Many make_cleanup uses in the code base are best eliminated by using a
"owning" smart pointer to manage ownership of the resource
automatically.
The question is _which_ smart pointer.
GDB currently supports building with a C++03 compiler. We have
std::auto_ptr in C++03, but, as is collective wisdom by now, that's
too easy to misuse, and has therefore been deprecated in C++11 and
finally removed in C++17.
It'd be nice to be able to use std::unique_ptr instead, which is the
modern, safe std::auto_ptr replacement in C++11.
In addition to extra safety -- moving (i.e., transfer of ownership of
the managed pointer between smart pointers) must be explicit --
std::unique_ptr has (among others) one nice feature that std::auto_ptr
doesn't --- ability to specify a custom deleter as template parameter.
In gdb's context, that allows easily creating a smart pointer for
memory allocated with xmalloc -- the smart pointer then knows to
release with xfree instead of delete. This is particularly
interesting when managing objects allocated in C libraries, and also,
for C++-fying parts of GDB that interact with other parts that still
return objects allocated with xmalloc.
Since std::unique_ptr's API is quite nice, and eventually we'd like to
move to C++11, this patch adds a C++03-compatible smart pointer that
exposes the subset of the std::unique_ptr API that we're interested
in. An advantage is that whenever we start requiring C++11, we won't
have to learn a new API. Meanwhile, this allows continuing to support
building with a C++03 compiler.
Since C++03 doesn't support rvalue references (boost gets close to
emulating them, but it's not fully transparent to user code), the
C++03 std::unique_ptr emulation here doesn't try hard to prevent
accidentally moving, which is where most of complication of a more
thorough emulation would be. Instead, we rely on the fact that GDB
will be usually compiled with a C++11 compiler, and use the real
std::unique_ptr in that case to catch such accidental moves. IOW, the
goal here is to allow code that would be correct using std::unique_ptr
to be equally correct in C++03 mode, and, just as efficient.
The C++03 version was originally based on GCC 7.0's std::auto_ptr and
then heavily customized to behave more like C++11's std::unique_ptr:
- Support for custom (stateless) deleters. (Support for stateful
deleters could be added, if necessary.)
- unique_ptr<T[]> partial specialization (auto_ptr<T> does not know
to use delete[]).
- Support for all of 'ptr != NULL', 'ptr == NULL' and 'if (ptr)'
using the safe bool idiom to emulate C++11's explicit bool
operator.
- There's no nullptr in C++03, so this allows initialization and
assignment from NULL instead (std::auto_ptr allows neither).
- Variable names un-uglified (ie., no leading __ prefix everywhere).
- Formatting made to follow GDB's coding conventions, including
comment style.
- Converting "move" constructors done differently in order to truly
support:
unique_ptr<Derived> func_returning_unique_ptr (.....);
...
unique_ptr<Base> ptr = func_returning_unique_ptr (.....);
At this point, it no longer shares much at all with the original file,
but, that's the history.
See comments in the code to find out more.
I thought of putting the "emulation" / shim in the "std" namespace, so
that when we start requiring C++11 at some point, no actual changes to
users of the smart pointer throughout would be necessary. Putting
things in the std namespace is technically undefined, however in
practice it doesn't cause any issue with any compiler. However,
thinking that people might be confused with seeing std::unique_ptr and
thinking that we're actually requiring C++11 already, I put the new
types in the "gdb" namespace instead.
For managing xmalloc pointers, this adds a gdb::unique_xmalloc_ptr<T>
"specialization" with a custom xfree deleter.
No actual use of any smart pointer is introduced in this patch.
That'll be done in following patches.
Tested (along with the rest of the series) on:
- NetBSD 5.1 (gcc70 on the compile farm), w/ gcc 4.1.3
- x86-64 Fedora 23, gcc 5.3.1 (gnu++03)
- x86-64 Fedora 23, and gcc 7.0 (gnu++14)
gdb/ChangeLog:
2016-10-18 Pedro Alves <palves@redhat.com>
* common/common-defs.h: Include "gdb_unique_ptr.h".
* common/gdb_unique_ptr.h: New.
Make sure the architecture is `bfd_arch_i386' before handling the `set
mpx' and `show mpx' commands, avoiding the issue with `i386_mpx_enabled'
interpreting `gdbarch->tdep' according to the `struct gdbarch_tdep'
definition in i386-tdep.h while indeed in a multi-target configuration
it may have a different layout and cause GDB to crash or at least
misbehave.
gdb/
* i386-tdep.c (i386_mpx_info_bounds): Make sure the architecture
is `bfd_arch_i386' before proceeding.
(i386_mpx_set_bounds): Likewise.
Fix a load of aliasing build errors:
cc1plus: warnings being treated as errors
.../gdb/tilegx-tdep.c: In function 'CORE_ADDR tilegx_analyze_prologue(gdbarch*, CORE_ADDR, CORE_ADDR, tilegx_frame_cache*, frame_info*)':
.../gdb/tilegx-tdep.c:609: error: dereferencing pointer 'operands' does break strict-aliasing rules
.../gdb/tilegx-tdep.c:592: error: dereferencing pointer 'operands' does break strict-aliasing rules
.../gdb/tilegx-tdep.c:571: error: dereferencing pointer 'operands' does break strict-aliasing rules
[...]
.../gdb/tilegx-tdep.c:601: error: dereferencing pointer '<anonymous>' does break strict-aliasing rules
.../gdb/tilegx-tdep.c:601: note: initialized from here
cc1plus: error: dereferencing pointer 'operands' does break strict-aliasing rules
cc1plus: error: dereferencing pointer 'operands' does break strict-aliasing rules
.../gdb/tilegx-tdep.c:452: note: initialized from here
cc1plus: error: dereferencing pointer 'pretmp.896' does break strict-aliasing rules
cc1plus: note: initialized from here
cc1plus: error: dereferencing pointer 'pretmp.896' does break strict-aliasing rules
cc1plus: note: initialized from here
make[1]: *** [tilegx-tdep.o] Error 1
from an attempt to cast a `long long' pointer to an `int64_t' pointer,
which may not necessarily be compatible types. Use the `long long' type
for the auxiliary variable then as this is the type of the structure
member referred.
gdb/
* tilegx-tdep.c (tilegx_analyze_prologue): Use the `long long'
type for `operands'.
Fix a commit 4df068de52 ("Add support for SVE addressing modes") build
regression:
cc1: warnings being treated as errors
.../opcodes/aarch64-dis.c: In function 'aarch64_ext_sve_addr_rr_lsl':
.../opcodes/aarch64-dis.c:1324: error: declaration of 'index' shadows a global declaration
/usr/include/string.h:303: error: shadowed declaration is here
make[3]: *** [aarch64-dis.lo] Error 1
in a way following commit 91d6fa6a03 ("Add -Wshadow to the gcc command
line options used when compiling the binutils.").
opcodes/
* aarch64-dis.c (aarch64_ext_sve_addr_rr_lsl): Rename `index'
local variable to `index_regno'.
Fix an aliasing build error:
cc1: warnings being treated as errors
.../bfd/elf32-nds32.c: In function 'nds32_elf_check_relocs':
.../bfd/elf32-nds32.c:6644: warning: dereferencing type-punned pointer will break strict-aliasing rules
make[3]: *** [elf32-nds32.lo] Error 1
in a way following commit 6edfbbad08 ("Fix up gcc4.1 aliasing
warnings"), <https://sourceware.org/ml/binutils/2005-10/msg00071.html>.
bfd/
* elf32-nds32.c (nds32_elf_check_relocs): Avoid aliasing warning
from GCC.
Fix a commit 5025eb7c0d ("Delete relocations associatesd with deleted
exidx entries.") build regression:
cc1: warnings being treated as errors
.../bfd/elf32-arm.c: In function 'elf32_arm_update_relocs':
.../bfd/elf32-arm.c:14951: warning: declaration of 'index' shadows a global declaration
/usr/include/string.h:304: warning: shadowed declaration is here
make[3]: *** [elf32-arm.lo] Error 1
in a way following commit 91d6fa6a03 ("Add -Wshadow to the gcc command
line options used when compiling the binutils.").
bfd/
* elf32-arm.c (elf32_arm_update_relocs): Rename `index' local
variable to `reloc_index'.
This fixes a comment I forgot to update in the previous patch.
gdb/testsuite/ChangeLog:
* gdb.trace/mi-trace-save.exp (test_trace_save_wrong_num_args):
Update comment.
-trace-save doesn't check whether an argument is passed, leading to a
segfault if you pass nothing.
I added a small test, which only tests the error conditions of
-trace-save.
gdb/ChangeLog:
* mi/mi-main.c (mi_cmd_trace_save): Check if argument is present
before using it.
gdb/testsuite/ChangeLog:
* gdb.trace/mi-trace-save.exp: New file.
Cross building gdb for mingw32 on Fedora 23 fails with:
x86_64-w64-mingw32-g++ -g -O2 [...] ../../src/gdb/charset.c
In file included from ../../src/gdb/charset.c:21:0:
../../src/gdb/charset.h:134:3: error: 'iconv_t' does not name a type
iconv_t m_desc;
^
../../src/gdb/charset.c: In constructor 'wchar_iterator::wchar_iterator(const gdb_byte*, size_t, const char*, size_t)':
../../src/gdb/charset.c:600:3: error: 'm_desc' was not declared in this scope
m_desc = iconv_open (INTERMEDIATE_ENCODING, charset);
^
../../src/gdb/charset.c: In destructor 'wchar_iterator::~wchar_iterator()':
../../src/gdb/charset.c:607:7: error: 'm_desc' was not declared in this scope
if (m_desc != (iconv_t) -1)
^
../../src/gdb/charset.c: In member function 'int wchar_iterator::iterate(wchar_iterate_result*, gdb_wchar_t**, const gdb_byte**, size_t*)':
../../src/gdb/charset.c:633:25: error: 'm_desc' was not declared in this scope
size_t r = iconv (m_desc, &inptr, &m_bytes, &outptr, &out_avail);
^
This is a regression caused by commit cda6c55bd3 (Turn wchar
iterator into a class). The problem is that iconv_t is now exposed in
gdb/charset.h, while before it was only used in gdb/charset.c.
gdb/charset.c, under #ifdef PHONY_ICONV, does:
#undef iconv_t
#define iconv_t int
So it seems the simplest is to use 'int' in the header file too.
gdb/ChangeLog:
2016-10-17 Pedro Alves <palves@redhat.com>
* charset.h (class wchar_iterator) [PHONY_ICONV] <m_desc>: Use
'int' as type.
* readelf.c (apply_relocations): Fail if the symbol table section
linked to by the reloc section does not have either the SHT_SYMTAB
or SHT_DYNSYM type.
(print_gnu_note): Decode the contents of NT_GNU_HWCAP notes.
Print the contents of unknown note types.
(process_note): Add the file and section to the parameter list.
Use print_symbol to display the note name.
Display the contents of unknown note types.
(process_corefile_note_segment): Rename to process_notes_at.
Add section parameter. Apply relocations to the notes when
loading from a section. Display section name when processing
notes in a section.
* testsuite/binutils-all/readelf.n: Update expected output.
The disassember was generating invXXX instructions for cases when in reality we
had llockd or scondd instrutions.
opcodes/ChangeLog:
Cupertino Miranda <cmiranda@synopsys.com>
arc-tbl.h: Removed any "inv.+" instructions from the table.
gas/ChangeLog:
Cupertino Miranda <cmiranda@synopsys.com>
testsuite/arc/dis-inv.s: Test to validate patch.
testsuite/arc/dis-inv.d: Likewise.
2016-09-19 Andrew Stubbs <ams@codesourcery.com>
* pex-win32.c (argv_to_cmdline): Quote zero-length parameters.
* testsuite/test-pexecute.c (main): Insert check for zero-length parameters.
2016-09-10 Mark Wielaard <mjw@redhat.com>
* cp-demangle.c (d_substitution): Change struct demangle_component
variable name from c to dc.
2016-08-12 Marek Polacek <polacek@redhat.com>
PR c/7652
* cp-demangle.c (d_print_mod): Add FALLTHRU.
2016-08-04 Marcel B?hme <boehme.marcel@gmail.com>
PR c++/71696
* cplus-dem.c: Prevent infinite recursion when there is a cycle
in the referencing of remembered mangled types.
(work_stuff): New stack to keep track of the remembered mangled
types that are currently being processed.
(push_processed_type): New method to push currently processed
remembered type onto the stack.
(pop_processed_type): New method to pop currently processed
remembered type from the stack.
(work_stuff_copy_to_from): Copy values of new variables.
(delete_non_B_K_work_stuff): Free stack memory.
(demangle_args): Push/Pop currently processed remembered type.
(do_type): Do not demangle a cyclic reference and push/pop
referenced remembered type.
So that GDB and other projects that share the top level can use them.
Bootstrapped with all default languages + jit on x86-64 Fedora 23.
gcc/ChangeLog:
2016-10-14 Pedro Alves <palves@redhat.com>
* coretypes.h (OVERRIDE, FINAL): Delete, moved to
include/ansidecl.h.
include/ChangeLog:
2016-10-14 Pedro Alves <palves@redhat.com>
* ansidecl.h (GCC_FINAL): Delete.
(OVERRIDE, FINAL): New, moved from gcc/coretypes.h.
The -trace-save MI command supports saving the trace in the CTF format,
as its CLI counterpart, but it's not documented.
gdb/doc/ChangeLog:
* gdb.texinfo (GDB/MI Tracepoint Commands): Document -ctf switch
of -trace-save.
As pointed out in PR binutils/19020, those tests were still failing when doing
remote testing. This is because the binary files weren't being copied over to
the remote host for testing.
FAIL: binutils-all/pr19020a
FAIL: binutils-all/pr19020b
This small patch fixes this up to make things pass.
binutils/ChangeLog:
2016-10-14 Luis Machado <lgustavo@codesourcery.com>
* testsuite/lib/utils-lib.exp (run_dump_test): Call remote_download
to copy file to remote host.
set sysroot command on AIX has no effect if a program depends on shared
library archives (.a). Fixed by using solib_find and solib_bfd_fopen
instead of gdb_bfd_open in solib_aix_bfd_open.
gdb/
2016-10-14 Sangamesh Mallayya <sangamesh.swamy@in.ibm.com>
Ulrich Weigand <uweigand@de.ibm.com>
* solib-aix.c (solib_aix_bfd_open): Call solib_find so that sysroot
path is set properly if program has a dependency on .a archive and
sysroot is set via set sysroot command.
Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
The ARC (short) instructions are using a special register number to
indicate is the instruction uses a long immediate (LIMM). In the case
of short instruction, this LIMM indicator depends on the ISA version
used. Thus, for ARCv1 processors, the LIMM indicator is 0x3E, the same
value used in "long" instructions. However, for the ARCv2 processors,
this LIMM indicator is 0x1E.
This patch fixes the LIMM detection for ARCv1 ISA and adds two tests.
gas/
2016-10-13 Claudiu Zissulescu <claziss@synopsys.com>
* testsuite/gas/arc/shortlimm_a7.d: New file.
* testsuite/gas/arc/shortlimm_a7.s: Likewise.
* testsuite/gas/arc/shortlimm_hs.d: Likewise.
* testsuite/gas/arc/shortlimm_hs.s: Likewise.
include/
2016-10-13 Claudiu Zissulescu <claziss@synopsys.com>
* opcode/arc.h (ARC_OPCODE_ARCV2): New define.
opcodes/
2016-10-13 Claudiu Zissulescu <claziss@synopsys.com>
* arc-dis.c (find_format_from_table): Discriminate LIMM indicator
usage on ISA basis.
gdb/ChangeLog
2016-10-14 Eli Zaretskii <eliz@gnu.org>
* common/common-defs.h [HAVE_STRINGS_H]: Include strings.h if
available, to get prototypes of 'strcasecmp' and 'strncasecmp'.
These don't make sense with C++.
gdb/ChangeLog:
2016-10-13 Pedro Alves <palves@redhat.com>
* contrib/ari/gdb_ari.sh (boolean): Suggest bool instead.
(false, true): Remove checks.
This patch skips some tests related to floating point in structs.exp
if gdb_skip_float_test return false.
gdb/testsuite:
2016-10-13 Yao Qi <yao.qi@linaro.org>
* gdb.base/structs.exp: Invoke gdb_skip_float_test, and do
floating point tests if $skip_float_test is false.
This changes macho_symfile_read_all_oso to use std::string. This
avoids a cleanup.
2016-10-12 Tom Tromey <tom@tromey.com>
* machoread.c (macho_symfile_read_all_oso): Use std::string.
This changes wchar_iterator from charset.c into a real C++ class, then
updates the users to use the class. This lets us remove some cleanups
in favor of the class' destructor.
2016-10-12 Tom Tromey <tom@tromey.com>
* valprint.c (generic_emit_char, count_next_character)
(generic_printstr): Update.
* charset.c (struct wchar_iterator): Move to charset.h.
(wchar_iterator::wchar_iterator): Rename from
make_wchar_iterator, turn into a constructor.
(wchar_iterator::~wchar_iterator): Rename from
do_cleanup_iterator, turn into a destructor.
(make_cleanup_wchar_iterator): Remove.
(wchar_iterator::iterate): Rename from wchar_iterate. Remove
"iter" argument. Update.
* charset.h: Include <vector>.
(class wchar_iterator): New class, from old struct
wchar_iterator.
(make_wchar_iterator, make_cleanup_wchar_iterator): Don't
declare.
This patch changes selftest.c to use std::vector rather than VEC.
I think this is a small net plus.
2016-10-12 Tom Tromey <tom@tromey.com>
* selftest.c: Include <vector>, not "vec.h".
(self_test_function_ptr): Remove.
(tests): Now a std::vector.
(register_self_test, run_self_tests): Update.
This converts tid_range_parser and get_number_or_range to be classes.
The various tid_range_parser_* and get_number_or_range_* functions
become methods on the respective classes. Then it updates the users
to follow.
The rationale for the change is that this provides better
encapsulation. For example, this forced me to think of a better
interface between tid_range_parser and get_number_or_range, since the
former peeked into the latter's internals a bit too much. That ended
up resulting mostly in these two not-just-straight-1-1 changes:
void
-tid_range_parser_skip (struct tid_range_parser *parser)
+tid_range_parser::skip_range ()
{
...
- tid_range_parser_init (parser, parser->range_parser.end_ptr,
- parser->default_inferior);
+ m_range_parser.skip_range ();
+ init (m_range_parser.string (), m_default_inferior);
}
and:
/* If we successfully parsed a thread number or finished parsing a
thread range, switch back to assuming the next TID is
inferior-qualified. */
- if (parser->range_parser.end_ptr == NULL
- || parser->range_parser.string == parser->range_parser.end_ptr)
+ if (!m_range_parser.in_range ())
{
For the same reason (encapsulation), this moves the enum
tid_range_state definition to within the tid_parser class's scope,
since that is private implementation detail.
While at it, switch to use "bool" for booleans.
gdb/ChangeLog:
2016-10-13 Pedro Alves <palves@redhat.com>
Tom Tromey <tom@tromey.com>
* tid-parse.h (tid_range_parser): New class.
(enum tid_range_state): Move into tid_range_parser's scope.
Remove TID_RANGE_ prefix from all values.
(tid_range_parser_get_tid, tid_range_parser_get_tid_range)
(tid_range_parser_star_range, tid_range_parser_finished)
(tid_range_parser_skip, tid_range_parser_qualified): Don't
declare.
(tid_is_in_list): Update comment.
* tid-parse.c (tid_range_parser::tid_range_parser): New.
(init, finished, get_string, skip, tid_is_qualified)
(get_tid_or_range, get_tid_range, get_tid, star_range): Rename;
turn into methods.
(tid_is_in_list): Adjust.
* cli/cli-utils.h (number_or_range_parser): New class.
(init_number_or_range, get_number_or_range)
(number_range_setup_range): Don't declare.
* cli/cli-utils.c
(number_or_range_parser::number_or_range_parser): New.
(init_number_or_range, get_number_or_range)
(number_range_setup_range): Rename; turn into methods.
(number_is_in_list): Adjust.
* breakpoint.c (map_breakpoint_numbers): Adjust. Use bool.
(trace_pass_command, get_tracepoint_by_number): Adjust.
* breakpoint.h (get_tracepoint_by_number): Adjust.
* inferior.c (detach_inferior_command, kill_inferior_command)
(remove_inferior_command): Adjust.
* linespec.c (decode_line_2): Adjust.
* memattr.c (mem_enable_command, mem_disable_command)
(mem_delete_command): Adjust.
* printcmd.c (map_display_numbers): Adjust.
* reverse.c (delete_bookmark_command, bookmarks_info): Adjust.
* thread.c (thread_apply_command): Adjust.
I noticed that testing aarch64-elf gdb with a physical board
ran into issues with gdb.python/py-value.exp. Further investigation showed
that we were actually trying to dereference a NULL pointer (argv) when trying
to access argv[0].
Being bare-metal, argv is not guaranteed to be valid. So we need to make sure
argv is sane before accessing argv[0].
The following patch fixes up the test program to check for a NULL argv and also
improves the testcase a bit so it doesn't have to work with a hardcoded argc
value.
Regression-tested on x86-64 Ubuntu 16.04.
gdb/testsuite/ChangeLog:
2016-10-12 Luis Machado <lgustavo@codesourcery.com>
* gdb.python/py-value.c (main): Check if argv is NULL before using it.
* gdb.python/py-value.exp (test_value_in_inferior): Don't use hardcoded
argc values.
Add 1 to argc so we guarantee distinct initial/modified argc values.
Add support for Newlib as an OS/ABI. The only thing that is specific to it
relatively to "generic" baremetal target is location of PC register in jump
buffer for longjump support.
Sniffer uses .ivt section to decide if ELF file is for ARC Newlib or not.
gdb/ChangeLog:
* arc-newlib-tdep.c: New file.
* configure.tgt: Add newlib support for ARC.
Standard get_longjmp_target implementation, similar to what is in arm-tdep.c.
Actual value of jb_pc should be set in init_osabi methods of particular OS/ABI
implementations.
gdb/ChangeLog:
* arc-tdep.h (struct gdbarch_tdep) <jb_pc>: New field.
* arc-tdep.c (arc_get_longjmp_target): New function.
(arc_gdbarch_init): Set get_longjmp_target if jb_pc is non-negative.
(arc_dump_tdep): Print jb_pc.
We don't track FP registers in aarch64 prologue analyzer, so this causes
an internal error when FP registers are saved by "stp" instruction in
prologue (stp d8, d9, [sp,#128]),
tbreak _Unwind_RaiseException^M
aarch64-tdep.c:335: internal-error: CORE_ADDR aarch64_analyze_prologue(gdbarch*, CORE_ADDR, CORE_ADDR, aarch64_prologue_cache*): Assertion `inst.operands[0].type == AARCH64_OPND_Rt' failed.^M
A problem internal to GDB has been detected,
This patch teaches GDB to track FP registers (D registers) in prologue
analyzer.
gdb:
2016-10-12 Yao Qi <yao.qi@linaro.org>
PR tdep/20682
* aarch64-tdep.c: Replace 32 with AARCH64_D_REGISTER_COUNT.
(aarch64_analyze_prologue): Extend array 'regs' for D registers.
Assert that operand 0 and 1 can be X or D registers. Update
register number for D registers. Update registers in frame
cache.
* aarch64-tdep.h (AARCH64_D_REGISTER_COUNT): New macro.
