbserver/ChangeLog:

* inferiors.c: Include gdbsupport/common-inferior.h.

Change-Id: Iae5ccb3e1dc37ce79f03f08465f603a0411e7af0

4 years agogdbserver: include hostio.h in hostio-errno.c
Simon Marchi [Mon, 13 Jan 2020 19:03:13 +0000 (14:03 -0500)]
gdbserver: include hostio.h in hostio-errno.c

... so that the definition of hostio_last_error_from_errno in hostio-errno.c
sees the declaration in hostio.h.

Fix this error:

      CXX    hostio-errno.o
    /home/smarchi/src/binutils-gdb/gdb/gdbserver/hostio-errno.c: In function ‘void hostio_last_error_from_errno(char*)’:
    /home/smarchi/src/binutils-gdb/gdb/gdbserver/hostio-errno.c:28:1: error: no previous declaration for ‘void hostio_last_error_from_errno(char*)’ [-Werror=missing-declarations]
     hostio_last_error_from_errno (char *buf)
     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~

gdb/gdbserver/ChangeLog:

* hostio-errno.c: Include hostio.h.

Change-Id: I056308fd4ce12810d0a1b826c423bd0c7eeb8944

4 years agogdb: add declaration to Python init function
Simon Marchi [Mon, 13 Jan 2020 19:03:04 +0000 (14:03 -0500)]
gdb: add declaration to Python init function

When I try to enable -Wmissing-declarations, I get this error:

      CXX    python/python.o
    /home/smarchi/src/binutils-gdb/gdb/python/python.c: In function ‘PyObject* init__gdb_module()’:
    /home/smarchi/src/binutils-gdb/gdb/python/python.c:1582:1: error: no previous declaration for ‘PyObject* init__gdb_module()’ [-Werror=missing-declarations]
     init__gdb_module (void)
     ^~~~~~~~~~~~~~~~

Prevent it by providing a declaration just before the definition.

gdb/ChangeLog:

* python/python.c (init__gdb_module): Add declaration.

Change-Id: I394bc691b7db624708cc4cb2cda28a56ab85a82b

4 years agogdb: add back declarations for _initialize functions
Simon Marchi [Mon, 13 Jan 2020 19:01:38 +0000 (14:01 -0500)]
gdb: add back declarations for _initialize functions

I'd like to enable the -Wmissing-declarations warning.  However, it
warns for every _initialize function, for example:

      CXX    dcache.o
    /home/smarchi/src/binutils-gdb/gdb/dcache.c: In function ‘void _initialize_dcache()’:
    /home/smarchi/src/binutils-gdb/gdb/dcache.c:688:1: error: no previous declaration for ‘void _initialize_dcache()’ [-Werror=missing-declarations]
     _initialize_dcache (void)
     ^~~~~~~~~~~~~~~~~~

The only practical way forward I found is to add back the declarations,
which were removed by this commit:

    commit 481695ed5f6e0a8a9c9c50bfac1cdd2b3151e6c9
    Author: John Baldwin <jhb@FreeBSD.org>
    Date:   Sat Sep 9 11:02:37 2017 -0700

        Remove unnecessary function prototypes.

I don't think it's a big problem to have the declarations for these
functions, but if anybody has a better solution for this, I'll be happy
to use it.

gdb/ChangeLog:

* aarch64-fbsd-nat.c (_initialize_aarch64_fbsd_nat): Add declaration.
* aarch64-fbsd-tdep.c (_initialize_aarch64_fbsd_tdep): Add declaration.
* aarch64-linux-nat.c (_initialize_aarch64_linux_nat): Add declaration.
* aarch64-linux-tdep.c (_initialize_aarch64_linux_tdep): Add declaration.
* aarch64-newlib-tdep.c (_initialize_aarch64_newlib_tdep): Add declaration.
* aarch64-tdep.c (_initialize_aarch64_tdep): Add declaration.
* ada-exp.y (_initialize_ada_exp): Add declaration.
* ada-lang.c (_initialize_ada_language): Add declaration.
* ada-tasks.c (_initialize_tasks): Add declaration.
* agent.c (_initialize_agent): Add declaration.
* aix-thread.c (_initialize_aix_thread): Add declaration.
* alpha-bsd-nat.c (_initialize_alphabsd_nat): Add declaration.
* alpha-linux-nat.c (_initialize_alpha_linux_nat): Add declaration.
* alpha-linux-tdep.c (_initialize_alpha_linux_tdep): Add declaration.
* alpha-nbsd-tdep.c (_initialize_alphanbsd_tdep): Add declaration.
* alpha-obsd-tdep.c (_initialize_alphaobsd_tdep): Add declaration.
* alpha-tdep.c (_initialize_alpha_tdep): Add declaration.
* amd64-darwin-tdep.c (_initialize_amd64_darwin_tdep): Add declaration.
* amd64-dicos-tdep.c (_initialize_amd64_dicos_tdep): Add declaration.
* amd64-fbsd-nat.c (_initialize_amd64fbsd_nat): Add declaration.
* amd64-fbsd-tdep.c (_initialize_amd64fbsd_tdep): Add declaration.
* amd64-linux-nat.c (_initialize_amd64_linux_nat): Add declaration.
* amd64-linux-tdep.c (_initialize_amd64_linux_tdep): Add declaration.
* amd64-nbsd-nat.c (_initialize_amd64nbsd_nat): Add declaration.
* amd64-nbsd-tdep.c (_initialize_amd64nbsd_tdep): Add declaration.
* amd64-obsd-nat.c (_initialize_amd64obsd_nat): Add declaration.
* amd64-obsd-tdep.c (_initialize_amd64obsd_tdep): Add declaration.
* amd64-sol2-tdep.c (_initialize_amd64_sol2_tdep): Add declaration.
* amd64-tdep.c (_initialize_amd64_tdep): Add declaration.
* amd64-windows-nat.c (_initialize_amd64_windows_nat): Add declaration.
* amd64-windows-tdep.c (_initialize_amd64_windows_tdep): Add declaration.
* annotate.c (_initialize_annotate): Add declaration.
* arc-newlib-tdep.c (_initialize_arc_newlib_tdep): Add declaration.
* arc-tdep.c (_initialize_arc_tdep): Add declaration.
* arch-utils.c (_initialize_gdbarch_utils): Add declaration.
* arm-fbsd-nat.c (_initialize_arm_fbsd_nat): Add declaration.
* arm-fbsd-tdep.c (_initialize_arm_fbsd_tdep): Add declaration.
* arm-linux-nat.c (_initialize_arm_linux_nat): Add declaration.
* arm-linux-tdep.c (_initialize_arm_linux_tdep): Add declaration.
* arm-nbsd-nat.c (_initialize_arm_netbsd_nat): Add declaration.
* arm-nbsd-tdep.c (_initialize_arm_netbsd_tdep): Add declaration.
* arm-obsd-tdep.c (_initialize_armobsd_tdep): Add declaration.
* arm-pikeos-tdep.c (_initialize_arm_pikeos_tdep): Add declaration.
* arm-symbian-tdep.c (_initialize_arm_symbian_tdep): Add declaration.
* arm-tdep.c (_initialize_arm_tdep): Add declaration.
* arm-wince-tdep.c (_initialize_arm_wince_tdep): Add declaration.
* auto-load.c (_initialize_auto_load): Add declaration.
* auxv.c (_initialize_auxv): Add declaration.
* avr-tdep.c (_initialize_avr_tdep): Add declaration.
* ax-gdb.c (_initialize_ax_gdb): Add declaration.
* bfin-linux-tdep.c (_initialize_bfin_linux_tdep): Add declaration.
* bfin-tdep.c (_initialize_bfin_tdep): Add declaration.
* break-catch-sig.c (_initialize_break_catch_sig): Add declaration.
* break-catch-syscall.c (_initialize_break_catch_syscall): Add declaration.
* break-catch-throw.c (_initialize_break_catch_throw): Add declaration.
* breakpoint.c (_initialize_breakpoint): Add declaration.
* bsd-uthread.c (_initialize_bsd_uthread): Add declaration.
* btrace.c (_initialize_btrace): Add declaration.
* charset.c (_initialize_charset): Add declaration.
* cli/cli-cmds.c (_initialize_cli_cmds): Add declaration.
* cli/cli-dump.c (_initialize_cli_dump): Add declaration.
* cli/cli-interp.c (_initialize_cli_interp): Add declaration.
* cli/cli-logging.c (_initialize_cli_logging): Add declaration.
* cli/cli-script.c (_initialize_cli_script): Add declaration.
* cli/cli-style.c (_initialize_cli_style): Add declaration.
* coff-pe-read.c (_initialize_coff_pe_read): Add declaration.
* coffread.c (_initialize_coffread): Add declaration.
* compile/compile-cplus-types.c (_initialize_compile_cplus_types): Add declaration.
* compile/compile.c (_initialize_compile): Add declaration.
* complaints.c (_initialize_complaints): Add declaration.
* completer.c (_initialize_completer): Add declaration.
* copying.c (_initialize_copying): Add declaration.
* corefile.c (_initialize_core): Add declaration.
* corelow.c (_initialize_corelow): Add declaration.
* cp-abi.c (_initialize_cp_abi): Add declaration.
* cp-namespace.c (_initialize_cp_namespace): Add declaration.
* cp-support.c (_initialize_cp_support): Add declaration.
* cp-valprint.c (_initialize_cp_valprint): Add declaration.
* cris-linux-tdep.c (_initialize_cris_linux_tdep): Add declaration.
* cris-tdep.c (_initialize_cris_tdep): Add declaration.
* csky-linux-tdep.c (_initialize_csky_linux_tdep): Add declaration.
* csky-tdep.c (_initialize_csky_tdep): Add declaration.
* ctfread.c (_initialize_ctfread): Add declaration.
* d-lang.c (_initialize_d_language): Add declaration.
* darwin-nat-info.c (_initialize_darwin_info_commands): Add declaration.
* darwin-nat.c (_initialize_darwin_nat): Add declaration.
* dbxread.c (_initialize_dbxread): Add declaration.
* dcache.c (_initialize_dcache): Add declaration.
* disasm-selftests.c (_initialize_disasm_selftests): Add declaration.
* disasm.c (_initialize_disasm): Add declaration.
* dtrace-probe.c (_initialize_dtrace_probe): Add declaration.
* dummy-frame.c (_initialize_dummy_frame): Add declaration.
* dwarf-index-cache.c (_initialize_index_cache): Add declaration.
* dwarf-index-write.c (_initialize_dwarf_index_write): Add declaration.
* dwarf2-frame-tailcall.c (_initialize_tailcall_frame): Add declaration.
* dwarf2-frame.c (_initialize_dwarf2_frame): Add declaration.
* dwarf2expr.c (_initialize_dwarf2expr): Add declaration.
* dwarf2loc.c (_initialize_dwarf2loc): Add declaration.
* dwarf2read.c (_initialize_dwarf2_read): Add declaration.
* elfread.c (_initialize_elfread): Add declaration.
* exec.c (_initialize_exec): Add declaration.
* extension.c (_initialize_extension): Add declaration.
* f-lang.c (_initialize_f_language): Add declaration.
* f-valprint.c (_initialize_f_valprint): Add declaration.
* fbsd-nat.c (_initialize_fbsd_nat): Add declaration.
* fbsd-tdep.c (_initialize_fbsd_tdep): Add declaration.
* filesystem.c (_initialize_filesystem): Add declaration.
* findcmd.c (_initialize_mem_search): Add declaration.
* findvar.c (_initialize_findvar): Add declaration.
* fork-child.c (_initialize_fork_child): Add declaration.
* frame-base.c (_initialize_frame_base): Add declaration.
* frame-unwind.c (_initialize_frame_unwind): Add declaration.
* frame.c (_initialize_frame): Add declaration.
* frv-linux-tdep.c (_initialize_frv_linux_tdep): Add declaration.
* frv-tdep.c (_initialize_frv_tdep): Add declaration.
* ft32-tdep.c (_initialize_ft32_tdep): Add declaration.
* gcore.c (_initialize_gcore): Add declaration.
* gdb-demangle.c (_initialize_gdb_demangle): Add declaration.
* gdb_bfd.c (_initialize_gdb_bfd): Add declaration.
* gdbarch-selftests.c (_initialize_gdbarch_selftests): Add declaration.
* gdbarch.c (_initialize_gdbarch): Add declaration.
* gdbtypes.c (_initialize_gdbtypes): Add declaration.
* gnu-nat.c (_initialize_gnu_nat): Add declaration.
* gnu-v2-abi.c (_initialize_gnu_v2_abi): Add declaration.
* gnu-v3-abi.c (_initialize_gnu_v3_abi): Add declaration.
* go-lang.c (_initialize_go_language): Add declaration.
* go32-nat.c (_initialize_go32_nat): Add declaration.
* guile/guile.c (_initialize_guile): Add declaration.
* h8300-tdep.c (_initialize_h8300_tdep): Add declaration.
* hppa-linux-nat.c (_initialize_hppa_linux_nat): Add declaration.
* hppa-linux-tdep.c (_initialize_hppa_linux_tdep): Add declaration.
* hppa-nbsd-nat.c (_initialize_hppanbsd_nat): Add declaration.
* hppa-nbsd-tdep.c (_initialize_hppanbsd_tdep): Add declaration.
* hppa-obsd-nat.c (_initialize_hppaobsd_nat): Add declaration.
* hppa-obsd-tdep.c (_initialize_hppabsd_tdep): Add declaration.
* hppa-tdep.c (_initialize_hppa_tdep): Add declaration.
* i386-bsd-nat.c (_initialize_i386bsd_nat): Add declaration.
* i386-cygwin-tdep.c (_initialize_i386_cygwin_tdep): Add declaration.
* i386-darwin-nat.c (_initialize_i386_darwin_nat): Add declaration.
* i386-darwin-tdep.c (_initialize_i386_darwin_tdep): Add declaration.
* i386-dicos-tdep.c (_initialize_i386_dicos_tdep): Add declaration.
* i386-fbsd-nat.c (_initialize_i386fbsd_nat): Add declaration.
* i386-fbsd-tdep.c (_initialize_i386fbsd_tdep): Add declaration.
* i386-gnu-nat.c (_initialize_i386gnu_nat): Add declaration.
* i386-gnu-tdep.c (_initialize_i386gnu_tdep): Add declaration.
* i386-go32-tdep.c (_initialize_i386_go32_tdep): Add declaration.
* i386-linux-nat.c (_initialize_i386_linux_nat): Add declaration.
* i386-linux-tdep.c (_initialize_i386_linux_tdep): Add declaration.
* i386-nbsd-nat.c (_initialize_i386nbsd_nat): Add declaration.
* i386-nbsd-tdep.c (_initialize_i386nbsd_tdep): Add declaration.
* i386-nto-tdep.c (_initialize_i386nto_tdep): Add declaration.
* i386-obsd-nat.c (_initialize_i386obsd_nat): Add declaration.
* i386-obsd-tdep.c (_initialize_i386obsd_tdep): Add declaration.
* i386-sol2-nat.c (_initialize_amd64_sol2_nat): Add declaration.
* i386-sol2-tdep.c (_initialize_i386_sol2_tdep): Add declaration.
* i386-tdep.c (_initialize_i386_tdep): Add declaration.
* i386-windows-nat.c (_initialize_i386_windows_nat): Add declaration.
* ia64-libunwind-tdep.c (_initialize_libunwind_frame): Add declaration.
* ia64-linux-nat.c (_initialize_ia64_linux_nat): Add declaration.
* ia64-linux-tdep.c (_initialize_ia64_linux_tdep): Add declaration.
* ia64-tdep.c (_initialize_ia64_tdep): Add declaration.
* ia64-vms-tdep.c (_initialize_ia64_vms_tdep): Add declaration.
* infcall.c (_initialize_infcall): Add declaration.
* infcmd.c (_initialize_infcmd): Add declaration.
* inflow.c (_initialize_inflow): Add declaration.
* infrun.c (_initialize_infrun): Add declaration.
* interps.c (_initialize_interpreter): Add declaration.
* iq2000-tdep.c (_initialize_iq2000_tdep): Add declaration.
* jit.c (_initialize_jit): Add declaration.
* language.c (_initialize_language): Add declaration.
* linux-fork.c (_initialize_linux_fork): Add declaration.
* linux-nat.c (_initialize_linux_nat): Add declaration.
* linux-tdep.c (_initialize_linux_tdep): Add declaration.
* linux-thread-db.c (_initialize_thread_db): Add declaration.
* lm32-tdep.c (_initialize_lm32_tdep): Add declaration.
* m2-lang.c (_initialize_m2_language): Add declaration.
* m32c-tdep.c (_initialize_m32c_tdep): Add declaration.
* m32r-linux-nat.c (_initialize_m32r_linux_nat): Add declaration.
* m32r-linux-tdep.c (_initialize_m32r_linux_tdep): Add declaration.
* m32r-tdep.c (_initialize_m32r_tdep): Add declaration.
* m68hc11-tdep.c (_initialize_m68hc11_tdep): Add declaration.
* m68k-bsd-nat.c (_initialize_m68kbsd_nat): Add declaration.
* m68k-bsd-tdep.c (_initialize_m68kbsd_tdep): Add declaration.
* m68k-linux-nat.c (_initialize_m68k_linux_nat): Add declaration.
* m68k-linux-tdep.c (_initialize_m68k_linux_tdep): Add declaration.
* m68k-tdep.c (_initialize_m68k_tdep): Add declaration.
* machoread.c (_initialize_machoread): Add declaration.
* macrocmd.c (_initialize_macrocmd): Add declaration.
* macroscope.c (_initialize_macroscope): Add declaration.
* maint-test-options.c (_initialize_maint_test_options): Add declaration.
* maint-test-settings.c (_initialize_maint_test_settings): Add declaration.
* maint.c (_initialize_maint_cmds): Add declaration.
* mdebugread.c (_initialize_mdebugread): Add declaration.
* memattr.c (_initialize_mem): Add declaration.
* mep-tdep.c (_initialize_mep_tdep): Add declaration.
* mi/mi-cmd-env.c (_initialize_mi_cmd_env): Add declaration.
* mi/mi-cmds.c (_initialize_mi_cmds): Add declaration.
* mi/mi-interp.c (_initialize_mi_interp): Add declaration.
* mi/mi-main.c (_initialize_mi_main): Add declaration.
* microblaze-linux-tdep.c (_initialize_microblaze_linux_tdep): Add declaration.
* microblaze-tdep.c (_initialize_microblaze_tdep): Add declaration.
* mips-fbsd-nat.c (_initialize_mips_fbsd_nat): Add declaration.
* mips-fbsd-tdep.c (_initialize_mips_fbsd_tdep): Add declaration.
* mips-linux-nat.c (_initialize_mips_linux_nat): Add declaration.
* mips-linux-tdep.c (_initialize_mips_linux_tdep): Add declaration.
* mips-nbsd-nat.c (_initialize_mipsnbsd_nat): Add declaration.
* mips-nbsd-tdep.c (_initialize_mipsnbsd_tdep): Add declaration.
* mips-sde-tdep.c (_initialize_mips_sde_tdep): Add declaration.
* mips-tdep.c (_initialize_mips_tdep): Add declaration.
* mips64-obsd-nat.c (_initialize_mips64obsd_nat): Add declaration.
* mips64-obsd-tdep.c (_initialize_mips64obsd_tdep): Add declaration.
* mipsread.c (_initialize_mipsread): Add declaration.
* mn10300-linux-tdep.c (_initialize_mn10300_linux_tdep): Add declaration.
* mn10300-tdep.c (_initialize_mn10300_tdep): Add declaration.
* moxie-tdep.c (_initialize_moxie_tdep): Add declaration.
* msp430-tdep.c (_initialize_msp430_tdep): Add declaration.
* nds32-tdep.c (_initialize_nds32_tdep): Add declaration.
* nios2-linux-tdep.c (_initialize_nios2_linux_tdep): Add declaration.
* nios2-tdep.c (_initialize_nios2_tdep): Add declaration.
* nto-procfs.c (_initialize_procfs): Add declaration.
* objc-lang.c (_initialize_objc_language): Add declaration.
* observable.c (_initialize_observer): Add declaration.
* opencl-lang.c (_initialize_opencl_language): Add declaration.
* or1k-linux-tdep.c (_initialize_or1k_linux_tdep): Add declaration.
* or1k-tdep.c (_initialize_or1k_tdep): Add declaration.
* osabi.c (_initialize_gdb_osabi): Add declaration.
* osdata.c (_initialize_osdata): Add declaration.
* p-valprint.c (_initialize_pascal_valprint): Add declaration.
* parse.c (_initialize_parse): Add declaration.
* ppc-fbsd-nat.c (_initialize_ppcfbsd_nat): Add declaration.
* ppc-fbsd-tdep.c (_initialize_ppcfbsd_tdep): Add declaration.
* ppc-linux-nat.c (_initialize_ppc_linux_nat): Add declaration.
* ppc-linux-tdep.c (_initialize_ppc_linux_tdep): Add declaration.
* ppc-nbsd-nat.c (_initialize_ppcnbsd_nat): Add declaration.
* ppc-nbsd-tdep.c (_initialize_ppcnbsd_tdep): Add declaration.
* ppc-obsd-nat.c (_initialize_ppcobsd_nat): Add declaration.
* ppc-obsd-tdep.c (_initialize_ppcobsd_tdep): Add declaration.
* printcmd.c (_initialize_printcmd): Add declaration.
* probe.c (_initialize_probe): Add declaration.
* proc-api.c (_initialize_proc_api): Add declaration.
* proc-events.c (_initialize_proc_events): Add declaration.
* proc-service.c (_initialize_proc_service): Add declaration.
* procfs.c (_initialize_procfs): Add declaration.
* producer.c (_initialize_producer): Add declaration.
* psymtab.c (_initialize_psymtab): Add declaration.
* python/python.c (_initialize_python): Add declaration.
* ravenscar-thread.c (_initialize_ravenscar): Add declaration.
* record-btrace.c (_initialize_record_btrace): Add declaration.
* record-full.c (_initialize_record_full): Add declaration.
* record.c (_initialize_record): Add declaration.
* regcache-dump.c (_initialize_regcache_dump): Add declaration.
* regcache.c (_initialize_regcache): Add declaration.
* reggroups.c (_initialize_reggroup): Add declaration.
* remote-notif.c (_initialize_notif): Add declaration.
* remote-sim.c (_initialize_remote_sim): Add declaration.
* remote.c (_initialize_remote): Add declaration.
* reverse.c (_initialize_reverse): Add declaration.
* riscv-fbsd-nat.c (_initialize_riscv_fbsd_nat): Add declaration.
* riscv-fbsd-tdep.c (_initialize_riscv_fbsd_tdep): Add declaration.
* riscv-linux-nat.c (_initialize_riscv_linux_nat): Add declaration.
* riscv-linux-tdep.c (_initialize_riscv_linux_tdep): Add declaration.
* riscv-tdep.c (_initialize_riscv_tdep): Add declaration.
* rl78-tdep.c (_initialize_rl78_tdep): Add declaration.
* rs6000-aix-tdep.c (_initialize_rs6000_aix_tdep): Add declaration.
* rs6000-lynx178-tdep.c (_initialize_rs6000_lynx178_tdep):
Add declaration.
* rs6000-nat.c (_initialize_rs6000_nat): Add declaration.
* rs6000-tdep.c (_initialize_rs6000_tdep): Add declaration.
* run-on-main-thread.c (_initialize_run_on_main_thread): Add declaration.
* rust-exp.y (_initialize_rust_exp): Add declaration.
* rx-tdep.c (_initialize_rx_tdep): Add declaration.
* s12z-tdep.c (_initialize_s12z_tdep): Add declaration.
* s390-linux-nat.c (_initialize_s390_nat): Add declaration.
* s390-linux-tdep.c (_initialize_s390_linux_tdep): Add declaration.
* s390-tdep.c (_initialize_s390_tdep): Add declaration.
* score-tdep.c (_initialize_score_tdep): Add declaration.
* ser-go32.c (_initialize_ser_dos): Add declaration.
* ser-mingw.c (_initialize_ser_windows): Add declaration.
* ser-pipe.c (_initialize_ser_pipe): Add declaration.
* ser-tcp.c (_initialize_ser_tcp): Add declaration.
* ser-uds.c (_initialize_ser_socket): Add declaration.
* ser-unix.c (_initialize_ser_hardwire): Add declaration.
* serial.c (_initialize_serial): Add declaration.
* sh-linux-tdep.c (_initialize_sh_linux_tdep): Add declaration.
* sh-nbsd-nat.c (_initialize_shnbsd_nat): Add declaration.
* sh-nbsd-tdep.c (_initialize_shnbsd_tdep): Add declaration.
* sh-tdep.c (_initialize_sh_tdep): Add declaration.
* skip.c (_initialize_step_skip): Add declaration.
* sol-thread.c (_initialize_sol_thread): Add declaration.
* solib-aix.c (_initialize_solib_aix): Add declaration.
* solib-darwin.c (_initialize_darwin_solib): Add declaration.
* solib-dsbt.c (_initialize_dsbt_solib): Add declaration.
* solib-frv.c (_initialize_frv_solib): Add declaration.
* solib-svr4.c (_initialize_svr4_solib): Add declaration.
* solib-target.c (_initialize_solib_target): Add declaration.
* solib.c (_initialize_solib): Add declaration.
* source-cache.c (_initialize_source_cache): Add declaration.
* source.c (_initialize_source): Add declaration.
* sparc-linux-nat.c (_initialize_sparc_linux_nat): Add declaration.
* sparc-linux-tdep.c (_initialize_sparc_linux_tdep): Add declaration.
* sparc-nat.c (_initialize_sparc_nat): Add declaration.
* sparc-nbsd-nat.c (_initialize_sparcnbsd_nat): Add declaration.
* sparc-nbsd-tdep.c (_initialize_sparcnbsd_tdep): Add declaration.
* sparc-obsd-tdep.c (_initialize_sparc32obsd_tdep): Add declaration.
* sparc-sol2-tdep.c (_initialize_sparc_sol2_tdep): Add declaration.
* sparc-tdep.c (_initialize_sparc_tdep): Add declaration.
* sparc64-fbsd-nat.c (_initialize_sparc64fbsd_nat): Add declaration.
* sparc64-fbsd-tdep.c (_initialize_sparc64fbsd_tdep): Add declaration.
* sparc64-linux-nat.c (_initialize_sparc64_linux_nat): Add declaration.
* sparc64-linux-tdep.c (_initialize_sparc64_linux_tdep): Add declaration.
* sparc64-nat.c (_initialize_sparc64_nat): Add declaration.
* sparc64-nbsd-nat.c (_initialize_sparc64nbsd_nat): Add declaration.
* sparc64-nbsd-tdep.c (_initialize_sparc64nbsd_tdep): Add declaration.
* sparc64-obsd-nat.c (_initialize_sparc64obsd_nat): Add declaration.
* sparc64-obsd-tdep.c (_initialize_sparc64obsd_tdep): Add declaration.
* sparc64-sol2-tdep.c (_initialize_sparc64_sol2_tdep): Add declaration.
* sparc64-tdep.c (_initialize_sparc64_adi_tdep): Add declaration.
* stabsread.c (_initialize_stabsread): Add declaration.
* stack.c (_initialize_stack): Add declaration.
* stap-probe.c (_initialize_stap_probe): Add declaration.
* std-regs.c (_initialize_frame_reg): Add declaration.
* symfile-debug.c (_initialize_symfile_debug): Add declaration.
* symfile-mem.c (_initialize_symfile_mem): Add declaration.
* symfile.c (_initialize_symfile): Add declaration.
* symmisc.c (_initialize_symmisc): Add declaration.
* symtab.c (_initialize_symtab): Add declaration.
* target.c (_initialize_target): Add declaration.
* target-connection.c (_initialize_target_connection): Add
declaration.
* target-dcache.c (_initialize_target_dcache): Add declaration.
* target-descriptions.c (_initialize_target_descriptions): Add declaration.
* thread.c (_initialize_thread): Add declaration.
* tic6x-linux-tdep.c (_initialize_tic6x_linux_tdep): Add declaration.
* tic6x-tdep.c (_initialize_tic6x_tdep): Add declaration.
* tilegx-linux-nat.c (_initialize_tile_linux_nat): Add declaration.
* tilegx-linux-tdep.c (_initialize_tilegx_linux_tdep): Add declaration.
* tilegx-tdep.c (_initialize_tilegx_tdep): Add declaration.
* tracectf.c (_initialize_ctf): Add declaration.
* tracefile-tfile.c (_initialize_tracefile_tfile): Add declaration.
* tracefile.c (_initialize_tracefile): Add declaration.
* tracepoint.c (_initialize_tracepoint): Add declaration.
* tui/tui-hooks.c (_initialize_tui_hooks): Add declaration.
* tui/tui-interp.c (_initialize_tui_interp): Add declaration.
* tui/tui-layout.c (_initialize_tui_layout): Add declaration.
* tui/tui-regs.c (_initialize_tui_regs): Add declaration.
* tui/tui-stack.c (_initialize_tui_stack): Add declaration.
* tui/tui-win.c (_initialize_tui_win): Add declaration.
* tui/tui.c (_initialize_tui): Add declaration.
* typeprint.c (_initialize_typeprint): Add declaration.
* ui-style.c (_initialize_ui_style): Add declaration.
* unittests/array-view-selftests.c (_initialize_array_view_selftests): Add declaration.
* unittests/child-path-selftests.c (_initialize_child_path_selftests): Add declaration.
* unittests/cli-utils-selftests.c (_initialize_cli_utils_selftests): Add declaration.
* unittests/common-utils-selftests.c (_initialize_common_utils_selftests): Add declaration.
* unittests/copy_bitwise-selftests.c (_initialize_copy_bitwise_utils_selftests): Add declaration.
* unittests/environ-selftests.c (_initialize_environ_selftests): Add declaration.
* unittests/filtered_iterator-selftests.c
(_initialize_filtered_iterator_selftests): Add declaration.
* unittests/format_pieces-selftests.c (_initialize_format_pieces_selftests): Add declaration.
* unittests/function-view-selftests.c (_initialize_function_view_selftests): Add declaration.
* unittests/help-doc-selftests.c (_initialize_help_doc_selftests): Add declaration.
* unittests/lookup_name_info-selftests.c (_initialize_lookup_name_info_selftests): Add declaration.
* unittests/main-thread-selftests.c
(_initialize_main_thread_selftests): Add declaration.
* unittests/memory-map-selftests.c (_initialize_memory_map_selftests): Add declaration.
* unittests/memrange-selftests.c (_initialize_memrange_selftests): Add declaration.
* unittests/mkdir-recursive-selftests.c (_initialize_mkdir_recursive_selftests): Add declaration.
* unittests/observable-selftests.c (_initialize_observer_selftest): Add declaration.
* unittests/offset-type-selftests.c (_initialize_offset_type_selftests): Add declaration.
* unittests/optional-selftests.c (_initialize_optional_selftests): Add declaration.
* unittests/parse-connection-spec-selftests.c (_initialize_parse_connection_spec_selftests): Add declaration.
* unittests/rsp-low-selftests.c (_initialize_rsp_low_selftests): Add declaration.
* unittests/scoped_fd-selftests.c (_initialize_scoped_fd_selftests): Add declaration.
* unittests/scoped_mmap-selftests.c (_initialize_scoped_mmap_selftests): Add declaration.
* unittests/scoped_restore-selftests.c (_initialize_scoped_restore_selftests): Add declaration.
* unittests/string_view-selftests.c (_initialize_string_view_selftests): Add declaration.
* unittests/style-selftests.c (_initialize_style_selftest): Add declaration.
* unittests/tracepoint-selftests.c (_initialize_tracepoint_selftests): Add declaration.
* unittests/tui-selftests.c (_initialize_tui_selftest): Add
declaration.
* unittests/unpack-selftests.c (_initialize_unpack_selftests): Add declaration.
* unittests/utils-selftests.c (_initialize_utils_selftests): Add declaration.
* unittests/vec-utils-selftests.c (_initialize_vec_utils_selftests): Add declaration.
* unittests/xml-utils-selftests.c (_initialize_xml_utils): Add declaration.
* user-regs.c (_initialize_user_regs): Add declaration.
* utils.c (_initialize_utils): Add declaration.
* v850-tdep.c (_initialize_v850_tdep): Add declaration.
* valops.c (_initialize_valops): Add declaration.
* valprint.c (_initialize_valprint): Add declaration.
* value.c (_initialize_values): Add declaration.
* varobj.c (_initialize_varobj): Add declaration.
* vax-bsd-nat.c (_initialize_vaxbsd_nat): Add declaration.
* vax-nbsd-tdep.c (_initialize_vaxnbsd_tdep): Add declaration.
* vax-tdep.c (_initialize_vax_tdep): Add declaration.
* windows-nat.c (_initialize_windows_nat): Add declaration.
(_initialize_check_for_gdb_ini): Add declaration.
(_initialize_loadable): Add declaration.
* windows-tdep.c (_initialize_windows_tdep): Add declaration.
* x86-bsd-nat.c (_initialize_x86_bsd_nat): Add declaration.
* x86-linux-nat.c (_initialize_x86_linux_nat): Add declaration.
* xcoffread.c (_initialize_xcoffread): Add declaration.
* xml-support.c (_initialize_xml_support): Add declaration.
* xstormy16-tdep.c (_initialize_xstormy16_tdep): Add declaration.
* xtensa-linux-nat.c (_initialize_xtensa_linux_nat): Add declaration.
* xtensa-linux-tdep.c (_initialize_xtensa_linux_tdep): Add declaration.
* xtensa-tdep.c (_initialize_xtensa_tdep): Add declaration.

Change-Id: I13eec7e0ed2b3c427377a7bdb055cf46da64def9

4 years agogdb: make regformats output a declaration for the init function
Simon Marchi [Mon, 13 Jan 2020 18:59:18 +0000 (13:59 -0500)]
gdb: make regformats output a declaration for the init function

When compiling gdbserver for an architecture that uses the regdat.sh
script (such as m68k) and the -Wmissing-declarations compiler flag, I
get:

  REGDAT reg-m68k-generated.c
  CXX    reg-m68k.o
reg-m68k-generated.c:30:1: error: no previous declaration for 'void init_registers_m68k()' [-Werror=missing-declarations]
   30 | init_registers_m68k (void)
      | ^~~~~~~~~~~~~~~~~~~

The same happens with other architectures, such as s390, but I'll be
using 68k as an example.

The init_registers_m68k function is defined in reg-m68k-generated.c,
which is produced by the regformats/regdat.sh script.  This script reads
the regformats/reg-m68k.dat file, containing a register description, and
produces C code that creates a corresponding target description at
runtime.

The init_registers_m68k function is invoked at initialization time in
linux-m68k-low.c.  The function must therefore be non-static, but does
not have a declaration at the moment.

The real clean way of fixing this would be to make regdat.sh generate a
.h file (in addition to the .c file) with declarations for whatever is
in the .c file.  The generated .c file would include the .h file, and
therefore the definition would have a corresponding declaration.  The
linux-m68k-low.c file would also include this .h file, instead of having
its own declaration of init_registers_m68k, like it does now.

However, this would be a quite big change for not much gain.  As far as
I understand, some common architectures (i386, x86-64, ARM, AArch64)
have been moved to dynamically building target descriptions based on
features (the linux-*-tdesc.c files in gdbserver) and don't use
regdat.sh anymore.  Logically (and given infinite development
resources), the other architectures would be migrated to this system too
and the regdat.sh script would be dropped.  A new architecture would
probably not use regdat.sh either.  So I therefore propose this simpler
patch instead, which just adds a local declaration in the generated
file.

gdb/ChangeLog:

* regformats/regdat.sh: Generate declaration for init function.

4 years agogdbserver: fix Makefile dependency of regformat-generated files on regdat.sh
Simon Marchi [Mon, 13 Jan 2020 18:57:32 +0000 (13:57 -0500)]
gdbserver: fix Makefile dependency of regformat-generated files on regdat.sh

The intent of the rules modified by this patch is that the *-generated.c
files generated by regdat.sh are re-generated in the event that
regdat.sh is modified.  However, if I build, touch regdat.sh, and build
again, the files are not re-generated during the second build.

This is because regdat.sh is specified as an order-only dependency [1],
after the pipe.  Make therefore only ensures that regdat.sh exists
before generating the target file, it doesn't check the timestamp of
regdat.sh.

This patch changes it to be a regular prerequisite.

The rules use the $< variable, which is substituted by the first
prerequisite only, so the command lines won't change.

[1] https://www.gnu.org/software/make/manual/html_node/Prerequisite-Types.html

gdb/gdbserver/ChangeLog:

* Makefile.in (%-generated.c): Make $(regdat_sh) a regular
prerequisite.

4 years agoMoev declaration of loop variable outside of the loop.
Nick Clifton [Mon, 13 Jan 2020 17:28:41 +0000 (17:28 +0000)]
Moev declaration of loop variable outside of the loop.

* objdump.c (disassemble_bytes): Remove C99-ism.

4 years agogdb: adjust remote-sim.c to multi-target
Simon Marchi [Mon, 13 Jan 2020 15:58:52 +0000 (10:58 -0500)]
gdb: adjust remote-sim.c to multi-target

The remote-sim.c file doesn't build since the main multi-target patch
(5b6d1e4f, "Multi-target support"), this patch is an attempt to fix it.
I have only build-tested it, so I'm not sure it runs fine, but it should
get us close at least.

I made these functions methods of the gdbsim_target, because they need
to pass the target down to some GDB core functions, like
find_inferior_ptid:

 - get_sim_inferior_data_by_ptid (renamed to get_inferior_data_by_ptid)
 - gdbsim_resume_inferior (renamed to resume_one_inferior)
 - gdbsim_close_inferior (renamed to close_one_inferior)

In the last two, I changed iterate_over_inferiors to a range-based for,
since that gives simpler code (no need to pass data through the void
pointer).

The next_pid variable, INITIAL_PID macro and sim_inferior_data structure
are simply moved up in the file, above gdbsim_target.

gdb/ChangeLog:

* remote-sim.c (next_pid, INITIAL_PID, sim_inferior_data): Move
up.
(gdbsim_target) <get_inferior_data_by_ptid, resume_one_inferior,
close_one_inferior>: New methods.
(get_sim_inferior_data_by_ptid): Move to gdbsim_target,
pass down target to find_inferior_pid.
(gdbsim_target::fetch_registers, gdbsim_target::store_registers):
Pass down target to find_inferior_ptid.
(gdbsim_target::create_inferior): Pass down target to
add_thread_silent.
(gdbsim_close_inferior): Move to gdbsim_close_inferior, pass
target down to find_inferior_ptid and switch_to_thread.
(gdbsim_target::close): Update to call close_one_inferior.
(struct resume_data): Remove.
(gdbsim_resume_inferior): Move to gdbsim_target.  Take arguments
directly, rather than through a void pointer.
(gdbsim_target::resume): Update to call resume_one_inferior.

4 years ago[gas][aarch64] Turn on SVE when using f32mm or f64mm extensions
Matthew Malcomson [Mon, 13 Jan 2020 15:31:39 +0000 (15:31 +0000)]
[gas][aarch64] Turn on SVE when using f32mm or f64mm extensions

There are no instructions under these matrix multiply extensions that
can be used without having SVE enabled.
Since these extensions require SVE, we make that explicit in the options
table.

Tested on aarch64-none-elf without regressions.

gas/ChangeLog:

2020-01-13  Matthew Malcomson  <matthew.malcomson@arm.com>

* config/tc-aarch64.c (f64mm, f32mm): Add sve as a feature
dependency.

4 years agoAdd test driver for the debuginfod support in the binutils sub-directory.
Nick Clifton [Mon, 13 Jan 2020 15:18:57 +0000 (15:18 +0000)]
Add test driver for the debuginfod support in the binutils sub-directory.

* testsuite/binutils-all/debuginfod.exp: New tests.

4 years agoAdd an option to objdump's disassembler to generate ascii art diagrams showing the...
Thomas Troeger [Mon, 13 Jan 2020 12:36:55 +0000 (12:36 +0000)]
Add an option to objdump's disassembler to generate ascii art diagrams showing the destinations of flow control instructions.

binutils* objdump.c (visualize_jumps, color_output, extended_color_output)
(detected_jumps): New variables.
(usage): Add the new jump visualization options.
(option_values): Add new option value.
(long_options): Add the new option.
(jump_info_new, jump_info_free): New functions.
(jump_info_min_address, jump_info_max_address): Likewise.
(jump_info_end_address, jump_info_is_start_address): Likewise.
(jump_info_is_end_address, jump_info_size): Likewise.
(jump_info_unlink, jump_info_insert): Likewise.
(jump_info_add_front, jump_info_move_linked): Likewise.
(jump_info_intersect, jump_info_merge): Likewise.
(jump_info_sort, jump_info_visualize_address): Likewise.
(disassemble_jumps): New function - used to locate jumps.
(disassemble_bytes): Add ascii art generation.
(disassemble_section): Add scan to locate jumps.
(main): Parse the new visualization option.
* doc/binutils.texi: Document the new feature.
* NEWS: Mention the new feature.

opcodes * arm-dis.c (print_insn_arm): Fill in insn info fields for control
flow instructions.
(print_insn_thumb16, print_insn_thumb32): Likewise.
(print_insn): Initialize the insn info.
* i386-dis.c (print_insn): Initialize the insn info fields, and
detect jumps.

4 years agoRe: PR23560, PR23561, readelf memory leaks
Alan Modra [Mon, 13 Jan 2020 12:23:02 +0000 (22:53 +1030)]
Re: PR23560, PR23561, readelf memory leaks

PR 25360
PR 25361
Dyslexia strikes again.

Fix git commit a788aedd86da983faf0afef3cb41461118a2e9f2 ChangeLog.

4 years agoRegen ld BLD-POTFILES.in
Alan Modra [Mon, 13 Jan 2020 12:14:04 +0000 (22:44 +1030)]
Regen ld BLD-POTFILES.in

* po/BLD-POTFILES.in: Regenerate.

4 years agoPR23560, PR23561, readelf memory leaks
Alan Modra [Mon, 13 Jan 2020 12:00:46 +0000 (22:30 +1030)]
PR23560, PR23561, readelf memory leaks

PR 23560
PR 23561
* dwarf.c (display_debug_frames): Move fde_fc earlier.  Free
fde_fc col_type and col_offset.
* readelf.c (apply_relocations): Move symsec check earlier.
(free_debug_section): Free reloc_info.
(process_notes_at): Free pnotes on error path.
(process_object): Free dump_sects here..
(process_archive): ..not here.

4 years agoPR25362, memory leak in nm
Alan Modra [Mon, 13 Jan 2020 09:06:13 +0000 (19:36 +1030)]
PR25362, memory leak in nm

PR 25362
* nm.c (display_rel_file): Free dyn_syms.

4 years ago[ARC][committed] Update test matching pattern.
Claudiu Zissulescu [Mon, 13 Jan 2020 09:16:47 +0000 (11:16 +0200)]
[ARC][committed] Update test matching pattern.

xxxx-xx-xx  Claudiu Zissulescu <claziss@gmail.com>

        * testsuite/ld-arc/relax-local-pic.d: Improve matching patterns.

4 years ago[ARC][committed] Code cleanup and improvements.
Claudiu Zissulescu [Mon, 13 Jan 2020 09:16:47 +0000 (11:16 +0200)]
[ARC][committed] Code cleanup and improvements.

Code clean up and improvements when changing the cpu from command
line. Also, remove unused/old emulations.

gas/
xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* config/tc-arc.c (arc_select_cpu): Re-init the bfd if we change
the CPU.
* config/tc-arc.h: Add header if/defs.
* testsuite/gas/arc/pseudos.d: Improve matching pattern.

ls/
xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* Makefile.am: Remove earcelf_prof.c and earclinux_prof.c
emulations.
* Makefile.in: Regenerate.
* configure.tgt: Likewise.
* emulparams/arcelf_prof.sh: Remove file.
* emulparams/arclinux_prof.sh: Likewise.

opcodes/
xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* arc-opc.c (C_NE): Make it required.

4 years ago[ARC][committed] Update ARC cpu list
Claudiu Zissulescu [Mon, 13 Jan 2020 09:16:47 +0000 (11:16 +0200)]
[ARC][committed] Update ARC cpu list

include/
xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* elf/arc-cpu.def: Update ARC cpu list.

4 years ago[ARC][committed] Use DWARF.sc in elf linker script templates.
Claudiu Zissulescu [Mon, 13 Jan 2020 09:16:47 +0000 (11:16 +0200)]
[ARC][committed] Use DWARF.sc in elf linker script templates.

xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* elfarcv2.sc : Allow interrupt vector table to be located at an
arbitrary address.  Use DWARF.sc file.
* elfarc.sc: Use DWARF.sc file.

4 years ago[ARC] [COMMITTED] Change ACCL/ACCH reg name to generic.
Claudiu Zissulescu [Mon, 13 Jan 2020 08:21:30 +0000 (10:21 +0200)]
[ARC] [COMMITTED] Change ACCL/ACCH reg name to generic.

ACCL/ACCH register names are only available for ARCv2 architecture,
leading to a confusion when disassembling for any other ARC
variants. This patch is changing the default names for ACCL/ACCH to
generic r58/r59.

2012-01-13  Claudiu Zissulescu <claziss@gmail.com>

        * opcode/arc-dis.c (regnames): Correct ACCL/ACCH naming, fix typo
reserved register name.

4 years agoasan: ns32k: wild memory write
Alan Modra [Mon, 13 Jan 2020 07:28:02 +0000 (17:58 +1030)]
asan: ns32k: wild memory write

index_offset isn't set up for "sfsr", resulting in a random offset
being used when trying to disassemble the following.

 .byte 0x3e, 0xf7, 0x07, 0x00

* ns32k-dis.c (Is_gen): Use strchr, add 'f'.
(print_insn_ns32k): Adjust ioffset for 'f' index_offset.

4 years agoubsan: wasm32: signed integer overflow
Alan Modra [Mon, 13 Jan 2020 03:57:19 +0000 (14:27 +1030)]
ubsan: wasm32: signed integer overflow

The signed integer overflow occurred when adding one to target_count
  for (i = 0; i < target_count + 1; i++)
but that's the least of the worries here.  target_count was long and i
int, leading to the possibility of a loop that never ended.

So to avoid this type of vulnerability, this patch uses what I believe
to be the proper types for arguments of various wasm32 opcodes, rather
than using "long" which may change in size.

gas/
* testsuite/gas/wasm32/allinsn.d: Update expected output.
opcodes/
* wasm32-dis.c (print_insn_wasm32): Localise variables.  Store
result of wasm_read_leb128 in a uint64_t and check that bits
are not lost when copying to other locals.  Use uint32_t for
most locals.  Use PRId64 when printing int64_t.

4 years agoscore formatting
Alan Modra [Mon, 13 Jan 2020 00:48:36 +0000 (11:18 +1030)]
score formatting

* score-dis.c: Formatting.
* score7-dis.c: Formatting.

4 years agoubsan: score: left shift of negative value
Alan Modra [Mon, 13 Jan 2020 00:16:55 +0000 (10:46 +1030)]
ubsan: score: left shift of negative value

* score-dis.c (print_insn_score48): Use unsigned variables for
unsigned values.  Don't left shift negative values.
(print_insn_score32): Likewise.
* score7-dis.c (print_insn_score32, print_insn_score16): Likewise.

4 years agoubsan: alpha-vma: timeout
Alan Modra [Sun, 12 Jan 2020 23:40:41 +0000 (10:10 +1030)]
ubsan: alpha-vma: timeout

* vms-alpha.c (_bfd_vms_slurp_egsd): Ensure minimum size even
for "ignored" records.

4 years agoMemory leaks and ineffective bounds checking in wasm_scan
Alan Modra [Sun, 12 Jan 2020 21:42:18 +0000 (08:12 +1030)]
Memory leaks and ineffective bounds checking in wasm_scan

It's always a bad idea to perform arithmetic on an unknown value read
from an object file before comparing against bounds.  Code like the
following attempting to bounds check "len", a 64-bit value, isn't
effective because the pointer arithmetic ignores the high 32 bits when
compiled for a 32-bit host.

      READ_LEB128 (len, p, end);
      if (p + len < p || p + len > end)
        goto error_return;

Instead, perform any arithmetic on known values where we don't need to
worry about overflows:

      READ_LEB128 (len, p, end);
      if (len > (size_t) (end - p))
        goto error_return;

I'll note that this check does do things the right way:

  READ_LEB128 (symcount, p, end);
  /* Sanity check: each symbol has at least two bytes.  */
  if (symcount > payload_size / 2)
    return FALSE;

"symcount * 2 > payload_size" would be wrong since the multiply could
overflow.

* wasm-module.c (wasm_scan_name_function_section): Formatting.
Delete asect name check.  Move asect NULL check to wasm_object_p.
Correct bounds check of sizes against end.  Replace uses of
bfd_zalloc with bfd_alloc, zeroing only necessary bytes.  Use
just one bfd_release.
(wasm_scan): Don't use malloc/strdup for section names,
bfd_alloc instead.  Simplify code prefixing section name.
Formatting.  Don't attempt to free memory here..
(wasm_object_p): ..do so here.  Formatting.

4 years agotic4x: sign extension using shifts
Alan Modra [Sun, 12 Jan 2020 09:46:22 +0000 (20:16 +1030)]
tic4x: sign extension using shifts

Don't do that.  Especially don't use shift counts that assume the type
being shifted is 32 bits when the type is long/unsigned long.  Also
reverts part of a change I made on 2019-12-11 to tic4x_print_register
that on closer inspection turns out to be unnecessary.

include/
* opcode/tic4x.h (EXTR): Delete.
(EXTRU, EXTRS, INSERTU, INSERTS): Rewrite without zero/sign
extension using shifts.  Do trim INSERTU value to specified bitfield.
opcodes/
* tic4x-dis.c (tic4x_print_register): Remove dead code.
gas/
* config/tc-tic4x.c (tic4x_operands_match): Correct tic3x trap
insertion.

4 years agoubsan: fr30: left shift of negative value
Alan Modra [Sat, 11 Jan 2020 02:02:11 +0000 (12:32 +1030)]
ubsan: fr30: left shift of negative value

cpu/
* fr30.cpu (f-disp9, f-disp10, f-s10, f-rel9, f-rel12): Don't
left shift signed values.
opcodes/
* fr30-ibld.c: Regenerate.

4 years agoubsan: xgate: left shift of negative value
Alan Modra [Sat, 11 Jan 2020 01:53:47 +0000 (12:23 +1030)]
ubsan: xgate: left shift of negative value

* xgate-dis.c (print_insn): Don't left shift signed value.
(ripBits): Formatting, use 1u.

4 years agoAutomatic date update in version.in
GDB Administrator [Mon, 13 Jan 2020 00:00:20 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agogdb: include gdb_wait.h in gdb_wait.c
Simon Marchi [Sun, 12 Jan 2020 20:11:07 +0000 (15:11 -0500)]
gdb: include gdb_wait.h in gdb_wait.c

When building for mingw with -Wmissing-declarations, I get:

  CXX    gdbsupport/gdb_wait.o
/binutils-gdb/gdb/gdbsupport/gdb_wait.c:52:1: error: no previous declaration for 'int windows_status_to_termsig(long unsigned int)' [-Wer
ror=missing-declarations]
   52 | windows_status_to_termsig (unsigned long status)
      | ^~~~~~~~~~~~~~~~~~~~~~~~~

Make gdb_wait.c include gdb_wait.h to fix it.

gdb/ChangeLog:

* gdbsupport/gdb_wait.c: Include gdb_wait.h.

4 years agogdbserver: include linux-arm-tdesc.h in linux-arm-tdesc.c
Simon Marchi [Sun, 12 Jan 2020 17:07:30 +0000 (12:07 -0500)]
gdbserver: include linux-arm-tdesc.h in linux-arm-tdesc.c

When building with -Wmissing-declarations, I get:

  CXX    linux-arm-tdesc.o
/binutils-gdb/gdb/gdbserver/linux-arm-tdesc.c:29:1: error: no previous declaration for 'const target_desc* arm_linux_read_description(arm_fp_type)' [-Werror=missing-declarations]
   29 | arm_linux_read_description (arm_fp_type fp_type)
      | ^~~~~~~~~~~~~~~~~~~~~~~~~~
/binutils-gdb/gdb/gdbserver/linux-arm-tdesc.c:49:1: error: no previous declaration for 'arm_fp_type arm_linux_get_tdesc_fp_type(const target_desc*)' [-Werror=missing-declarations]
   49 | arm_linux_get_tdesc_fp_type (const target_desc *tdesc)
      | ^~~~~~~~~~~~~~~~~~~~~~~~~~~

Include linux-arm-tdesc.h in linux-arm-tdesc.c to fix it.  And because
linux-arm-tdesc.h uses the arm_fp_type, it should include arch/arm.h as
well.

gdb/gdbserver/ChangeLog:

* linux-arm-tdesc.c: Include linux-arm-tdesc.h.
* linux-arm-tdesc.h: Include arch/arm.h.

4 years agogdbserver: make aarch64_write_goto_address static
Simon Marchi [Sun, 12 Jan 2020 16:06:23 +0000 (11:06 -0500)]
gdbserver: make aarch64_write_goto_address static

This function is only used in this file, so make it static.  It fixes
this error, when building with -Wmissing-declarations:

  CXX    linux-aarch64-low.o
/home/simark/src/binutils-gdb/gdb/gdbserver/linux-aarch64-low.c:2642:1: error: no previous declaration for 'void aarch64_write_goto_address(CORE_ADDR, CORE_ADDR, int)' [-Werror=missing-declarations]
 aarch64_write_goto_address (CORE_ADDR from, CORE_ADDR to, int size)
 ^~~~~~~~~~~~~~~~~~~~~~~~~~

gdb/gdbserver/ChangeLog:

* linux-aarch64-low.c (aarch64_write_goto_address): Make static.

4 years agogdbserver: include aarch32/aarch64 header file in corresponding source file
Simon Marchi [Sun, 12 Jan 2020 16:06:21 +0000 (11:06 -0500)]
gdbserver: include aarch32/aarch64 header file in corresponding source file

When building gdbserver for an aarch64 host with -Wmissing-declarations,
I see:

  CXX    linux-aarch32-tdesc.o
/home/simark/src/binutils-gdb/gdb/gdbserver/linux-aarch32-tdesc.c:28:1: error: no previous declaration for 'const target_desc* aarch32_linux_read_description()' [-Werror=missing-declarations]
 aarch32_linux_read_description ()
 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/simark/src/binutils-gdb/gdb/gdbserver/linux-aarch32-tdesc.c:43:1: error: no previous declaration for 'bool is_aarch32_linux_description(const target_desc*)' [-Werror=missing-declarations]
 is_aarch32_linux_description (const target_desc *tdesc)
 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~

  CXX    linux-aarch64-tdesc.o
/home/simark/src/binutils-gdb/gdb/gdbserver/linux-aarch64-tdesc.c:32:1: error: no previous declaration for 'const target_desc* aarch64_linux_read_description(uint64_t, bool)' [-Werror=missing-declarations]
 aarch64_linux_read_description (uint64_t vq, bool pauth_p)
 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Fix it by including linux-aarch32-tdesc.h in linux-aarch32-tdesc.c and
linux-aarch64-tdesc.h in linux-aarch64-tdesc.c.

gdb/gdbserver/ChangeLog:

* linux-aarch32-tdesc.c: Include linux-aarch32-tdesc.h.
* linux-aarch64-tdesc.c: Include linux-aarch64-tdesc.h.

4 years agoRemove last traces of discard_all_inferiors
Pedro Alves [Sun, 12 Jan 2020 00:40:02 +0000 (00:40 +0000)]
Remove last traces of discard_all_inferiors

The multi-target patch should have removed all traces of
discard_all_inferiors, but somehow one use stayed behind along with
the definition of the function.

discard_all_inferiors is bad now because it blindly exits inferiors of
all target connections.  It's best to remove it.

gdb/ChangeLog:
2020-01-12  Pedro Alves  <palves@redhat.com>

* bsd-kvm.c (bsd_kvm_target::close): Call exit_inferior_silent
directly for the current inferior instead of
discard_all_inferiors.
(discard_all_inferiors): Delete.

4 years agoAutomatic date update in version.in
GDB Administrator [Sun, 12 Jan 2020 00:01:17 +0000 (00:01 +0000)]
Automatic date update in version.in

4 years agoMake TUI borders respect "set style enabled"
Tom Tromey [Sat, 4 Jan 2020 21:35:02 +0000 (14:35 -0700)]
Make TUI borders respect "set style enabled"

When adding support for styling the TUI borders, I neglected to have
this code check cli_styling.  As a result, "set style enabled off"
does not affect the borders.

This patch fixes this oversight.  While doing this, I found that
running gdb without an executable, enabling the TUI, and then trying
"set style enabled off" would fail with the mysterious "No registers".
The fix for this is to use deprecated_safe_get_selected_frame in
tui_source_window_base::refill.

gdb/ChangeLog
2020-01-11  Tom Tromey  <tom@tromey.com>

* tui/tui-wingeneral.c (box_win): Check cli_styling.
* tui/tui-winsource.c (tui_source_window_base::refill): Use
deprecated_safe_get_selected_frame.

Change-Id: I36acda25dd9014d994d366b4a0e8faee9d95d0f8

4 years agoAutomatic date update in version.in
GDB Administrator [Sat, 11 Jan 2020 00:00:11 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agoSwitch the inferior before outputting its id in "info inferiors"
Tankut Baris Aktemur [Fri, 10 Jan 2020 20:06:17 +0000 (20:06 +0000)]
Switch the inferior before outputting its id in "info inferiors"

GDB uses the 'current_top_target' when displaying the description of
an inferior.  This leads to same target being used for each inferior
and, in turn, yields incorrect output when the inferior has a target
that is supposed to give a specialized output.  For instance, the
remote target outputs "Remote target" instead of "process XYZ" as the
description if the multi-process feature is not supported or turned
off.

E.g.: Suppose we have a native and a remote target, and the native is
the current inferior.  The remote target does not support multi-process.
For "info inferiors", we would expect to see:

~~~
(gdb) i inferiors
  Num  Description       Connection       Executable
* 1    process 29060     1 (native)       /a/path
  2    Remote target     2 (remote ...)
~~~

but instead we get

~~~
(gdb) i inferiors
  Num  Description       Connection       Executable
* 1    process 29060     1 (native)       /a/path
  2    process 42000     2 (remote ...)
~~~

Similarly, if the current inferior is the remote one, we would expect
to see

~~~
(gdb) i inferiors
  Num  Description       Connection       Executable
  1    process 29060     1 (native)       /a/path
* 2    Remote target     2 (remote ...)
~~~

but we get

~~~
(gdb) i inferiors
  Num  Description       Connection       Executable
* 1    Remote target     1 (native)       /a/path
  2    Remote target     2 (remote ...)
~~~

With this patch, we switch to the inferior when outputting its
description, so that the current_top_target will be aligned to the
inferior we are displaying.

For testing, this patch expands the "info inferiors" test for the
multi-target feature.  The test was checking for the output of the
info commands after setup, only when the current inferior is the last
added inferior.

This patch does the following to the testcase:

1. The "info inferiors" and "info connections" test is extracted out
   from the "setup" procedure to a separate procedure.

2. The test is enriched to check the output after switching to each
   inferior, not just the last one.

3. The test is performed twice; one for when the multi-process feature
   is turned on, one for off.

gdb/ChangeLog:
2020-01-10  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* inferior.c (print_inferior): Switch inferior before printing it.

gdb/testsuite/ChangeLog:
2020-01-10  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* gdb.multi/multi-target.exp (setup): Factor out "info
connections" and "info inferiors" tests to ...
(test_info_inferiors): ... this new procedure.
(top level): Run new "info-inferiors" tests.

4 years agoSwitch the inferior too in switch_to_program_space_and_thread
Pedro Alves [Fri, 10 Jan 2020 20:06:16 +0000 (20:06 +0000)]
Switch the inferior too in switch_to_program_space_and_thread

With multi-target, each inferior now has its own target connection.
The problem in switch_to_program_space_and_thread is that in the
current state GDB switches to "no thread" and also sets the program
space but because the inferior is not switched, potentially an
incorrect target remains selected.

Here is a sample scenario that exploits this flow:

On terminal 1, start a gdbserver on a program named foo:

 $ gdbserver :1234 ./foo

On terminal 2, start gdb on a program named bar.  Suppose foo and bar
are compiled from foo.c and bar.c.  They are completely separate.  So,
bar.c:2 has no meaning for foo.

 $ gdb -q ./bar
 Reading symbols from ./bar...
 (gdb) add-inferior
 [New inferior 2]
 Added inferior 2
 (gdb) inferior 2
 [Switching to inferior 2 [<null>] (<noexec>)]
 (gdb) target remote :1234
 ...
 (gdb) set debug remote 2
 (gdb) break bar.c:2
 Sending packet: $Hgp0.0#ad...Packet received: OK
 Sending packet: $m5fa,12#f8...Packet received: E01
 Sending packet: $m5fa,1#c6...Packet received: E01
 Sending packet: $m5fb,3#c9...Packet received: E01
 Sending packet: $m5fe,1#ca...Packet received: E01
 Breakpoint 1 at 0x5fe: file bar.c, line 2.
 (gdb)

Here we have an unnecessary sending of the packets to the gdbserver.

With this fix in progspace-and-thread.c, we'll get this:

 (gdb) break bar.c:2
 Breakpoint 1 at 0x5fe: file bar.c, line 2.
 (gdb)

Now there is no sending of the packets to gdbserver.

The changes around clear_symtab_users calls are necessary because
otherwise we regress gdb.base/step-over-exit.exp, hitting the new
assertion in switch_to_program_space_and_thread.  The problem is, a
forked child terminates, and when GDB decides to auto-purge that
inferior, GDB tries to switch to the pspace of that no-longer-existing
inferior.

The root of the problem is within the program_space destructor:

program_space::~program_space ()
{
...
  set_current_program_space (this);        # (1)
...
  breakpoint_program_space_exit (this);    # (2)
...
  free_all_objfiles ();                    # (3)
...
}

We get here from delete_inferior -> delete_program_space.

So we're deleting an inferior, and the inferior to be
deleted is no longer in the inferior list.

At (2), we've deleted all the breakpoints and locations for the
program space being deleted.

The crash happens while doing a breakpoint re-set, called by
clear_symtab_users at the tail end of (3).  That is, while recreating
breakpoints for the current program space, which is the program space
we're tearing down.  During breakpoint re-set, we try to switch to the
new location's pspace (the current pspace set in (1), so the pspace
we're tearing down) with switch_to_program_space_and_thread, and that
hits the failed assertion.  It's the fact that we recreate breakpoints
in the program_space destructor that is the latent bug here.  Just
don't do that, and we don't end up in the crash situation.

My first approach to fix this added a symfile_add_flags parameter to
program_space::free_all_objfiles, and then passed that down to
clear_symtab_users.  The program_space dtor would then pass down
SYMFILE_DEFER_BP_RESET to free_all_objfiles.  I couldn't help feeling
that adding that parameter to free_all_objfiles looked a little
awkward, so I settled on something a little different -- hoist the
clear_symtab_users call to the callers.  There are only two callers.
I felt that that didn't look as odd, particularly since
remove_symbol_file_command also does:

  objf->unlink ();
  clear_symtab_users (0);

I.e., objfile deletion is already separate from calling
clear_symtab_users in some places.

gdb/ChangeLog:
2020-01-10  Aleksandar Paunovic  <aleksandar.paunovic@intel.com>
    Pedro Alves  <palves@redhat.com>

* progspace-and-thread.c (switch_to_program_space_and_thread):
Assert there's an inferior for PSPACE.  Use
switch_to_inferior_no_thread to switch the inferior too.
* progspace.c (program_space::~program_space): Call
clear_symtab_users here, with SYMFILE_DEFER_BP_RESET.
(program_space::free_all_objfiles): Don't call clear_symtab_users
here.
* symfile.c (symbol_file_clear): Call clear_symtab_users here.

gdb/testsuite/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.server/bkpt-other-inferior.exp: New file.

4 years agoMulti-target: NEWS and user manual
Pedro Alves [Fri, 10 Jan 2020 20:06:15 +0000 (20:06 +0000)]
Multi-target: NEWS and user manual

This commit documents the new multi-target features in both NEWS and
user manual.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* NEWS: Mention multi-target debugging, "info connections", and
"add-inferior -no-connection".

gdb/doc/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.texinfo (Starting): Say "current inferior not connected"
instead of "GDB not connected".
(Inferiors and Programs): Rename node to ...
(Inferiors Connections and Programs): ... this.  Update all
references.  Talk about multiple target connections.  Update "info
inferiors" info to mention the connections column.  Describe "info
connections".  Document "add-inferior -no-connection".
* guile.texi, python.texi: Update cross references.

4 years agoRequire always-non-stop for multi-target resumptions
Pedro Alves [Fri, 10 Jan 2020 20:06:14 +0000 (20:06 +0000)]
Require always-non-stop for multi-target resumptions

Currently, we can only support resuming multiple targets at the same
time if all targets are in non-stop mode (or user-visible all-stop
mode with target backend in non-stop mode).

This patch makes GDB error out if the user tries to resume more than
one target at the same time and one of the resumed targets isn't in
non-stop mode:

 (gdb) info inferiors
   Num  Description       Connection                Executable
   1    process 15303     1 (native)                a.out
 * 2    process 15286     2 (extended-remote :9999) a.out
 (gdb) set schedule-multiple on
 (gdb) c
 Continuing.
 Connection 2 (extended-remote :9999) does not support multi-target resumption.

This is here later in the series instead of in the main multi-target
patch because it depends the previous patch, which added
process_stratum_target::connection_string().

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* infrun.c: Include "target-connection.h".
(check_multi_target_resumption): New.
(proceed): Call it.
* target-connection.c (make_target_connection_string): Make
extern.
* target-connection.h (make_target_connection_string): Declare.

4 years agoAdd "info connections" command, "info inferiors" connection number/string
Pedro Alves [Fri, 10 Jan 2020 20:06:14 +0000 (20:06 +0000)]
Add "info connections" command, "info inferiors" connection number/string

This commit extends the CLI a bit for multi-target, in three ways.

#1 - New "info connections" command.

This is a new command that lists the open connections (process_stratum
targets).  For example, if you're debugging two remote connections, a
couple local/native processes, and a core dump, all at the same time,
you might see something like this:

 (gdb) info connections
   Num  What                     Description
   1    remote 192.168.0.1:9999  Remote serial target in gdb-specific protocol
   2    remote 192.168.0.2:9998  Remote serial target in gdb-specific protocol
 * 3    native                   Native process
   4    core                     Local core dump file

#2 - New "info inferiors" "Connection" column

You'll also see a new matching "Connection" column in "info
inferiors", showing you which connection an inferior is bound to:

 (gdb) info inferiors
   Num  Description       Connection                   Executable
   1    process 18526     1 (remote 192.168.0.1:9999)  target:/tmp/a.out
   2    process 18531     2 (remote 192.168.0.2:9998)  target:/tmp/a.out
   3    process 19115     3 (native)                   /tmp/prog1
   4    process 6286      4 (core)                     myprogram
 * 5    process 19122     3 (native)                   /bin/hello

#3 - Makes "add-inferior" show the inferior's target connection

"add-inferior" now shows you the connection you've just bound the
inferior to, which is the current process_stratum target:

 (gdb) add-inferior
 [New inferior 2]
 Added inferior 2 on connection 1 (extended-remote localhost:2346)

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* Makefile.in (COMMON_SFILES): Add target-connection.c.
* inferior.c (uiout_field_connection): New function.
(print_inferior): Add new "connection-id" column.
(add_inferior_command): Show connection number/string of added
inferior.
* process-stratum-target.h
(process_stratum_target::connection_string): New virtual method.
(process_stratum_target::connection_number): New field.
* remote.c (remote_target::connection_string): New override.
* target-connection.c: New file.
* target-connection.h: New file.
* target.c (decref_target): Remove process_stratum targets from
the connection list.
(target_stack::push): Add process_stratum targets to the
connection list.

gdb/testsuite/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.base/kill-detach-inferiors-cmd.exp: Adjust expected output
of "add-inferior".
* gdb.base/quit-live.exp: Likewise.
* gdb.base/remote-exec-file.exp: Likewise.
* gdb.guile/scm-progspace.exp: Likewise.
* gdb.linespec/linespec.exp: Likewise.
* gdb.mi/new-ui-mi-sync.exp: Likewise.
* gdb.mi/user-selected-context-sync.exp: Likewise.
* gdb.multi/multi-target.exp (setup): Add "info connection" and
"info inferiors" tests.
* gdb.multi/remove-inferiors.exp: Adjust expected output of
"add-inferior".
* gdb.multi/watchpoint-multi.exp: Likewise.
* gdb.python/py-inferior.exp: Likewise.
* gdb.server/extended-remote-restart.exp: Likewise.
* gdb.threads/fork-plus-threads.exp: Adjust expected output of
"info inferiors".
* gdb.threads/forking-threads-plus-breakpoint.exp: Likewise.
* gdb.trace/report.exp: Likewise.

4 years agoRevert 'Remove unused struct serial::name field'
Pedro Alves [Fri, 10 Jan 2020 20:06:11 +0000 (20:06 +0000)]
Revert 'Remove unused struct serial::name field'

This commit reverts:

 commit 5f5219fc34f7557296272230123a3837960a6f09
 Author:     Pedro Alves <palves@redhat.com>
 AuthorDate: Tue Apr 12 16:49:30 2016 +0100

     Remove unused struct serial::name field

The following patches will add uses for the field.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

Revert:
2016-04-12  Pedro Alves  <palves@redhat.com>
* serial.c (serial_open, serial_fdopen_ops, do_serial_close):
Remove references to name.
* serial.h (struct serial) <name>: Delete.

4 years agogdbarch-selftests.c: No longer error out if debugging something
Pedro Alves [Fri, 10 Jan 2020 20:06:10 +0000 (20:06 +0000)]
gdbarch-selftests.c: No longer error out if debugging something

Since each inferior has its own target stack, the stratum condition
for the "error out if debugging something" check is always false.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdbarch-selftests.c (register_to_value_test): Remove "target
already pushed" check.

4 years agoAdd multi-target tests
Pedro Alves [Fri, 10 Jan 2020 20:06:09 +0000 (20:06 +0000)]
Add multi-target tests

This adds a testcase exercising multi-target features.  It spawns 6
inferiors, like this:

 inferior 1 -> native
 inferior 2 -> extended-remote 1
 inferior 3 -> core
 inferior 4 -> native
 inferior 5 -> extended-remote 2
 inferior 6 -> core

and then tests various details, including:

 - running to breakpoints

 - interrupting with Ctrl-C and "interrupt -a"

 - "next" bouncing between two breakpoints in two threads running in
   different targets.

 - since we have cores and live inferiors mixed in the same session,
   this makes sure that gdb doesn't try to remove a core dump's
   threads.

 - all-stop and non-stop modes.

This testcase caught a _lot_ of bugs in development.

gdb/testsuite/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.multi/multi-target.c: New file.
* gdb.multi/multi-target.exp: New file.
* lib/gdbserver-support.exp (gdb_target_cmd): Handle "Non-stop
mode requested, but remote does not support non-stop".

4 years agoMulti-target support
Pedro Alves [Fri, 10 Jan 2020 20:06:08 +0000 (20:06 +0000)]
Multi-target support

This commit adds multi-target support to GDB.  What this means is that
with this commit, GDB can now be connected to different targets at the
same time.  E.g., you can debug a live native process and a core dump
at the same time, connect to multiple gdbservers, etc.

Actually, the word "target" is overloaded in gdb.  We already have a
target stack, with pushes several target_ops instances on top of one
another.  We also have "info target" already, which means something
completely different to what this patch does.

So from here on, I'll be using the "target connections" term, to mean
an open process_stratum target, pushed on a target stack.  This patch
makes gdb have multiple target stacks, and multiple process_stratum
targets open simultaneously.  The user-visible changes / commands will
also use this terminology, but of course it's all open to debate.

User-interface-wise, not that much changes.  The main difference is
that each inferior may have its own target connection.

A target connection (e.g., a target extended-remote connection) may
support debugging multiple processes, just as before.

Say you're debugging against gdbserver in extended-remote mode, and
you do "add-inferior" to prepare to spawn a new process, like:

 (gdb) target extended-remote :9999
 ...
 (gdb) start
 ...
 (gdb) add-inferior
 Added inferior 2
 (gdb) inferior 2
 [Switching to inferior 2 [<null>] (<noexec>)]
 (gdb) file a.out
 ...
 (gdb) start
 ...

At this point, you have two inferiors connected to the same gdbserver.

With this commit, GDB will maintain a target stack per inferior,
instead of a global target stack.

To preserve the behavior above, by default, "add-inferior" makes the
new inferior inherit a copy of the target stack of the current
inferior.  Same across a fork - the child inherits a copy of the
target stack of the parent.  While the target stacks are copied, the
targets themselves are not.  Instead, target_ops is made a
refcounted_object, which means that target_ops instances are
refcounted, which each inferior counting for a reference.

What if you want to create an inferior and connect it to some _other_
target?  For that, this commit introduces a new "add-inferior
-no-connection" option that makes the new inferior not share the
current inferior's target.  So you could do:

 (gdb) target extended-remote :9999
 Remote debugging using :9999
 ...
 (gdb) add-inferior -no-connection
 [New inferior 2]
 Added inferior 2
 (gdb) inferior 2
 [Switching to inferior 2 [<null>] (<noexec>)]
 (gdb) info inferiors
   Num  Description       Executable
   1    process 18401     target:/home/pedro/tmp/main
 * 2    <null>
 (gdb) tar extended-remote :10000
 Remote debugging using :10000
 ...
 (gdb) info inferiors
   Num  Description       Executable
   1    process 18401     target:/home/pedro/tmp/main
 * 2    process 18450     target:/home/pedro/tmp/main
 (gdb)

A following patch will extended "info inferiors" to include a column
indicating which connection an inferior is bound to, along with a
couple other UI tweaks.

Other than that, debugging is the same as before.  Users interact with
inferiors and threads as before.  The only difference is that
inferiors may be bound to processes running in different machines.

That's pretty much all there is to it in terms of noticeable UI
changes.

On to implementation.

Since we can be connected to different systems at the same time, a
ptid_t is no longer a unique identifier.  Instead a thread can be
identified by a pair of ptid_t and 'process_stratum_target *', the
later being the instance of the process_stratum target that owns the
process/thread.  Note that process_stratum_target inherits from
target_ops, and all process_stratum targets inherit from
process_stratum_target.  In earlier patches, many places in gdb were
converted to refer to threads by thread_info pointer instead of
ptid_t, but there are still places in gdb where we start with a
pid/tid and need to find the corresponding inferior or thread_info
objects.  So you'll see in the patch many places adding a
process_stratum_target parameter to functions that used to take only a
ptid_t.

Since each inferior has its own target stack now, we can always find
the process_stratum target for an inferior.  That is done via a
inf->process_target() convenience method.

Since each inferior has its own target stack, we need to handle the
"beneath" calls when servicing target calls.  The solution I settled
with is just to make sure to switch the current inferior to the
inferior you want before making a target call.  Not relying on global
context is just not feasible in current GDB.  Fortunately, there
aren't that many places that need to do that, because generally most
code that calls target methods already has the current context
pointing to the right inferior/thread.  Note, to emphasize -- there's
no method to "switch to this target stack".  Instead, you switch the
current inferior, and that implicitly switches the target stack.

In some spots, we need to iterate over all inferiors so that we reach
all target stacks.

Native targets are still singletons.  There's always only a single
instance of such targets.

Remote targets however, we'll have one instance per remote connection.

The exec target is still a singleton.  There's only one instance.  I
did not see the point of instanciating more than one exec_target
object.

After vfork, we need to make sure to push the exec target on the new
inferior.  See exec_on_vfork.

For type safety, functions that need a {target, ptid} pair to identify
a thread, take a process_stratum_target pointer for target parameter
instead of target_ops *.  Some shared code in gdb/nat/ also need to
gain a target pointer parameter.  This poses an issue, since gdbserver
doesn't have process_stratum_target, only target_ops.  To fix this,
this commit renames gdbserver's target_ops to process_stratum_target.
I think this makes sense.  There's no concept of target stack in
gdbserver, and gdbserver's target_ops really implements a
process_stratum-like target.

The thread and inferior iterator functions also gain
process_stratum_target parameters.  These are used to be able to
iterate over threads and inferiors of a given target.  Following usual
conventions, if the target pointer is null, then we iterate over
threads and inferiors of all targets.

I tried converting "add-inferior" to the gdb::option framework, as a
preparatory patch, but that stumbled on the fact that gdb::option does
not support file options yet, for "add-inferior -exec".  I have a WIP
patchset that adds that, but it's not a trivial patch, mainly due to
need to integrate readline's filename completion, so I deferred that
to some other time.

In infrun.c/infcmd.c, the main change is that we need to poll events
out of all targets.  See do_target_wait.  Right after collecting an
event, we switch the current inferior to an inferior bound to the
target that reported the event, so that target methods can be used
while handling the event.  This makes most of the code transparent to
multi-targets.  See fetch_inferior_event.

infrun.c:stop_all_threads is interesting -- in this function we need
to stop all threads of all targets.  What the function does is send an
asynchronous stop request to all threads, and then synchronously waits
for events, with target_wait, rinse repeat, until all it finds are
stopped threads.  Now that we have multiple targets, it's not
efficient to synchronously block in target_wait waiting for events out
of one target.  Instead, we implement a mini event loop, with
interruptible_select, select'ing on one file descriptor per target.
For this to work, we need to be able to ask the target for a waitable
file descriptor.  Such file descriptors already exist, they are the
descriptors registered in the main event loop with add_file_handler,
inside the target_async implementations.  This commit adds a new
target_async_wait_fd target method that just returns the file
descriptor in question.  See wait_one / stop_all_threads in infrun.c.

The 'threads_executing' global is made a per-target variable.  Since
it is only relevant to process_stratum_target targets, this is where
it is put, instead of in target_ops.

You'll notice that remote.c includes some FIXME notes.  These refer to
the fact that the global arrays that hold data for the remote packets
supported are still globals.  For example, if we connect to two
different servers/stubs, then each might support different remote
protocol features.  They might even be different architectures, like
e.g., one ARM baremetal stub, and a x86 gdbserver, to debug a
host/controller scenario as a single program.  That isn't going to
work correctly today, because of said globals.  I'm leaving fixing
that for another pass, since it does not appear to be trivial, and I'd
rather land the base work first.  It's already useful to be able to
debug multiple instances of the same server (e.g., a distributed
cluster, where you have full control over the servers installed), so I
think as is it's already reasonable incremental progress.

Current limitations:

 - You can only resume more that one target at the same time if all
   targets support asynchronous debugging, and support non-stop mode.
   It should be possible to support mixed all-stop + non-stop
   backends, but that is left for another time.  This means that
   currently in order to do multi-target with gdbserver you need to
   issue "maint set target-non-stop on".  I would like to make that
   mode be the default, but we're not there yet.  Note that I'm
   talking about how the target backend works, only.  User-visible
   all-stop mode works just fine.

 - As explained above, connecting to different remote servers at the
   same time is likely to produce bad results if they don't support the
   exact set of RSP features.

FreeBSD updates courtesy of John Baldwin.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>
    John Baldwin  <jhb@FreeBSD.org>

* aarch64-linux-nat.c
(aarch64_linux_nat_target::thread_architecture): Adjust.
* ada-tasks.c (print_ada_task_info): Adjust find_thread_ptid call.
(task_command_1): Likewise.
* aix-thread.c (sync_threadlists, aix_thread_target::resume)
(aix_thread_target::wait, aix_thread_target::fetch_registers)
(aix_thread_target::store_registers)
(aix_thread_target::thread_alive): Adjust.
* amd64-fbsd-tdep.c: Include "inferior.h".
(amd64fbsd_get_thread_local_address): Pass down target.
* amd64-linux-nat.c (ps_get_thread_area): Use ps_prochandle
thread's gdbarch instead of target_gdbarch.
* break-catch-sig.c (signal_catchpoint_print_it): Adjust call to
get_last_target_status.
* break-catch-syscall.c (print_it_catch_syscall): Likewise.
* breakpoint.c (breakpoints_should_be_inserted_now): Consider all
inferiors.
(update_inserted_breakpoint_locations): Skip if inferiors with no
execution.
(update_global_location_list): When handling moribund locations,
find representative inferior for location's pspace, and use thread
count of its process_stratum target.
* bsd-kvm.c (bsd_kvm_target_open): Pass target down.
* bsd-uthread.c (bsd_uthread_target::wait): Use
as_process_stratum_target and adjust thread_change_ptid and
add_thread calls.
(bsd_uthread_target::update_thread_list): Use
as_process_stratum_target and adjust find_thread_ptid,
thread_change_ptid and add_thread calls.
* btrace.c (maint_btrace_packet_history_cmd): Adjust
find_thread_ptid call.
* corelow.c (add_to_thread_list): Adjust add_thread call.
(core_target_open): Adjust add_thread_silent and thread_count
calls.
(core_target::pid_to_str): Adjust find_inferior_ptid call.
* ctf.c (ctf_target_open): Adjust add_thread_silent call.
* event-top.c (async_disconnect): Pop targets from all inferiors.
* exec.c (add_target_sections): Push exec target on all inferiors
sharing the program space.
(remove_target_sections): Remove the exec target from all
inferiors sharing the program space.
(exec_on_vfork): New.
* exec.h (exec_on_vfork): Declare.
* fbsd-nat.c (fbsd_add_threads): Add fbsd_nat_target parameter.
Pass it down.
(fbsd_nat_target::update_thread_list): Adjust.
(fbsd_nat_target::resume): Adjust.
(fbsd_handle_debug_trap): Add fbsd_nat_target parameter.  Pass it
down.
(fbsd_nat_target::wait, fbsd_nat_target::post_attach): Adjust.
* fbsd-tdep.c (fbsd_corefile_thread): Adjust
get_thread_arch_regcache call.
* fork-child.c (gdb_startup_inferior): Pass target down to
startup_inferior and set_executing.
* gdbthread.h (struct process_stratum_target): Forward declare.
(add_thread, add_thread_silent, add_thread_with_info)
(in_thread_list): Add process_stratum_target parameter.
(find_thread_ptid(inferior*, ptid_t)): New overload.
(find_thread_ptid, thread_change_ptid): Add process_stratum_target
parameter.
(all_threads()): Delete overload.
(all_threads, all_non_exited_threads): Add process_stratum_target
parameter.
(all_threads_safe): Use brace initialization.
(thread_count): Add process_stratum_target parameter.
(set_resumed, set_running, set_stop_requested, set_executing)
(threads_are_executing, finish_thread_state): Add
process_stratum_target parameter.
(switch_to_thread): Use is_current_thread.
* i386-fbsd-tdep.c: Include "inferior.h".
(i386fbsd_get_thread_local_address): Pass down target.
* i386-linux-nat.c (i386_linux_nat_target::low_resume): Adjust.
* inf-child.c (inf_child_target::maybe_unpush_target): Remove
have_inferiors check.
* inf-ptrace.c (inf_ptrace_target::create_inferior)
(inf_ptrace_target::attach): Adjust.
* infcall.c (run_inferior_call): Adjust.
* infcmd.c (run_command_1): Pass target to
scoped_finish_thread_state.
(proceed_thread_callback): Skip inferiors with no execution.
(continue_command): Rename 'all_threads' local to avoid hiding
'all_threads' function.  Adjust get_last_target_status call.
(prepare_one_step): Adjust set_running call.
(signal_command): Use user_visible_resume_target.  Compare thread
pointers instead of inferior_ptid.
(info_program_command): Adjust to pass down target.
(attach_command): Mark target's 'thread_executing' flag.
(stop_current_target_threads_ns): New, factored out from ...
(interrupt_target_1): ... this.  Switch inferior before making
target calls.
* inferior-iter.h
(struct all_inferiors_iterator, struct all_inferiors_range)
(struct all_inferiors_safe_range)
(struct all_non_exited_inferiors_range): Filter on
process_stratum_target too.  Remove explicit.
* inferior.c (inferior::inferior): Push dummy target on target
stack.
(find_inferior_pid, find_inferior_ptid, number_of_live_inferiors):
Add process_stratum_target parameter, and pass it down.
(have_live_inferiors): Adjust.
(switch_to_inferior_and_push_target): New.
(add_inferior_command, clone_inferior_command): Handle
"-no-connection" parameter.  Use
switch_to_inferior_and_push_target.
(_initialize_inferior): Mention "-no-connection" option in
the help of "add-inferior" and "clone-inferior" commands.
* inferior.h: Include "process-stratum-target.h".
(interrupt_target_1): Use bool.
(struct inferior) <push_target, unpush_target, target_is_pushed,
find_target_beneath, top_target, process_target, target_at,
m_stack>: New.
(discard_all_inferiors): Delete.
(find_inferior_pid, find_inferior_ptid, number_of_live_inferiors)
(all_inferiors, all_non_exited_inferiors): Add
process_stratum_target parameter.
* infrun.c: Include "gdb_select.h" and <unordered_map>.
(target_last_proc_target): New global.
(follow_fork_inferior): Push target on new inferior.  Pass target
to add_thread_silent.  Call exec_on_vfork.  Handle target's
reference count.
(follow_fork): Adjust get_last_target_status call.  Also consider
target.
(follow_exec): Push target on new inferior.
(struct execution_control_state) <target>: New field.
(user_visible_resume_target): New.
(do_target_resume): Call target_async.
(resume_1): Set target's threads_executing flag.  Consider resume
target.
(commit_resume_all_targets): New.
(proceed): Also consider resume target.  Skip threads of inferiors
with no execution.  Commit resumtion in all targets.
(start_remote): Pass current inferior to wait_for_inferior.
(infrun_thread_stop_requested): Consider target as well.  Pass
thread_info pointer to clear_inline_frame_state instead of ptid.
(infrun_thread_thread_exit): Consider target as well.
(random_pending_event_thread): New inferior parameter.  Use it.
(do_target_wait): Rename to ...
(do_target_wait_1): ... this.  Add inferior parameter, and pass it
down.
(threads_are_resumed_pending_p, do_target_wait): New.
(prepare_for_detach): Adjust calls.
(wait_for_inferior): New inferior parameter.  Handle it.  Use
do_target_wait_1 instead of do_target_wait.
(fetch_inferior_event): Adjust.  Switch to representative
inferior.  Pass target down.
(set_last_target_status): Add process_stratum_target parameter.
Save target in global.
(get_last_target_status): Add process_stratum_target parameter and
handle it.
(nullify_last_target_wait_ptid): Clear 'target_last_proc_target'.
(context_switch): Check inferior_ptid == null_ptid before calling
inferior_thread().
(get_inferior_stop_soon): Pass down target.
(wait_one): Rename to ...
(poll_one_curr_target): ... this.
(struct wait_one_event): New.
(wait_one): New.
(stop_all_threads): Adjust.
(handle_no_resumed, handle_inferior_event): Adjust to consider the
event's target.
(switch_back_to_stepped_thread): Also consider target.
(print_stop_event): Update.
(normal_stop): Update.  Also consider the resume target.
* infrun.h (wait_for_inferior): Remove declaration.
(user_visible_resume_target): New declaration.
(get_last_target_status, set_last_target_status): New
process_stratum_target parameter.
* inline-frame.c (clear_inline_frame_state(ptid_t)): Add
process_stratum_target parameter, and use it.
(clear_inline_frame_state (thread_info*)): New.
* inline-frame.c (clear_inline_frame_state(ptid_t)): Add
process_stratum_target parameter.
(clear_inline_frame_state (thread_info*)): Declare.
* linux-fork.c (delete_checkpoint_command): Pass target down to
find_thread_ptid.
(checkpoint_command): Adjust.
* linux-nat.c (linux_nat_target::follow_fork): Switch to thread
instead of just tweaking inferior_ptid.
(linux_nat_switch_fork): Pass target down to thread_change_ptid.
(exit_lwp): Pass target down to find_thread_ptid.
(attach_proc_task_lwp_callback): Pass target down to
add_thread/set_running/set_executing.
(linux_nat_target::attach): Pass target down to
thread_change_ptid.
(get_detach_signal): Pass target down to find_thread_ptid.
Consider last target status's target.
(linux_resume_one_lwp_throw, resume_lwp)
(linux_handle_syscall_trap, linux_handle_extended_wait, wait_lwp)
(stop_wait_callback, save_stop_reason, linux_nat_filter_event)
(linux_nat_wait_1, resume_stopped_resumed_lwps): Pass target down.
(linux_nat_target::async_wait_fd): New.
(linux_nat_stop_lwp, linux_nat_target::thread_address_space): Pass
target down.
* linux-nat.h (linux_nat_target::async_wait_fd): Declare.
* linux-tdep.c (get_thread_arch_regcache): Pass target down.
* linux-thread-db.c (struct thread_db_info::process_target): New
field.
(add_thread_db_info): Save target.
(get_thread_db_info): New process_stratum_target parameter.  Also
match target.
(delete_thread_db_info): New process_stratum_target parameter.
Also match target.
(thread_from_lwp): Adjust to pass down target.
(thread_db_notice_clone): Pass down target.
(check_thread_db_callback): Pass down target.
(try_thread_db_load_1): Always push the thread_db target.
(try_thread_db_load, record_thread): Pass target down.
(thread_db_target::detach): Pass target down.  Always unpush the
thread_db target.
(thread_db_target::wait, thread_db_target::mourn_inferior): Pass
target down.  Always unpush the thread_db target.
(find_new_threads_callback, thread_db_find_new_threads_2)
(thread_db_target::update_thread_list): Pass target down.
(thread_db_target::pid_to_str): Pass current inferior down.
(thread_db_target::get_thread_local_address): Pass target down.
(thread_db_target::resume, maintenance_check_libthread_db): Pass
target down.
* nto-procfs.c (nto_procfs_target::update_thread_list): Adjust.
* procfs.c (procfs_target::procfs_init_inferior): Declare.
(proc_set_current_signal, do_attach, procfs_target::wait): Adjust.
(procfs_init_inferior): Rename to ...
(procfs_target::procfs_init_inferior): ... this and adjust.
(procfs_target::create_inferior, procfs_notice_thread)
(procfs_do_thread_registers): Adjust.
* ppc-fbsd-tdep.c: Include "inferior.h".
(ppcfbsd_get_thread_local_address): Pass down target.
* proc-service.c (ps_xfer_memory): Switch current inferior and
program space as well.
(get_ps_regcache): Pass target down.
* process-stratum-target.c
(process_stratum_target::thread_address_space)
(process_stratum_target::thread_architecture): Pass target down.
* process-stratum-target.h
(process_stratum_target::threads_executing): New field.
(as_process_stratum_target): New.
* ravenscar-thread.c
(ravenscar_thread_target::update_inferior_ptid): Pass target down.
(ravenscar_thread_target::wait, ravenscar_add_thread): Pass target
down.
* record-btrace.c (record_btrace_target::info_record): Adjust.
(record_btrace_target::record_method)
(record_btrace_target::record_is_replaying)
(record_btrace_target::fetch_registers)
(get_thread_current_frame_id, record_btrace_target::resume)
(record_btrace_target::wait, record_btrace_target::stop): Pass
target down.
* record-full.c (record_full_wait_1): Switch to event thread.
Pass target down.
* regcache.c (regcache::regcache)
(get_thread_arch_aspace_regcache, get_thread_arch_regcache): Add
process_stratum_target parameter and handle it.
(current_thread_target): New global.
(get_thread_regcache): Add process_stratum_target parameter and
handle it.  Switch inferior before calling target method.
(get_thread_regcache): Pass target down.
(get_thread_regcache_for_ptid): Pass target down.
(registers_changed_ptid): Add process_stratum_target parameter and
handle it.
(registers_changed_thread, registers_changed): Pass target down.
(test_get_thread_arch_aspace_regcache): New.
(current_regcache_test): Define a couple local test_target_ops
instances and use them for testing.
(readwrite_regcache): Pass process_stratum_target parameter.
(cooked_read_test, cooked_write_test): Pass mock_target down.
* regcache.h (get_thread_regcache, get_thread_arch_regcache)
(get_thread_arch_aspace_regcache): Add process_stratum_target
parameter.
(regcache::target): New method.
(regcache::regcache, regcache::get_thread_arch_aspace_regcache)
(regcache::registers_changed_ptid): Add process_stratum_target
parameter.
(regcache::m_target): New field.
(registers_changed_ptid): Add process_stratum_target parameter.
* remote.c (remote_state::supports_vCont_probed): New field.
(remote_target::async_wait_fd): New method.
(remote_unpush_and_throw): Add remote_target parameter.
(get_current_remote_target): Adjust.
(remote_target::remote_add_inferior): Push target.
(remote_target::remote_add_thread)
(remote_target::remote_notice_new_inferior)
(get_remote_thread_info): Pass target down.
(remote_target::update_thread_list): Skip threads of inferiors
bound to other targets.  (remote_target::close): Don't discard
inferiors.  (remote_target::add_current_inferior_and_thread)
(remote_target::process_initial_stop_replies)
(remote_target::start_remote)
(remote_target::remote_serial_quit_handler): Pass down target.
(remote_target::remote_unpush_target): New remote_target
parameter.  Unpush the target from all inferiors.
(remote_target::remote_unpush_and_throw): New remote_target
parameter.  Pass it down.
(remote_target::open_1): Check whether the current inferior has
execution instead of checking whether any inferior is live.  Pass
target down.
(remote_target::remote_detach_1): Pass down target.  Use
remote_unpush_target.
(extended_remote_target::attach): Pass down target.
(remote_target::remote_vcont_probe): Set supports_vCont_probed.
(remote_target::append_resumption): Pass down target.
(remote_target::append_pending_thread_resumptions)
(remote_target::remote_resume_with_hc, remote_target::resume)
(remote_target::commit_resume): Pass down target.
(remote_target::remote_stop_ns): Check supports_vCont_probed.
(remote_target::interrupt_query)
(remote_target::remove_new_fork_children)
(remote_target::check_pending_events_prevent_wildcard_vcont)
(remote_target::remote_parse_stop_reply)
(remote_target::process_stop_reply): Pass down target.
(first_remote_resumed_thread): New remote_target parameter.  Pass
it down.
(remote_target::wait_as): Pass down target.
(unpush_and_perror): New remote_target parameter.  Pass it down.
(remote_target::readchar, remote_target::remote_serial_write)
(remote_target::getpkt_or_notif_sane_1)
(remote_target::kill_new_fork_children, remote_target::kill): Pass
down target.
(remote_target::mourn_inferior): Pass down target.  Use
remote_unpush_target.
(remote_target::core_of_thread)
(remote_target::remote_btrace_maybe_reopen): Pass down target.
(remote_target::pid_to_exec_file)
(remote_target::thread_handle_to_thread_info): Pass down target.
(remote_target::async_wait_fd): New.
* riscv-fbsd-tdep.c: Include "inferior.h".
(riscv_fbsd_get_thread_local_address): Pass down target.
* sol2-tdep.c (sol2_core_pid_to_str): Pass down target.
* sol-thread.c (sol_thread_target::wait, ps_lgetregs, ps_lsetregs)
(ps_lgetfpregs, ps_lsetfpregs, sol_update_thread_list_callback):
Adjust.
* solib-spu.c (spu_skip_standalone_loader): Pass down target.
* solib-svr4.c (enable_break): Pass down target.
* spu-multiarch.c (parse_spufs_run): Pass down target.
* spu-tdep.c (spu2ppu_sniffer): Pass down target.
* target-delegates.c: Regenerate.
* target.c (g_target_stack): Delete.
(current_top_target): Return the current inferior's top target.
(target_has_execution_1): Refer to the passed-in inferior's top
target.
(target_supports_terminal_ours): Check whether the initial
inferior was already created.
(decref_target): New.
(target_stack::push): Incref/decref the target.
(push_target, push_target, unpush_target): Adjust.
(target_stack::unpush): Defref target.
(target_is_pushed): Return bool.  Adjust to refer to the current
inferior's target stack.
(dispose_inferior): Delete, and inline parts ...
(target_preopen): ... here.  Only dispose of the current inferior.
(target_detach): Hold strong target reference while detaching.
Pass target down.
(target_thread_name): Add assertion.
(target_resume): Pass down target.
(target_ops::beneath, find_target_at): Adjust to refer to the
current inferior's target stack.
(get_dummy_target): New.
(target_pass_ctrlc): Pass the Ctrl-C to the first inferior that
has a thread running.
(initialize_targets): Rename to ...
(_initialize_target): ... this.
* target.h: Include "gdbsupport/refcounted-object.h".
(struct target_ops): Inherit refcounted_object.
(target_ops::shortname, target_ops::longname): Make const.
(target_ops::async_wait_fd): New method.
(decref_target): Declare.
(struct target_ops_ref_policy): New.
(target_ops_ref): New typedef.
(get_dummy_target): Declare function.
(target_is_pushed): Return bool.
* thread-iter.c (all_matching_threads_iterator::m_inf_matches)
(all_matching_threads_iterator::all_matching_threads_iterator):
Handle filter target.
* thread-iter.h (struct all_matching_threads_iterator, struct
all_matching_threads_range, class all_non_exited_threads_range):
Filter by target too.  Remove explicit.
* thread.c (threads_executing): Delete.
(inferior_thread): Pass down current inferior.
(clear_thread_inferior_resources): Pass down thread pointer
instead of ptid_t.
(add_thread_silent, add_thread_with_info, add_thread): Add
process_stratum_target parameter.  Use it for thread and inferior
searches.
(is_current_thread): New.
(thread_info::deletable): Use it.
(find_thread_ptid, thread_count, in_thread_list)
(thread_change_ptid, set_resumed, set_running): New
process_stratum_target parameter.  Pass it down.
(set_executing): New process_stratum_target parameter.  Pass it
down.  Adjust reference to 'threads_executing'.
(threads_are_executing): New process_stratum_target parameter.
Adjust reference to 'threads_executing'.
(set_stop_requested, finish_thread_state): New
process_stratum_target parameter.  Pass it down.
(switch_to_thread): Also match inferior.
(switch_to_thread): New process_stratum_target parameter.  Pass it
down.
(update_threads_executing): Reimplement.
* top.c (quit_force): Pop targets from all inferior.
(gdb_init): Don't call initialize_targets.
* windows-nat.c (windows_nat_target) <get_windows_debug_event>:
Declare.
(windows_add_thread, windows_delete_thread): Adjust.
(get_windows_debug_event): Rename to ...
(windows_nat_target::get_windows_debug_event): ... this.  Adjust.
* tracefile-tfile.c (tfile_target_open): Pass down target.
* gdbsupport/common-gdbthread.h (struct process_stratum_target):
Forward declare.
(switch_to_thread): Add process_stratum_target parameter.
* mi/mi-interp.c (mi_on_resume_1): Add process_stratum_target
parameter.  Use it.
(mi_on_resume): Pass target down.
* nat/fork-inferior.c (startup_inferior): Add
process_stratum_target parameter.  Pass it down.
* nat/fork-inferior.h (startup_inferior): Add
process_stratum_target parameter.
* python/py-threadevent.c (py_get_event_thread): Pass target down.

gdb/gdbserver/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* fork-child.c (post_fork_inferior): Pass target down to
startup_inferior.
* inferiors.c (switch_to_thread): Add process_stratum_target
parameter.
* lynx-low.c (lynx_target_ops): Now a process_stratum_target.
* nto-low.c (nto_target_ops): Now a process_stratum_target.
* linux-low.c (linux_target_ops): Now a process_stratum_target.
* remote-utils.c (prepare_resume_reply): Pass the target to
switch_to_thread.
* target.c (the_target): Now a process_stratum_target.
(done_accessing_memory): Pass the target to switch_to_thread.
(set_target_ops): Ajust to use process_stratum_target.
* target.h (struct target_ops): Rename to ...
(struct process_stratum_target): ... this.
(the_target, set_target_ops): Adjust.
(prepare_to_access_memory): Adjust comment.
* win32-low.c (child_xfer_memory): Adjust to use
process_stratum_target.
(win32_target_ops): Now a process_stratum_target.

4 years agoFix reconnecting to a gdbserver already debugging multiple processes, II
Pedro Alves [Fri, 10 Jan 2020 20:05:54 +0000 (20:05 +0000)]
Fix reconnecting to a gdbserver already debugging multiple processes, II

Another bug exposed by gdb.server/extended-remote-restart.exp in the
multi-target work is that remote_target::start_remote can leave
inferior_ptid and current_inferior() out of sync:

 (top-gdb) p current_inferior_->pid
 $1 = 29541
 (top-gdb) p inferior_ptid
 $2 = {m_pid = 29540, m_lwp = 29540, m_tid = 0}

This is caused by writing to inferior_ptid directly instead of using
switch_to_thread.  Also, "inferior_list->thread_list->ptid" assumes
that we want the first thread of the first inferior, but that inferior
may not have threads, or with multi-target, that target may be
connected to some other target.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* remote.c (remote_target::start_remote): Don't set inferior_ptid
directly.  Instead find the first thread in the thread list and
use switch_to_thread.

4 years agoFix reconnecting to a gdbserver already debugging multiple processes, I
Pedro Alves [Fri, 10 Jan 2020 20:05:53 +0000 (20:05 +0000)]
Fix reconnecting to a gdbserver already debugging multiple processes, I

The multi-target patch will change the remote target's behavior when:

- the current inferior is connected to an extended-remote target.
- the current inferior is attached to any process.
- some other inferior than than the current one is live.

In current master, we get:

 (gdb) tar extended-remote :9999
 A program is being debugged already.  Kill it? (y or n)

While after multi-target, since each inferior may have its own target
connection, we'll get:

 (gdb) tar extended-remote :9999
 Already connected to a remote target.  Disconnect? (y or n)

That change made gdb.server/extended-remote-restart.exp expose a gdb
bug, because it made "target remote", via gdb_reconnect, just
disconnect from the previous connection, while in current master that
command would kill the inferior before disconnecting.  In turn, that
would make a multi-target gdb find processes already running under
control of gdbserver as soon as it reconnects, while in current master
there is never any process around when gdb reconnects, since they'd
all been killed prior to disconnection.

The bug this exposed is that remote_target::remote_add_inferior was
always reusing current_inferior() for the new process, even if the
current inferior was already bound to a process.  In the testcase's
case, when we reconnect, the remote is debugging two processes.  So
we'd bind the first remote process to the empty current inferior the
first time, and then bind the second remote process to the same
inferior again, essencially losing track of the first process.  That
resulted in failed assertions when we look up the inferior for the
first process by PID.  The fix is to still prefer binding to the
current inferior (so that plain "target remote" keeps doing what you'd
expect), but not reuse the current inferior if it is already bound to
a process.

This patch tweaks the test to explicitly disconnect before
reconnecting, to avoid GDB killing processes, thus making current GDB
behave the same as it will behave when the multi-target work lands.
That change alone without the GDB fix exposes the bug like so:

 (gdb) PASS: gdb.server/extended-remote-restart.exp: kill: 0, follow-child 0: disconnect
 target extended-remote localhost:2350
 Remote debugging using localhost:2350
 src/gdb/thread.c:93: internal-error: thread_info* inferior_thread(): Assertion `tp' failed.
 A problem internal to GDB has been detected,
 further debugging may prove unreliable.
 Quit this debugging session? (y or n)

The original bug that the testcase was written for was related to
killing, (git 9d4a934ce604 ("gdb: Fix assert for extended-remote
target (PR gdb/18050)")), but since the testcase tries reconnecting
with both explicitly killing and not explicitly killing, I think we're
covering the original bug with this testcase change.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* remote.c (remote_target::remote_add_inferior): Don't bind a
process to the current inferior if the current inferior is already
bound to a process.

gdb/testsuite/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.server/extended-remote-restart.exp (test_reload): Explicitly
disconnect before reconnecting.

4 years agoAvoid another inferior_ptid reference in gdb/remote.c
Tankut Baris Aktemur [Fri, 10 Jan 2020 20:05:52 +0000 (20:05 +0000)]
Avoid another inferior_ptid reference in gdb/remote.c

The multi-target patch makes inferior_ptid point to null_ptid before
calling into target_wait, which catches bad uses of inferior_ptid,
since the current selected thread in gdb shouldn't have much relation
to the thread that reports an event.

One such bad use is found in remote_target::remote_parse_stop_reply,
where we handle the 'W' or 'X' packets (process exit), and the remote
target does not support the multi-process extensions, i.e., it does
not report the PID of the process that exited.

With the multi-target patch, that would result in a failed assertion,
trying to find the inferior for process pid 0.

gdb/ChangeLog:
2020-01-10  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>
    Pedro Alves  <palves@redhat.com>

* remote.c (remote_target::remote_parse_stop_reply) <W/X packets>:
If no process is specified, return null_ptid instead of
inferior_ptid.
(remote_target::wait_as): Handle TARGET_WAITKIND_EXITED /
TARGET_WAITKIND_SIGNALLED with no pid.

gdb/testsuite/ChangeLog:
2020-01-10  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>
    Pedro Alves  <palves@redhat.com>

* gdb.server/connect-without-multi-process.exp: Also test
continuing to end.

4 years agoTweak handling of remote errors in response to resumption packet
Pedro Alves [Fri, 10 Jan 2020 20:05:52 +0000 (20:05 +0000)]
Tweak handling of remote errors in response to resumption packet

With current master, on a Fedora 27 machine with a kernel with buggy
watchpoint support, I see:

  (gdb) PASS: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: hardware breakpoints work
  continue
  Continuing.
  warning: Remote failure reply: E01
  Remote communication error.  Target disconnected.: Connection reset by peer.
  (gdb) FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: watchpoints work
  continue
  The program is not being run.
  (gdb) FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: breakpoint after the first fork (the program is no longer running)

The FAILs themselves aren't what's interesting here.  What is
interesting is that with the main multi-target patch applied, I was getting this:

  (gdb) PASS: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: hardware breakpoints work
  continue
  Continuing.
  warning: Remote failure reply: E01
  /home/pedro/brno/pedro/gdb/binutils-gdb-2/build/../src/gdb/inferior.c:285: internal-error: inferior* find_inferior_pid(process_stratum_target*, int): Assertion `pid != 0' failed.
  A problem internal to GDB has been detected,
  further debugging may prove unreliable.
  Quit this debugging session? (y or n) FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: watchpoints work (GDB internal error)

The problem is that in remote_target::wait_as, we're hitting this:

  switch (buf[0])
    {
    case 'E': /* Error of some sort. */
      /* We're out of sync with the target now.  Did it continue or
 not?  Not is more likely, so report a stop.  */
      rs->waiting_for_stop_reply = 0;

      warning (_("Remote failure reply: %s"), buf);
      status->kind = TARGET_WAITKIND_STOPPED;
      status->value.sig = GDB_SIGNAL_0;
      break;

which leaves event_ptid as null_ptid.  At the end of the function, we then reach:

  else if (status->kind != TARGET_WAITKIND_EXITED
   && status->kind != TARGET_WAITKIND_SIGNALLED)
    {
      if (event_ptid != null_ptid)
record_currthread (rs, event_ptid);
      else
event_ptid = inferior_ptid;                 <<<<< here
    }

and the trouble is that with the multi-target patch, we'll get here
with inferior_ptid as null_ptid too.  That is done exactly to find
these implicit assumptions that inferior_ptid is a good choice for
default thread, which isn't generaly true.

I first thought of fixing this in the "case 'E'" path, but, given that
this "event_ptid = inferior_ptid" path is also taken when the remote
target does not support threads at all, no thread-related packets or
extensions, it's better to fix it in latter path, to handle all
scenarios that miss reporting a thread.

That's what this patch does.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* remote.c (first_remote_resumed_thread): New.
(remote_target::wait_as): Use it as default event_ptid instead of
inferior_ptid.

4 years agoUse all_non_exited_inferiors in infrun.c
Pedro Alves [Fri, 10 Jan 2020 20:05:51 +0000 (20:05 +0000)]
Use all_non_exited_inferiors in infrun.c

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* infrun.c (handle_no_resumed): Use all_non_exited_inferiors.

4 years agotfile_target::close: trace_fd can't be -1
Pedro Alves [Fri, 10 Jan 2020 20:05:50 +0000 (20:05 +0000)]
tfile_target::close: trace_fd can't be -1

It's not possible to open a tfile target with an invalid trace_fd, and
it's not possible to close a closed target, so this early return is dead.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* tracefile-tfile.c (tfile_target::close): Assert that trace_fd is
not -1.

4 years agoSome get_last_target_status tweaks
Pedro Alves [Fri, 10 Jan 2020 20:05:49 +0000 (20:05 +0000)]
Some get_last_target_status tweaks

- Make get_last_target_status arguments optional.  A following patch
  will add another argument to get_last_target_status (the event's
  target), and passing nullptr when we don't care for some piece of
  info is handier than creating dummy local variables.

- Declare nullify_last_target_wait_ptid in a header, and remove the
  local extern declaration from linux-fork.c.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* break-catch-sig.c (signal_catchpoint_print_it): Don't pass a
ptid to get_last_target_status.
* break-catch-syscall.c (print_it_catch_syscall): Don't pass a
ptid to get_last_target_status.
* infcmd.c (continue_command): Don't pass a target_waitstatus to
get_last_target_status.
(info_program_command): Don't pass a target_waitstatus to
get_last_target_status.
* infrun.c (init_wait_for_inferior): Use
nullify_last_target_wait_ptid.
(get_last_target_status): Handle nullptr arguments.
(nullify_last_target_wait_ptid): Clear target_last_waitstatus.
(print_stop_event): Don't pass a ptid to get_last_target_status.
(normal_stop): Don't pass a ptid to get_last_target_status.
* infrun.h (get_last_target_status, set_last_target_status): Move
comments here and update.
(nullify_last_target_wait_ptid): Declare.
* linux-fork.c (fork_load_infrun_state): Remove local extern
declaration of nullify_last_target_wait_ptid.
* linux-nat.c (get_detach_signal): Don't pass a target_waitstatus
to get_last_target_status.

4 years agoswitch inferior/thread before calling target methods
Pedro Alves [Fri, 10 Jan 2020 20:05:48 +0000 (20:05 +0000)]
switch inferior/thread before calling target methods

Once each inferior has its own target stack, we'll need to make sure
that the right inferior is selected before we call into target
methods.

It kind of sounds worse than it is in practice.  Not that many places
need to be concerned.

In thread.c, we add a new switch_to_thread_if_alive function that
centralizes the switching before calls to target_thread_alive.  Other
cases are handled with explicit switching.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdbthread.h (scoped_restore_current_thread)
<dont_restore, restore, m_dont_restore>: Declare.
* thread.c (thread_alive): Add assertion.  Return bool.
(switch_to_thread_if_alive): New.
(prune_threads): Switch inferior/thread.
(print_thread_info_1): Switch thread before calling target methods.
(scoped_restore_current_thread::restore): New, factored out from
...
(scoped_restore_current_thread::~scoped_restore_current_thread):
... this.
(scoped_restore_current_thread::scoped_restore_current_thread):
Add assertion.
(thread_apply_all_command, thread_select): Use
switch_to_thread_if_alive.
* infrun.c (proceed, restart_threads, handle_signal_stop)
(switch_back_to_stepped_thread): Switch current thread before
calling target methods.

4 years agoIntroduce switch_to_inferior_no_thread
Pedro Alves [Fri, 10 Jan 2020 20:05:47 +0000 (20:05 +0000)]
Introduce switch_to_inferior_no_thread

Several places want to switch context to an inferior and its pspace,
while at the same time switch to "no thread selected".  This commit
adds a function that does that, and uses it in a few places.

gdb/ChangeLog:
2020-01-10  Pedro Alves <palves@redhat.com>

* inferior.c (switch_to_inferior_no_thread): New function,
factored out from ...
(inferior_command): ... here.
* inferior.h (switch_to_inferior_no_thread): Declare.
* mi/mi-main.c (run_one_inferior): Use
switch_to_inferior_no_thread.

4 years agoDelete unnecessary code from kill_command
Pedro Alves [Fri, 10 Jan 2020 20:05:46 +0000 (20:05 +0000)]
Delete unnecessary code from kill_command

I believe this comment:

      /* Killing off the inferior can leave us with a core file.  If
 so, print the state we are left in.  */

Referred to the fact that a decade ago, by design, GDB would let you
type "run" when debugging a core dump, keeping the core open.  That
"run" would push a process_stratum target on the target stack for the
live process, and, the core would remain open -- we used to have a
core_stratum.  When the live process was killed/detached or exited,
GDB would go back to debugging the core, since the core_stratum target
was now at the top of the stack.  That design had a number of
problems, see here for example:

  https://sourceware.org/ml/gdb-patches/2008-08/msg00290.html

In 2010, core_stratum was finaly eliminated and cores now have
process_stratum too, with commit c0edd9edadfe ("Make core files the
process_stratum.").  Pushing a live process on the stack while you're
debugging a core discards the core completely.

I also thought that this might be in use with checkpoints, but it does
not -- "kill" when you have multiple checkpoints kills all the
checkpoints.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* infcmd.c (kill_command): Remove dead code.

4 years agoDon't check target is running in remote_target::mourn_inferior
Pedro Alves [Fri, 10 Jan 2020 20:05:45 +0000 (20:05 +0000)]
Don't check target is running in remote_target::mourn_inferior

I believe the tail end of remote_target::mourn_inferior is broken, and
it's been broken for too long to even bother trying to fix.  Most
probably nobody needs it.  If the code is reached and we find the
target is running, we'd need to resync the thread list, at least,
since generic_mourn_inferior got rid of all the threads in the
inferior, otherwise, we'd hit an assertion on the next call to
inferior_thread(), for example.  A "correct" fix would probably
involve restarting the whole remote_target::start_remote requence,
exactly as if we had completely disconnected and reconnected from
scratch.

Note that regular stub debugging usually uses plain target remote, but
this code is only reachable in target extended-mode:

- The !remote_multi_process_p check means that it's only reacheable if
  the stub does not support multi-process.  I.e., there can only ever
  be one live process.

- remote_target::mourn_inferior has this at the top:

  /* In 'target remote' mode with one inferior, we close the connection.  */
  if (!rs->extended && number_of_live_inferiors () <= 1)
    {
      unpush_target (this);

      /* remote_close takes care of doing most of the clean up.  */
      generic_mourn_inferior ();
      return;
    }

  Which means that if we only had one live inferior (which for our
  case, must be true), we'll have closed the connection already,
  unless we're in extended-remote mode.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* remote.c (remote_target::mourn_inferior): No longer check
whether the target is running.

4 years agoMake target_ops::has_execution take an 'inferior *' instead of a ptid_t
Pedro Alves [Fri, 10 Jan 2020 20:05:44 +0000 (20:05 +0000)]
Make target_ops::has_execution take an 'inferior *' instead of a ptid_t

With the multi-target work, each inferior will have its own target
stack, so to call a target method, we'll need to make sure that the
inferior in question is the current one, otherwise target->beneath()
calls will find the target beneath in the wrong inferior.

In some places, it's much more convenient to be able to check whether
an inferior has execution without having to switch to it in order to
call target_has_execution on the right inferior/target stack, to avoid
side effects with switching inferior/thread/program space.

The current target_ops::has_execution method takes a ptid_t as
parameter, which, in a multi-target world, isn't sufficient to
identify the target.  This patch prepares to address that, by changing
the parameter to an inferior pointer instead.  From the inferior,
we'll be able to query its target stack to tell which target is
beneath.

Also adds a new inferior::has_execution() method to make callers a bit
more natural to read.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* corelow.c (core_target::has_execution): Change parameter type to
inferior pointer.
* inferior.c (number_of_live_inferiors): Use
inferior::has_execution instead of target_has_execution_1.
* inferior.h (inferior::has_execution): New.
* linux-thread-db.c (thread_db_target::update_thread_list): Use
inferior::has_execution instead of target_has_execution_1.
* process-stratum-target.c
(process_stratum_target::has_execution): Change parameter type to
inferior pointer.  Check the inferior's PID instead of
inferior_ptid.
* process-stratum-target.h
(process_stratum_target::has_execution): Change parameter type to
inferior pointer.
* record-full.c (record_full_core_target::has_execution): Change
parameter type to inferior pointer.
* target.c (target_has_execution_1): Change parameter type to
inferior pointer.
(target_has_execution_current): Adjust.
* target.h (target_ops::has_execution): Change parameter type to
inferior pointer.
(target_has_execution_1): Change parameter type to inferior
pointer.  Change return type to bool.
* tracefile.h (tracefile_target::has_execution): Change parameter
type to inferior pointer.

4 years agoexceptions.c:print_flush: Remove obsolete check
Pedro Alves [Fri, 10 Jan 2020 20:05:43 +0000 (20:05 +0000)]
exceptions.c:print_flush: Remove obsolete check

Commit 20f0d60db4fb ("Avoid crash when calling warning too early"),
added a "current_top_target () != NULL" check to
target_supports_terminal_ours, so this check in exceptions.c is now
obsolete.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* exceptions.c (print_flush): Remove current_top_target() check.

4 years agoMake "show remote exec-file" inferior-aware
Pedro Alves [Fri, 10 Jan 2020 20:05:42 +0000 (20:05 +0000)]
Make "show remote exec-file" inferior-aware

The "set remote exec-file" setting is per-inferior, but the "show
remote exec-file" command always shows the last set exec-file,
irrespective of the current inferior.  E.g.:

 # Set inferior 1's exec-file:
 (gdb) set remote exec-file prog1

 # Add inferior 2, switch to it, and set its exec-file:
 (gdb) add-inferior
 Added inferior 2
 (gdb) inferior 2
 (gdb) set remote exec-file prog2

 # Switch back to inferior 1, and show its exec-file:
 (gdb) inferior 1
 (gdb) show remote exec-file
 prog2
 ^^^^^ should show "prog1" instead here.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* remote.c (show_remote_exec_file): Show the current inferior's
exec-file instead of the command variable's value.

gdb/testsuite/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.base/remote-exec-file.exp: New file.

4 years agoDon't rely on inferior_ptid in record_full_wait
Pedro Alves [Fri, 10 Jan 2020 20:05:41 +0000 (20:05 +0000)]
Don't rely on inferior_ptid in record_full_wait

The multi-target patch sets inferior_ptid to null_ptid before handling
a target event, and thus before calling target_wait, in order to catch
places in target_ops::wait implementations that are incorrectly
relying on inferior_ptid (which could otherwise be a ptid of a
different target, for example).  That caught this instance in
record-full.c.

Fix it by saving the last resumed ptid, and then using it in
record_full_wait_1, just like how the last "step" argument passed to
record_full_target::resume is handled too.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* record-full.c (record_full_resume_ptid): New global.
(record_full_target::resume): Set it.
(record_full_wait_1): Use record_full_resume_ptid instead of
inferior_ptid.

4 years agoPreserve selected thread in all-stop w/ background execution
Pedro Alves [Fri, 10 Jan 2020 20:05:41 +0000 (20:05 +0000)]
Preserve selected thread in all-stop w/ background execution

In non-stop mode, if you resume the program in the background (with
"continue&", for example), then gdb makes sure to not switch the
current thread behind your back.  That means that you can be sure that
the commands you type apply to the thread you selected, even if some
other thread that was running in the background hits some event just
while you're typing.

In all-stop mode, however, if you resume the program in the
background, gdb let's the current thread switch behind your back.

This is bogus, of course.  All-stop and non-stop background
resumptions should behave the same.

This patch fixes that, and adds a testcase that exposes the bad
behavior in current master.

The fork-running-state.exp changes are necessary because that
preexisting testcase was expecting the old behavior:

Before:

  continue &
  Continuing.
  (gdb)
  [Attaching after process 8199 fork to child process 8203]
  [New inferior 2 (process 8203)]
  info threads
    Id   Target Id                      Frame
    1.1  process 8199 "fork-running-st" (running)
  * 2.1  process 8203 "fork-running-st" (running)
  (gdb)

After:

  continue &
  Continuing.
  (gdb)
  [Attaching after process 24660 fork to child process 24664]
  [New inferior 2 (process 24664)]
  info threads
    Id   Target Id                       Frame
  * 1.1  process 24660 "fork-running-st" (running)
    2.1  process 24664 "fork-running-st" (running)
  (gdb)

Here we see that before this patch GDB switches current inferior to
the new inferior behind the user's back, as a side effect of handling
the fork.

The delete_exited_threads call in inferior_appeared is there to fix an
issue that Baris found in a previous version of this patch.  The
fetch_inferior_event change increases the refcount of the current
thread, and in case the fetched inferior event denotes a thread exit,
the thread will not be deleted right away.  A non-deleted but exited
thread stays in the inferior's thread list.  This, in turn, causes the
"init_thread_list" call in inferior.c to be skipped.  A consequence is
that the global thread ID counter is not restarted if the current
thread exits, and then the inferior is restarted:

 (gdb) start
 Temporary breakpoint 1 at 0x4004d6: file main.c, line 21.
 Starting program: /tmp/main

 Temporary breakpoint 1, main () at main.c:21
 21        foo ();
 (gdb) info threads -gid
   Id   GId  Target Id            Frame
 * 1    1    process 16106 "main" main () at main.c:21
 (gdb) c
 Continuing.
 [Inferior 1 (process 16106) exited normally]
 (gdb) start
 Temporary breakpoint 2 at 0x4004d6: file main.c, line 21.
 Starting program: /tmp/main

 Temporary breakpoint 2, main () at main.c:21
 21        foo ();
 (gdb) info threads -gid
   Id   GId  Target Id            Frame
 * 1    2    process 16138 "main" main () at main.c:21
       ^^^

Notice that GId == 2 above.  It should have been "1" instead.

The new tids-git-reset.exp testcase exercises the problem above.

gdb/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdbthread.h (scoped_restore_current_thread)
<dont_restore, restore, m_dont_restore>: Declare.
* thread.c (thread_alive): Add assertion.  Return bool.
(switch_to_thread_if_alive): New.
(prune_threads): Switch inferior/thread.
(print_thread_info_1): Switch thread before calling target methods.
(scoped_restore_current_thread::restore): New, factored out from
...
(scoped_restore_current_thread::~scoped_restore_current_thread):
... this.
(scoped_restore_current_thread::scoped_restore_current_thread):
Add assertion.
(thread_apply_all_command, thread_select): Use
switch_to_thread_if_alive.

gdb/testsuite/ChangeLog:
2020-01-10  Pedro Alves  <palves@redhat.com>

* gdb.base/fork-running-state.exp (do_test): Adjust expected
output.
* gdb.threads/async.c: New.
* gdb.threads/async.exp: New.
* gdb.multi/tids-gid-reset.c: New.
* gdb.multi/tids-gid-reset.exp: New.

4 years agoFix handling of null stap semaphores
George Barrett [Fri, 10 Jan 2020 19:30:28 +0000 (06:30 +1100)]
Fix handling of null stap semaphores

According to the SystemTap documentation on user-space probes[0], stap
probe points without semaphores are denoted by setting the semaphore
address in the probe's note to zero. At present the code does do a
comparison of the semaphore address against zero, but only after it's
been relocated; as such it will (almost?) always fail, commonly
resulting in GDB trying to overwrite the ELF magic located at the
image's base address.

This commit tests the address as specified in the SDT note rather than
the relocated value in order to correctly detect absent probe
semaphores.

[0]: https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation

gdb/Changelog:
2020-01-11  George Barrett  <bob@bob131.so>

* stap-probe.c (stap_modify_semaphore): Don't check for null
semaphores.
(stap_probe::set_semaphore, stap_probe::clear_semaphore): Check
for null semaphores.

gdb/testsuite/ChangeLog:
2020-01-11  George Barrett  <bob@bob131.so>

* gdb.base/stap-probe.c (relocation_marker): Add dummy variable
to help in finding the image relocation offset.
* gdb.base/stap-probe.exp (stap_test): Accept arbitrary compile
options in arguments.
(stap_test_no_debuginfo): Likewise.
(stap-probe-nosem-noopt-pie, stap-probe-nosem-noopt-nopie): Add
test variants.
(stap_test): Add null semaphore relocation test.

4 years agogdb/testsuite/gdb.base/stap-probe: Minor clean-up
George Barrett [Fri, 10 Jan 2020 19:30:01 +0000 (06:30 +1100)]
gdb/testsuite/gdb.base/stap-probe: Minor clean-up

This patch resolves a couple of issues with the test case for SystemTap
user-space probe points:
  1. The preprocessor macro guarding the semaphore variables in the C
     file is (rather confusingly) named USE_PROBES. This has been
     renamed to USE_SEMAPHORES, to better reflect its function.
  2. The test procedures in the expect file improperly pass the flag
     defining USE_PROBES to prepare_for_testing; as such, the test
     binary that's supposed to have probes with semaphores is the same
     as the one without. This has also been fixed.
  3. No test is performed to check that `info probes' returns
     information about probe semaphores. Such a test is included in this
     patch.

gdb/testsuite/ChangeLog
2020-01-10  George Barrett  <bob@bob131.so>

* gdb.base/stap-probe.c: Rename USE_PROBES to USE_SEMAPHORES.
* gdb.base/stap-probe.exp: Likewise.
(stap_test): Pass argument as an additional flag.
(stap_test_no_debuginfo): Likewise.
(stap_test): Check `info probes stap' output for semaphore
addresses if the test binary is supposed to have them.

4 years ago[PR ld/22269] arm: Avoid dynamic relocs for undefweak symbols in static PIE
Szabolcs Nagy [Thu, 9 Jan 2020 17:20:56 +0000 (17:20 +0000)]
[PR ld/22269] arm: Avoid dynamic relocs for undefweak symbols in static PIE

With static PIE linking undefined weak symbols are resolved to 0, so no
dynamic relocation is needed for them. The UNDEFWEAK_NO_DYNAMIC_RELOC
macro was introduced so this case can be handled easily, but it was not
applied consistently in the first attempt to fix ld/22269 for arm:

  commit 95b03e4ad68e7a90f5096b47df595636344b783a
  arm: Check UNDEFWEAK_NO_DYNAMIC_RELOC

This patch fixes spurious relative relocs in static PIE binaries against
GOT entries created for undefined weak symbols on arm*-*, this fixes

FAIL: pr22269-1 (static pie undefined weak)

bfd/ChangeLog:

PR ld/22269
* elf32-arm.c (elf32_arm_final_link_relocate): Use
UNDEFWEAK_NO_DYNAMIC_RELOC.
(allocate_dynrelocs_for_symbol): Likewise.

4 years agoAArch64: Revert setting of elf class in linker stub.
Tamar Christina [Fri, 10 Jan 2020 13:48:57 +0000 (13:48 +0000)]
AArch64: Revert setting of elf class in linker stub.

This changes the fix to PR 25210 by removing the ELF class change.
As it turns out the correct change was only the change in compress.c.

Everything else is unneeded and setting the elf class is making the linker
behave very oddly under LTO.  The first stub is correctly written out but for
the rest the suddenly don't have a pointer to the stub section anymore.

This caused SPEC to fail as the program would branch to the stub and it wouldn't
be filled in.

Committed to master under the trivial rule as this is partially reverting a previous commit.

bfd/ChangeLog:

PR 25210
* elfnn-aarch64.c (_bfd_aarch64_create_stub_section): Remove elfclass.

4 years agoHPUX gas testsuite fixes
Alan Modra [Fri, 10 Jan 2020 06:59:59 +0000 (17:29 +1030)]
HPUX gas testsuite fixes

* testsuite/gas/elf/pr14891.s: Don't start directives in first column.
* testsuite/gas/elf/pr21661.d: Don't run on hpux.

4 years agoubsan: tilepro: signed integer overflow
Alan Modra [Thu, 9 Jan 2020 21:57:33 +0000 (08:27 +1030)]
ubsan: tilepro: signed integer overflow

* tilepro-opc.c (parse_insn_tilepro): Make opval unsigned.
* tilegx-opc.c (parse_insn_tilegx): Likewise.  Delete raw_opval.

4 years agoubsan: m10300: shift exponent -4
Alan Modra [Wed, 8 Jan 2020 20:29:42 +0000 (06:59 +1030)]
ubsan: m10300: shift exponent -4

* m10300-dis.c (disassemble): Move extraction of DREG, AREG, RREG,
and XRREG value earlier to avoid a shift with negative exponent.
* m10200-dis.c (disassemble): Similarly.

4 years agoubsan: spu: left shift of negative value
Alan Modra [Wed, 8 Jan 2020 20:14:16 +0000 (06:44 +1030)]
ubsan: spu: left shift of negative value

Also fixes a real bug.  The DECODE_INSN_I9a and DECODE_INSN_I9b both
use UNSIGNED_EXTRACT for 7 low bits of the result, but this was an
unsigned value due to "insn" being unsigned.  DECODE_INSN_I9* is
therefore unsigned too, leading to a zero extension in an expression
using a bfd_vma if bfd_vma is 64 bits.

* opcode/spu.h: Formatting.
(UNSIGNED_EXTRACT): Use 1u.
(SIGNED_EXTRACT): Don't sign extend with shifts.
(DECODE_INSN_I9a, DECODE_INSN_I9b): Avoid left shift of signed value.
Keep result signed.
(DECODE_INSN_U9a, DECODE_INSN_U9b): Delete.

4 years agoubsan: alpha-coff: signed integer overflow
Alan Modra [Wed, 8 Jan 2020 20:11:25 +0000 (06:41 +1030)]
ubsan: alpha-coff: signed integer overflow

* coff-alpha.c (alpha_ecoff_object_p): Calculate size in bfd_size_type.

4 years agoAutomatic date update in version.in
GDB Administrator [Fri, 10 Jan 2020 00:00:24 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agogdb/tui: Link source and assembler scrolling .... again
Andrew Burgess [Tue, 7 Jan 2020 11:39:17 +0000 (11:39 +0000)]
gdb/tui: Link source and assembler scrolling .... again

Until recently when the source window was scrolled the assembler
window would scroll in sync - keeping the disassembly for the current
line in view.

This was broken in commit:

  commit b4b49dcbff6b437fa8b4e2fc0c3f27b457f11310
  Date:   Wed Nov 13 16:47:58 2019 -0700

      Don't call tui_show_source from tui_ui_out

This commit restores the synchronised scrolling and also maintains the
horizontal scroll within the source view when it is vertically
scrolled, something that was broken before.

This commit does not mean that scrolling the assembler view scrolls
the source view.  The connection this way never existed, though maybe
it should, but I'll leave adding this feature for a separate commit.

gdb/ChangeLog:

* tui/tui-source.c (tui_source_window::do_scroll_vertical): Update
all source windows, and maintain horizontal scroll status while
doing so.

gdb/testsuite/ChangeLog:

* gdb.tui/basic.exp: Add more scrolling tests.

Change-Id: I250114a3bc670040a6a759d41905776771b2f818

4 years agogdb: Fix scrolling in TUI
Tom Tromey [Sun, 22 Dec 2019 23:52:56 +0000 (16:52 -0700)]
gdb: Fix scrolling in TUI

Hannes Domani pointed out that my previous patch to fix the "list"
command in the TUI instead broke vertical scrolling.  While looking at
this, I found that do_scroll_vertical calls print_source_lines, which
seems like a very roundabout way to change the source window.  This
patch removes this oddity and fixes the bug at the same time.

I've added a new test case.  This is somewhat tricky, because the
obvious approach of sending a dummy command after the scroll did not
work -- due to how the TUI works, sennding a command causes the scroll
to take effect.

gdb/ChangeLog
2019-12-22  Tom Tromey  <tom@tromey.com>

PR tui/18932:
* tui/tui-source.c (tui_source_window::do_scroll_vertical): Call
update_source_window, not print_source_lines.

gdb/testsuite/ChangeLog
2019-12-22  Tom Tromey  <tom@tromey.com>

PR tui/18932:
* lib/tuiterm.exp (Term::wait_for): Rename from _accept.  Return a
meangingful value.
(Term::command, Term::resize): Update.
* gdb.tui/basic.exp: Add scrolling test.

Change-Id: I9636a7c8a8cade37431c6165ee996a9d556ef1c8

4 years agogdb/tui: Fix 'layout asm' before the inferior has started
Andrew Burgess [Tue, 7 Jan 2020 00:41:08 +0000 (00:41 +0000)]
gdb/tui: Fix 'layout asm' before the inferior has started

Currently if a user starts the tui with 'layout asm' then they will be
presented with the 'src' layout.

What happens is:

  1. Layout command enables TUI, selecting the SRC layout by default.

  2. As part of tui_enable we call tui_display_main, which calls
     tui_get_begin_asm_address, which calls
     set_default_source_symtab_and_line.  This changes core GDBs
     current symtab and line, which triggers a call to the symtab
     changed hook tui_symtab_changed, which sets the flag
     from_source_symtab.

  3. Back in the layout command, the layout is changed from SRC to
     ASM.  After this the layout command completes and we return to
     core GDB which prints the prompt, however...

  4. The before prompt hook is called which sees the
     from_source_symtab flag is set and forces the SRC window to be
     displayed.  This switches us back to SRC view.

The solution I propose here is to delay installing the hooks into core
GDB until after we have finished setting up the tui and selecting the
default frame to view.  In this way we effectively ignore the first
symtab changed event triggered when making main the default symtab.

gdb/ChangeLog:

* tui/tui.c (tui_enable): Register tui hooks after calling
tui_display_main.

gdb/testsuite/ChangeLog:

* gdb.tui/tui-layout-asm.exp: New file.

Change-Id: I858ab81a17ffb4aa72deb3f36c3755228a9c9d9a

4 years agogdb/testsuite/tui: Introduce check_box_contents
Andrew Burgess [Tue, 7 Jan 2020 00:35:02 +0000 (00:35 +0000)]
gdb/testsuite/tui: Introduce check_box_contents

A new test procedure for matching the contents of one screen box
against a regexp.  This can be used to match the contents of one TUI
window against a regexp without any of the borders, or other windows
being included in the matched output (as is currently the case with
check_contents).

This will be used in a later commit.

gdb/testsuite/ChangeLog:

* lib/tuiterm.exp (Term::check_box_contents): New proc.

Change-Id: Icf795bf38dd9295e282a34eecc318a9cdbc73926

4 years agogdb/testsuite/tui: Split enter_tui into two procs
Andrew Burgess [Tue, 7 Jan 2020 00:30:35 +0000 (00:30 +0000)]
gdb/testsuite/tui: Split enter_tui into two procs

Split Term::enter_tui into two procedures, a core which does the
setup, but doesn't actually enable tui mode, and the old enter_tui
that calls the new core, and then enables tui mode.

This is going to be useful in a later commit.

gdb/testsuite/ChangeLog:

* lib/tuiterm.exp (Term::prepare_for_tui): New proc.
(Term::enter_tui): Use Term::prepare_for_tui.

Change-Id: I501dfb2ddaa4a4e7246a5ad319ab428e4f42b3af

4 years agogdb/testsuite/tui: Always dump_screen when asked
Andrew Burgess [Tue, 7 Jan 2020 00:26:22 +0000 (00:26 +0000)]
gdb/testsuite/tui: Always dump_screen when asked

The Term::dump_screen routine currently dumps the screen using calls
to 'verbose', this means it will only dump the screen when the
testsuite is running in verbose mode.

However, the Term::dump_screen is most often called when a test fails,
in this case I think it is useful to have the screen dumped even when
we're not in verbose mode.

This commit changes the calls to 'verbose' to be 'verbose -log' so we
always get the screen dump.

gdb/testsuite/ChangeLog:

* lib/tuiterm.exp (Term::dump_screen): Always dump the screen when
called.

Change-Id: I5f0a7f5ac2ece04d6fe6e9c5a28ea2a0dda38955

4 years agogdb/testsuite: Fix race condition in gdb.base/skip.exp
Andrew Burgess [Thu, 9 Jan 2020 22:38:22 +0000 (22:38 +0000)]
gdb/testsuite: Fix race condition in gdb.base/skip.exp

In this commit:

  commit 5024637fac653914d471808288dc3221bc7ec089
  Date:   Sun Dec 15 11:05:47 2019 +0100

      Fix skip.exp test failure observed with gcc-9.2.0

A race condition was introduced into the gdb.base/skip.exp test when
this line:

    gdb_test "step" "foo \\(\\) at.*" "step 3"

Was changed to this:

    gdb_test "step" "foo \\(\\) at.*" "step 3" "main \\(\\) at .*" "step"

Before the above change we expected GDB to behave like this:

  (gdb) step
  foo () at /path/to/gdb/testsuite/gdb.base/skip.c:42
  42        return 0;
  (gdb)

However, when the test is compiled with GCC 9.2.0 we get a different
behaviour, and so we need a second 'step', like this:

  (gdb) step
  main () at /path/to/gdb.base/skip.c:32
  32        x = baz ((bar (), foo ()));
  (gdb) step
  foo () at /path/to/gdb/testsuite/gdb.base/skip.c:42
  42        return 0;
  (gdb)

Now the change to the test matches against 'main () at .*', however if
GDB or expect is being slow then we might only get to see output like
this:

  (gdb) step
  main () at /path/to/g

This will happily match the question pattern, so we send 'step' to GDB
again.  Now GDB continues to produce output which expect accepts, we
now see this:

  b.base/skip.c:32
  32        x = baz ((bar (), foo ()));
  (gdb)

This has carried on from where the previous block of output left off.
This doesn't match the final pattern 'foo \\(\\) at.*', but it does
match the prompt pattern that gdb_test_multiple adds, and so we report
the test as failing.

The solution is to simply ensure that the question consumes everything
up to, and including the prompt.  This ensures that the prompt can't
then match the failure case.  The new test line becomes:

    gdb_test "step" "foo \\(\\) at.*" "step 3" \
       "main \\(\\) at .*\r\n$gdb_prompt " "step"

gdb/testsuite/ChangeLog:

* gdb.base/skip.exp: Fix race condition in test.

Change-Id: I9f0b0b52ef1b4f980bfaa8fe405ff06d520f3482

4 years agoDon't define _FORTIFY_SOURCE on MinGW
Christian Biesinger [Wed, 18 Dec 2019 18:06:43 +0000 (12:06 -0600)]
Don't define _FORTIFY_SOURCE on MinGW

Recent MinGW versions require -lssp when using _FORTIFY_SOURCE, which
gdb does (in common-defs.h)
https://github.com/msys2/MINGW-packages/issues/5868#issuecomment-544107564

To avoid all the complications with checking for -lssp and making sure it's
linked statically, just don't define it.

gdb/ChangeLog:

2020-01-09  Christian Biesinger  <cbiesinger@google.com>

* gdbsupport/common-defs.h: Don't define _FORTIFY_SOURCE on MinGW.

Change-Id: Ide6870ab57198219a2ef78bc675768a789ca2b1d

4 years agoFix indentation in print_thread_info_1
Simon Marchi [Wed, 8 Jan 2020 21:55:03 +0000 (16:55 -0500)]
Fix indentation in print_thread_info_1

The body of this this big "for" loop is missing an indentation level,
this patch fixes that.

gdb/ChangeLog:

* thread.c (print_thread_info_1): Fix indentation.

4 years agoFix memory leak of the demangled symbol name
Christian Biesinger [Wed, 8 Jan 2020 01:10:40 +0000 (19:10 -0600)]
Fix memory leak of the demangled symbol name

compute_and_set_names would only free the name if we did not find the name
in the hashtable, but it needs to always free it.  Solve this by moving the
smart pointer outside the if.

Thanks to PhilippeW for finding this.

gdb/ChangeLog:

2020-01-09  Christian Biesinger  <cbiesinger@google.com>

* symtab.c (general_symbol_info::compute_and_set_names): Move the
unique_xmalloc_ptr outside the if to always free the demangled name.

Change-Id: Id7c6b8408432183700ccb5ff634818d6c5a3ac95

4 years agoFix an attempt to free a static pointer when using objcopy's symbol addition feature.
Nick Clifton [Thu, 9 Jan 2020 16:51:04 +0000 (16:51 +0000)]
Fix an attempt to free a static pointer when using objcopy's symbol addition feature.

PR 25220
* objcopy.c (empty_name): New variable.
(need_sym_before): Prevent an attempt to free a static variable.
(filter_symbols): Avoid strcmp test by checking for pointer
equality.

4 years agoFix an illegal memory access triggered when trying to examine an input file containin...
Nick Clifton [Thu, 9 Jan 2020 15:49:08 +0000 (15:49 +0000)]
Fix an illegal memory access triggered when trying to examine an input file containing corrupt compressed sections.

PR 25221
* bfd.c (bfd_convert_section_contents): Check for a compress
header size that is larger than the actual section size.

4 years agoFix the cast used to prevent compile time warning about an always false test.
Nick Clifton [Thu, 9 Jan 2020 14:32:49 +0000 (14:32 +0000)]
Fix the cast used to prevent compile time warning about an always false test.

PR 25224
* z80-dis.c (ld_ii_ii): Use correct cast.

4 years agooops - toplevel changelog entry for previous delta.
Aaron Merey [Thu, 9 Jan 2020 13:37:26 +0000 (13:37 +0000)]
oops - toplevel changelog entry for previous delta.

       * config/debuginfod.m4: New file. Add macro AC_DEBUGINFOD. Adds
        new configure option --with-debuginfod.
        * configure: Regenerate.
        * configure.ac: Call AC_DEBUGINFOD.

4 years agoUpdate Traditional Chinese translation for the binutils sub-directory.
Nick Clifton [Thu, 9 Jan 2020 13:36:22 +0000 (13:36 +0000)]
Update Traditional Chinese translation for the binutils sub-directory.

* po/zh_TW.po: Updated Traditional Chinese translation.

4 years agoAdd support for debuginfod to the binutils (disable by default, enabled via a configu...
Aaron Merey [Thu, 9 Jan 2020 13:19:20 +0000 (13:19 +0000)]
Add support for debuginfod to the binutils (disable by default, enabled via a configure time option).

debuginfod is a lightweight web service that indexes ELF/DWARF
debugging resources by build-id and serves them over HTTP. This patch
enables objdump and readelf to query debuginfod servers when they are
otherwise not able to find separate debug files. Binutils can be built
with debuginfod using the --with-debuginfod configure option. This
requires that libdebuginfod be installed and found at configure time.
debuginfod is packaged with elfutils, starting with version 0.178. For
more information see https://sourceware.org/elfutils/.

toplevel* config/debuginfod.m4: New file. Add macro AC_DEBUGINFOD. Adds
        new configure option --with-debuginfod.
        * configure: Regenerate.
        * configure.ac: Call AC_DEBUGINFOD.

binutils* Makefile.am (readelf_LDADD, objdump_LDADD): Add libdebuginfod.
        * Makefile.in: Regenerate.
        * NEWS: Update.
        * config.in: Regenerate.
        * configure: Regenerate.
        * configure.ac: Call AC_DEBUGINFOD.
        * doc/Makefile.in: Regenerate.
        * doc/binutils.texi: Add section on using binutils
        with debuginfod.
        * dwarf.c (debuginfod_fetch_separate_debug_info): New function.
        Query debuginfod servers for the target debug file.
        (load_separate_debug_info): Call
        debuginfod_fetch_separate_debug_info if configured with
        debuginfod.
        (load_separate_debug_files): Add file argument to
        load_separate_debug_info calls.
        * dwarf.h (get_build_id): Add declaration.
        * objdump.c (get_build_id): New function. Get build-id of file.
        * readelf.c (get_build_id): Likewise.
        * testsuite/binutils-all/debuginfod.exp: New tests.
        * testsuite/binutils-all/linkdebug.s: Add .note.gnu.build-id
        section.

4 years agoFix compile time warnings about comparisons always being false.
Sergey Belyashov [Thu, 9 Jan 2020 11:47:44 +0000 (11:47 +0000)]
Fix compile time warnings about comparisons always being false.

PR 25224
gas * config/tc-z80.c (emit_ld_m_rr): Use integer types when checking
opcode byte values.
(emit_ld_r_r): Likewise.
(emit_ld_rr_m): Likewise.
(emit_ld_rr_nn): Likewise.

opcodes * z80-dis.c (ld_ii_ii): Use character constant when checking
opcode byte value.