This commit enhances type_align to handle TYPE_CODE_RANGE types
the same as integers and enums, rather than returning zero,
which means for this function that it could not determine its
alignment.
gdb/ChangeLog:
* gdbtypes.c (type_align): Handle TYPE_CODE_RANGE the same as
integers and enumeration types.
Tested on x86_64-linux. Also tested on a variety of platforms
(with CPUs being ARM, AArch64, Leon3 (SPARC-like), PowerPC,
PowerPC64, RV64, Visium, x86, x86_64).
Trying to print a packed array sometimes leads to a crash (see
attached testcase for an example of when this happens):
| (gdb) p bad
| [1] 65571 segmentation fault gdb -q foo
Variable "bad" is declared in the debug information as an array where
the array's type name has an XPnnn suffix:
| .uleb128 0xc # (DIE (0x566) DW_TAG_typedef)
| .long .LASF200 # DW_AT_name: "pck__t___XP1"
| [loc info attributes snipped]
| .long 0x550 # DW_AT_type
| .byte 0x1 # DW_AT_alignment
The signals to GDB that the debugging information follows a GNAT encoding
used for packed arrays, and an in order to decode it, we need to find
the type whose name is the same minus the "___XPnnn" suffix: "pck__t".
For that, we make a call to ada-lang.c::standard_lookup, which is
a simple function which essentially does:
| /* Return the result of a standard (literal, C-like) lookup of NAME in
| given DOMAIN, visible from lexical block BLOCK. */
|
| [...]
| sym = lookup_symbol_in_language (name, block, domain, language_c, 0);
Unfortunately for us, while the intent of this call was to perform
an exact-match lookup, in our case, it returns ... type pck__t___XP1
instead! In other words, it finds itself back. The reason why it finds
this type is a confluence of two factors:
(1) Forcing the lookup into language_c currently does not affect
how symbol matching is done anymore, because we look at the symbol's
language to determine which kind of matching should be done;
(2) The lookup searches the local context (via block) first, beforei
doing a more general lookup. And looking at the debug info for
the main subprogram, we see that type "pck__t" is not declared
there, only in the debug info for pck.ads. In other words,
there is no way that we accidently find "pck__t" by random chance.
I believe Pedro added a new function called ada_lookup_encoded_symbol
for that specific purpose, so I started by replacing the lookup
by language above by this. Unfortunately, still no joy.
This was because, even though ada_lookup_encoded_symbol puts angle-
brackets around the search name to signal that we want a verbatim
search, we end up losing that information in the function called
to compare a symbol with the search name:
| static bool
| do_full_match (const char *symbol_search_name,
| const lookup_name_info &lookup_name,
| completion_match_result *comp_match_res)
| {
| return full_match (symbol_search_name, ada_lookup_name (lookup_name));
^^^^^^^^^^^^^^^
|
<=> lookup_name.m_ada.m_encoded_name
(no angle brackets)
The way I fixed this was by introducing a new function called
do_exact_match, and then adjust ada_get_symbol_name_matcher to
return that function when seeing that we have a verbatim non-wild-match
search.
As it happens, this fixes an incorrect test in gdb.ada/homony.exp,
where we were inserting a breakpoint on a symbol using the angle-brackets
notation, and got 2 locations for that breakpoint...
(gdb) b <homonym__get_value>
Breakpoint 1 at 0x4029fc: <homonym__get_value>. (2 locations)
... each location being in a different function:
(gdb) info break
Num Type Disp Enb Address What
1 breakpoint keep y <MULTIPLE>
1.1 y 0x00000000004029fc in homonym.get_value
at /[...]/homonym.adb:32
1.2 y 0x0000000000402a3a in homonym.get_value
at /[...]/homonym.adb:50
(gdb) x /i 0x00000000004029fc
0x4029fc <homonym__get_value+8>: movl $0x1d,-0x4(%rbp)
(gdb) x /i 0x0000000000402a3a
0x402a3a <homonym__get_value__2+8>: movl $0x11,-0x4(%rbp)
Since we used angle-brackets, we shouldn't be matching the second one,
something this patch fixes.
gdb/ChangeLog:
* ada-lang.c (standard_lookup): Use ada_lookup_encoded_symbol
instead of lookup_symbol_in_language
(do_exact_match): New function.
(ada_get_symbol_name_matcher): Return do_exact_match when
doing a verbatim match.
gdb/testsuite/ChangeLog:
* gdb.ada/big_packed_array: New testcase.
* gdb.ada/homonym.exp: Fix incorrect expected output for
"break <homonym__get_value>" test.
Tested on x86_64-linux.
ravenscar-thread.c intercepts resume and wait target requests and
replaces the requested ptid with the ptid of the underlying CPU.
However, this is incorrect when a request is made with a wildcard
ptid.
This patch adds a special case to ravenscar-thread.c for
minus_one_ptid. I don't believe a special case for process wildcards
is necessary, so I have not added that.
Joel's description explains the bug well:
At the user level, we noticed the issue because we had a test were
we insert a breakpoint one some code which is only run from, say,
CPU #2, whereas we unfortunately resumed the execution after having
stopped somewhere in CPU #1. As a result, we sent an order to resume
CPU #1, which starves CPU #2 forever, because the code in CPU #1
waits for some of the Ada tasks allocated to CPU #2 (and we never
reach our breakpoint either).
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* ravenscar-thread.c (ravenscar_thread_target::resume)
(ravenscar_thread_target::wait): Special case wildcard requests.
This changes ravenscar-thread.c to make it ready for multi-target.
This is done by moving globals into the target, and then arranging to
allocate the target with "new" and delete the target in its "close"
method.
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* ravenscar-thread.c (base_ptid): Remove.
(struct ravenscar_thread_target) <close>: New method.
<m_base_ptid>: New member.
<update_inferior_ptid, active_task, task_is_currently_active,
runtime_initialized>: Declare methods.
<ravenscar_thread_target>: Add constructor.
(ravenscar_thread_target::task_is_currently_active)
(ravenscar_thread_target::update_inferior_ptid)
(ravenscar_runtime_initialized): Rename. Now methods.
(ravenscar_thread_target::resume, ravenscar_thread_target::wait)
(ravenscar_thread_target::update_thread_list): Update.
(ravenscar_thread_target::active_task): Now method.
(ravenscar_thread_target::store_registers)
(ravenscar_thread_target::prepare_to_store)
(ravenscar_thread_target::prepare_to_store)
(ravenscar_thread_target::mourn_inferior): Update.
(ravenscar_inferior_created): Use "new" to create target.
(ravenscar_thread_target::get_ada_task_ptid): Update.
(_initialize_ravenscar): Don't initialize base_ptid.
(ravenscar_ops): Remove global.
This adds a push_target overload that takes a "target_ops_up &&".
This removes some calls to release a target_ops_up, and makes the
intent here clearer.
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* target.h (push_target): Declare new overload.
* target.c (push_target): New overload, taking an rvalue reference.
* remote.c (remote_target::open_1): Use push_target overload.
* corelow.c (core_target_open): Use push_target overload.
This changes some functions in ravenscar-thread.c to return "bool"
rather than int, where appropriate, and also changes "(void)" to "()".
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* ravenscar-thread.c (is_ravenscar_task)
(ravenscar_task_is_currently_active): Return bool.
(ravenscar_update_inferior_ptid, get_running_thread_msymbol)
(_initialize_ravenscar): Remove "(void)".
(has_ravenscar_runtime, ravenscar_runtime_initialized): Likewise.
Return bool.
This fixes some incorrect formatting in ravenscar-thread.c.
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* ravenscar-thread.c (ravenscar_runtime_initializer)
(has_ravenscar_runtime, get_running_thread_id)
(ravenscar_thread_target::resume): Fix indentation.
This turns ravenscar_arch_ops into an abstract base class and updates
all the places where it is used. This is an improvement because it
avoids any possibility of forgetting to set one of the function
pointers. It also makes clear that these functions aren't intended to
be changed dynamically.
This version of the patch removes the prepare_to_store method, as it
is unused, and it is easy enough to add if it is ever needed.
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* sparc-ravenscar-thread.c (struct sparc_ravenscar_ops): Derive
from ravenscar_arch_ops.
(sparc_ravenscar_ops::fetch_registers)
(sparc_ravenscar_ops::store_registers): Now methods.
(sparc_ravenscar_prepare_to_store): Remove.
(sparc_ravenscar_ops): Redefine.
* ravenscar-thread.h (struct ravenscar_arch_ops): Add virtual
methods and destructor. Remove members.
* ravenscar-thread.c (ravenscar_thread_target::fetch_registers)
(ravenscar_thread_target::store_registers)
(ravenscar_thread_target::prepare_to_store): Update.
* ppc-ravenscar-thread.c (ppc_ravenscar_generic_prepare_to_store):
Remove.
(struct ppc_ravenscar_powerpc_ops): Derive from
ravenscar_arch_ops.
(ppc_ravenscar_powerpc_ops::fetch_registers)
(ppc_ravenscar_powerpc_ops::store_registers): Now methods.
(ppc_ravenscar_powerpc_ops): Redefine.
(struct ppc_ravenscar_e500_ops): Derive from ravenscar_arch_ops.
(ppc_ravenscar_e500_ops::fetch_registers)
(ppc_ravenscar_e500_ops::store_registers): Now methods.
(ppc_ravenscar_e500_ops): Redefine.
* aarch64-ravenscar-thread.c
(aarch64_ravenscar_generic_prepare_to_store): Remove.
(struct aarch64_ravenscar_ops): Derive from ravenscar_arch_ops.
(aarch64_ravenscar_fetch_registers)
(aarch64_ravenscar_store_registers): Now methods.
(aarch64_ravenscar_ops): Redefine.
This changes some code in ravenscar-thread.c to use scoped_restore. I
am not sure if it matters in practice, but this makes these methods
exception-safe in case the methods lower in the target stack can
throw.
gdb/ChangeLog
2019-02-15 Tom Tromey <tromey@adacore.com>
* ravenscar-thread.c (ravenscar_thread_target::stopped_by_sw_breakpoint)
(ravenscar_thread_target::stopped_by_hw_breakpoint)
(ravenscar_thread_target::stopped_by_watchpoint)
(ravenscar_thread_target::stopped_data_address)
(ravenscar_thread_target::core_of_thread): Use scoped_restore.
Phillipe noticed that create_ada_exception_catchpoint was not freeing
the "addr_string" memory:
==14141== 114 bytes in 4 blocks are definitely lost in loss record 1,054 of 3,424
==14141== at 0x4C2BE6D: malloc (vg_replace_malloc.c:309)
==14141== by 0x405107: xmalloc (common-utils.c:44)
==14141== by 0x7563F9: xstrdup (xstrdup.c:34)
==14141== by 0x381B21: ada_exception_sal (ada-lang.c:13217)
==14141== by 0x381B21: create_ada_exception_catchpoint(gdbarch*, ada_exception_catchpoint_kind, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int, int, int) (ada-lang.c:13251)
==14141== by 0x3820A8: catch_ada_exception_command(char const*, int, cmd_list_element*) (ada-lang.c:13285)
==14141== by 0x3F4828: cmd_func(cmd_list_element*, char const*, int) (cli-decode.c:1892)
This patch fixes the problem by changing ada_exception_sal to return a
std::string via its out parameter.
gdb/ChangeLog
2019-02-15 Philippe Waroquiers <philippe.waroquiers@skynet.be>
Tom Tromey <tromey@adacore.com>
* ada-lang.c (ada_exception_sal): Change addr_string to a
std::string.
(create_ada_exception_catchpoint): Update.
Philippe noticed a memory leak coming from ada_catchpoint_location --
it was not freeing the "function_name" member from its base class:
==14141== 114 bytes in 4 blocks are definitely lost in loss record 1,055 of 3,424
==14141== at 0x4C2BE6D: malloc (vg_replace_malloc.c:309)
==14141== by 0x405107: xmalloc (common-utils.c:44)
==14141== by 0x7563F9: xstrdup (xstrdup.c:34)
==14141== by 0x3B82B3: set_breakpoint_location_function(bp_location*, int) (breakpoint.c:7156)
==14141== by 0x3C112B: add_location_to_breakpoint(breakpoint*, symtab_and_line const*) (breakpoint.c:8609)
==14141== by 0x3C127A: init_raw_breakpoint(breakpoint*, gdbarch*, symtab_and_line, bptype, breakpoint_ops const*) (breakpoint.c:7187)
==14141== by 0x3C1B52: init_ada_exception_breakpoint(breakpoint*, gdbarch*, symtab_and_line, char const*, breakpoint_ops const*, int, int, int) (breakpoint.c:11262)
==14141== by 0x381C2E: create_ada_exception_catchpoint(gdbarch*, ada_exception_catchpoint_kind, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int, int, int) (ada-lang.c:13255)
This patch fixes the problem by further C++-ifying bp_location. In
particular, bp_location_ops is now removed, and the "dtor" function
pointer is replaced with an ordinary destructor.
gdb/ChangeLog
2019-02-15 Philippe Waroquiers <philippe.waroquiers@skynet.be>
Tom Tromey <tromey@adacore.com>
* breakpoint.c (~bp_location): Rename from bp_location_dtor.
(bp_location_ops): Remove.
(base_breakpoint_allocate_location): Update.
(free_bp_location): Update.
* ada-lang.c (class ada_catchpoint_location)
<ada_catchpoint_location>: Remove ops parameter.
(ada_catchpoint_location_dtor): Remove.
(ada_catchpoint_location_ops): Remove.
(allocate_location_exception): Update.
* breakpoint.h (struct bp_location_ops): Remove.
(class bp_location) <bp_location>: Remove bp_location_ops
parameter.
<~bp_location>: Add destructor.
<ops>: Remove.
... as a follow-up to commit d981640286 "Run more
ld tests when not native", which replaced by a proper solution the following
mess before present in 'ld/configure.host':
-*-*-gnu*)
- # When creating static executables, we ought to use crt0.o instead of crt1.o,
- # <http://www.gnu.org/software/hurd/open_issues/binutils.html#static>,
- # but the testing infrastructure is not prepared for that. This is not
- # relevant for most tests, and the few remaining ones have been XFAILed.
- HOSTING_CRT0='[...]'
- HOSTING_LIBS='[...]'
ld/
* testsuite/ld-elf/elf.exp: Remove Hurd XFAILs.
..., which is not defined in GNU/Hurd systems, and so commit
94585166df "Extended-remote follow-exec" caused:
[...]/gdb/remote.c: In member function 'void remote_target::remote_parse_stop_reply(const char*, stop_reply*)':
[...]/gdb/remote.c:7343:22: error: 'PATH_MAX' was not declared in this scope
char pathname[PATH_MAX];
^~~~~~~~
gdb/
* remote.c (remote_target::remote_parse_stop_reply): Avoid using
'PATH_MAX'.
Hurd's commit baf7e5c8ce176aead15c2559952d8bdf0da41ffd "hurd: Use polymorphic
port types to return some rights" causes in the GDB build:
/usr/bin/ld: process_reply_S.o: in function `_Xproc_pid2proc_reply':
[...]/gdb/process_reply_S.c:754: undefined reference to `S_proc_pid2proc_reply'
/usr/bin/ld: [...]/gdb/process_reply_S.c:730: undefined reference to `S_proc_pid2proc_reply'
/usr/bin/ld: process_reply_S.o: in function `_Xproc_task2proc_reply':
[...]/gdb/process_reply_S.c:589: undefined reference to `S_proc_task2proc_reply'
/usr/bin/ld: [...]/gdb/process_reply_S.c:565: undefined reference to `S_proc_task2proc_reply'
/usr/bin/ld: process_reply_S.o: in function `_Xproc_getmsgport_reply':
[...]/gdb/process_reply_S.c:204: undefined reference to `S_proc_getmsgport_reply'
/usr/bin/ld: [...]/gdb/process_reply_S.c:180: undefined reference to `S_proc_getmsgport_reply'
collect2: error: ld returned 1 exit status
gdb/
* gnu-nat.c (S_proc_getmsgport_reply, S_proc_task2proc_reply)
(S_proc_pid2proc_reply): Adjust to Hurd "proc" interface changes.
... that appeared with 9bf2a70066
"-Wwrite-strings: Remove -Wno-write-strings".
gdb/
* gnu-nat.c (gnu_write_inferior, parse_int_arg, _parse_bool_arg)
(check_empty): Use "const char *".
..., that is commit 00431a78b2 causing:
[...]/gdb/gnu-nat.c: In member function 'virtual void gnu_nat_target::detach(inferior*, int)':
[...]/gdb/gnu-nat.c:2284:23: error: invalid conversion from 'int' to 'inferior*' [-fpermissive]
detach_inferior (pid);
^
In file included from [...]/gdb/gnu-nat.c:61:0:
[...]/gdb/inferior.h:523:13: note: initializing argument 1 of 'void detach_inferior(inferior*)'
extern void detach_inferior (inferior *inf);
^~~~~~~~~~~~~~~
Fixed by inlining the removed code.
gdb/
* gnu-nat.c (gnu_nat_target::detach): Instead of
'detach_inferior (pid)' call
'detach_inferior (find_inferior_pid (pid))'.
..., that is commit f6ac5f3d63 causing:
In file included from [...]/gdb/gnu-nat.c:24:0:
[...]/gdb/gnu-nat.h:123:1: error: expected class-name before '{' token
{
^
[...]/gdb/gnu-nat.h:128:16: error: 'inferior' has not been declared
void detach (inferior *, int) override;
^~~~~~~~
[...]/gdb/gnu-nat.h:132:8: error: use of enum 'target_xfer_status' without previous declaration
enum target_xfer_status xfer_partial (enum target_object object,
^~~~~~~~~~~~~~~~~~
[...]/gdb/gnu-nat.h:132:46: error: use of enum 'target_object' without previous declaration
enum target_xfer_status xfer_partial (enum target_object object,
^~~~~~~~~~~~~
[...]/gdb/gnu-nat.h:124:8: error: 'void gnu_nat_target::attach(const char*, int)' marked 'override', but does not override
void attach (const char *, int) override;
^~~~~~
[...]
[...]/gdb/gnu-nat.c: In member function 'virtual void gnu_nat_target::detach(inferior*, int)':
[...]/gdb/gnu-nat.c:2286:34: error: 'ops' was not declared in this scope
inf_child_maybe_unpush_target (ops);
^~~
[...]/gdb/gnu-nat.c:2286:34: note: suggested alternative: 'open'
inf_child_maybe_unpush_target (ops);
^~~
open
[...]/gdb/gnu-nat.c:2286:3: error: 'inf_child_maybe_unpush_target' was not declared in this scope
inf_child_maybe_unpush_target (ops);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[...]/gdb/gnu-nat.c:2286:3: note: suggested alternative: 'maybe_unpush_target'
inf_child_maybe_unpush_target (ops);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
maybe_unpush_target
[...]/gdb/i386-gnu-nat.c:200:1: warning: 'void gnu_store_registers(target_ops*, regcache*, int)' defined but not used [-Wunused-function]
gnu_store_registers (struct target_ops *ops,
^~~~~~~~~~~~~~~~~~~
[...]/gdb/i386-gnu-nat.c:109:1: warning: 'void gnu_fetch_registers(target_ops*, regcache*, int)' defined but not used [-Wunused-function]
gnu_fetch_registers (struct target_ops *ops,
^~~~~~~~~~~~~~~~~~~
[...]
/usr/bin/ld: i386-gnu-nat.o:(.data.rel+0x0): undefined reference to `vtable for i386_gnu_nat_target'
collect2: error: ld returned 1 exit status
gdb/
* gnu-nat.c (gnu_nat_target::detach): Instead of
'inf_child_maybe_unpush_target (ops)' call 'maybe_unpush_target'.
* gnu-nat.h: #include "inf-child.h".
* i386-gnu-nat.c (gnu_fetch_registers): Rename/move to
'i386_gnu_nat_target::fetch_registers'.
(gnu_store_registers): Rename/move to
'i386_gnu_nat_target::store_registers'.
In file included from ./nm.h:25:0,
from [...]/gdb/defs.h:423,
from [...]/gdb/gdb.c:19:
[...]/gdb/regcache.h:35:46: warning: 'get_thread_regcache' initialized and declared 'extern'
extern struct regcache *get_thread_regcache (thread_info *thread);
^~~~~~~~~~~
[...]/gdb/regcache.h:35:46: error: 'regcache* get_thread_regcache' redeclared as different kind of symbol
[...]
[...]/gdb/gdbarch.h:1203:69: error: 'thread_info' is not a type
extern LONGEST gdbarch_get_syscall_number (struct gdbarch *gdbarch, thread_info *thread);
^~~~~~~~~~~
Fixed with a different (self-contained, more maintainable?) approach compared
to what has been done in commit 7aabaf9d4a
"Create private_thread_info hierarchy", and commit
75cbc781e3 "gdb: For macOS, s/thread_info/struct
thread_info/". We don't want to change all the GDB code to everywhere use
'class thread_info' or 'struct thread_info' instead of plain 'thread_info'.
gdb/
* config/i386/nm-i386gnu.h: Don't "#include" any files.
* gnu-nat.h (mach_thread_info): New function.
* gnu-nat.c (thread_takeover_sc_cmd): Use it.
Using the gdb.ada/call_pn.exp testcase, and running it by hand on
riscv64-elf, we get the following error:
(gdb) call pn(55)
Could not compute alignment of type
The problem occurs because the parameter's type is a TYPE_CODE_RANGE,
and that type code is not handled by riscv_type_alignment. So this patch
fixes the issue by handling TYPE_CODE_RANGE the same way we handle other
integral types.
gdb/ChangeLog:
* riscv-rdep.c (riscv_type_alignment): Handle TYPE_CODE_RANGE.
Tested on riscv64-elf using AdaCore's testsuite.
This is a followup on a recent patch which, among other things
introduced the exit notification of the main thread in order
to be symetrical with the fact that a thread notification was
emitted before signaling its creation.
This patch takes the opposite approach of removing both creation
and exit notifications for that main thread, which is consistent
with what is done on other platforms such as GNU/Linux for instance.
gdb/ChangeLog
* windows-nat.c (windows_add_thread): Add new parameter
"main_thread_p" with default value set to false. Update
function documentation as well as all callers.
(windows_delete_thread): Likewise.
(fake_create_process): Update call to windows_add_thread.
(get_windows_debug_event) <CREATE_THREAD_DEBUG_EVENT>
<CREATE_PROCESS_DEBUG_EVENT>: Likewise.
<EXIT_THREAD_DEBUG_EVENT, EXIT_PROCESS_DEBUG_EVENT>: Update
call to windows_delete_thread.
Tested on x86-windows (MinGW) using AdaCore's testsuite.
gdb/testsuite/ChangeLog
2019-02-12 Weimin Pan <weimin.pan@oracle.com>
PR breakpoints/21870
* gdb.arch/aarch64-dbreg-contents.exp: New file.
* gdb.arch/aarch64-dbreg-contents.c: New file.
Object file paths passed to find_separate_debug_file are always
canonical paths with symbolic links resolved. If a sysroot path
traverses a symbolic link, it will not match the object file paths.
Generate a canonical version of the sysroot directory. If it is
valid, use it instead of gdb_sysroot with child_path to determine if
an object file is under a system root.
gdb/ChangeLog:
* symfile.c (find_separate_debug_file): Use canonical path of
sysroot with child_path instead of gdb_sysroot if it is valid.
This fixes the case where the sysroot happens to end in a trailing
'/'. Note that the path returned from child_path always skips over
the directory separator at the start of the base path, so a separator
must always be explicitly added before the base path.
gdb/ChangeLog:
* symfile.c (find_separate_debug_file): Use child_path to
determine if an object file is under a sysroot.
child_path returns a pointer to the first component in a child path
that comes after a parent path. This does not depend on trying to
stat() the paths since they may describe remote paths but instead
relies on filename parsing. The function requires that the child path
describe a filename that contains at least one component below the
parent path and returns a pointer to the first component.
gdb/ChangeLog:
* Makefile.in (SUBDIR_UNITTESTS_SRCS): Add
unittests/child-path-selftests.c.
* common/pathstuff.c (child_path): New function.
* common/pathstuff.h (child_path): New prototype.
* unittests/child-path-selftests.c: New file.
When an object file is present in a system root, GDB currently looks
for separate debug files under the global debugfile directories. For
example, if the sysroot is set to "/myroot" and hte global debugfile
directory is set to "/usr/lib/debug", GDB will look for a separate
debug file for "/myroot/lib/libc.so.7" in the following paths:
/myroot/lib/libc.so.7.debug
/myroot/lib/.debug/libc.so.7.debug
/usr/lib/debug//myroot/lib/libc.so.7.debug
/usr/lib/debug/lib/libc.so.7.debug
However, some system roots include a full system installation
including a nested global debugfile directory under the sysroot. This
patch adds an additional check to support such systems. In the
example above the additional path searched is:
/myroot/usr/lib/debug/lib/libc.so.7.debug
To try to preserve existing behavior as much as possible, this new
path is searched last for each global debugfile directory.
gdb/ChangeLog:
* symfile.c (find_separate_debug_file): Look for separate debug
files in debug directories under the sysroot.
Note that print_msymbol_info does not (yet?) print data msymbol
using variable_name_style, as otherwise 'info variables'
would show the non debugging symbols in variable name style,
but 'real' variables would be not styled.
2019-02-12 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* symtab.h (struct minimal_symbol data_p): New const method.
(struct minimal_symbol text_p): Likewise.
* symtab.c (output_source_filename): Use file name style
to print file name.
(print_symbol_info): Likewise.
(print_msymbol_info): Use address style to print addresses.
Use function name style to print executable text symbols.
(expand_symtab_containing_pc): Use data_p.
(find_pc_sect_compunit_symtab): Likewise.
Note that ada-typeprint.c print_func_type is called with
types representing functions and is also called to print
a function NAME together with its type. In such a case, the function
name will be printed using function name style.
Similarly, c_print_type_1 is called to print a type, optionally
with the name of an object of this type in the VARSTRING arg.
So, c_print_type_1 uses function name style to print varstring
when the type code indicates that c_print_type_1 TYPE is some
'real code'.
gdb/ChangeLog
2019-02-12 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* ada-typeprint.c (print_func_type): Print function name
style to print function name.
* c-typeprint.c (c_print_type_1): Likewise.
When a node is removed from a splay tree, the splay tree was
not using the function splay_tree_delete_key_fn to release the key.
This was causing a leak, fixed by Tom Tromey.
This patch fixes another key leak, that happens when a key equal to
a key already present is inserted. In such a case, we have to release
the old KEY.
Note that this is based on the assumption that the caller always
allocates a new KEY when doing an insert.
Also, clarify the documentation about when the release functions are
called.
2019-02-11 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* splay-tree.h (splay_tree_delete_key_fn): Update comment.
(splay_tree_delete_value_fn): Likewise.
libiberty/ChangeLog
2019-02-11 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* splay-tree.c (splay_tree_insert): Also release old KEY in case
of insertion of a key equal to an already present key.
(splay_tree_new_typed_alloc): Update comment.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@268793 138bc75d-0d04-0410-961f-82ee72b054a4
Checking the syscall number when stopped on entry/exit relies on checking
the value in register X8.
However, on exit from an execve syscall, the registers will all be cleared.
Given this is only checked on syscall entry/exit, then a cleared register
state either means execve exit or syscall 0 (io_setup) entry with invalid
parameters and an invalid FR and LR, which in reality should never happen.
Use this to detect execve exit.
Move function to allow use of aarch64_sys_execve enum, and use newer
regcache functions.
Fixes gdb.base/catch-syscall.exp on Aarch64.
gdb/ChangeLog:
* aarch64-linux-tdep.c (aarch64_linux_get_syscall_number): Check
for execve.
ommit 3ae729d5a4
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Wed Mar 7 04:18:45 2018 -0800
x86: Rewrite NOP generation for fill and alignment
increased MAX_MEM_FOR_RS_ALIGN_CODE to 4095 which resulted in increase
of assembler time and memory usage by 5 times for inputs with many
.p2align directives, which is typical for LTO output. This patch passes
max_bytes to TC_FRAG_INIT so that MAX_MEM_FOR_RS_ALIGN_CODE can be set
as needed and tracked by backend it so that HANDLE_ALIGN can check the
maximum alignment for each rs_align_code frag. Wall time to assemble
the same cc1plus.s:
before:
423.78user 0.89system 7:05.71elapsed 99%CPU
after:
102.35user 0.27system 1:42.89elapsed 99%CPU
PR gas/24165
* frags.c (frag_var_init): Pass max_chars to TC_FRAG_INIT as
max_bytes.
* config/tc-aarch64.h (TC_FRAG_INIT): Add and pass max_bytes to
aarch64_init_frag.
* /config/tc-arm.h (TC_FRAG_INIT): And and pass max_bytes to
arm_init_frag.
* config/tc-avr.h (TC_FRAG_INIT): And and ignore max_bytes.
* config/tc-ia64.h (TC_FRAG_INIT): Likewise.
* config/tc-mmix.h (TC_FRAG_INIT): Likewise.
* config/tc-nds32.h (TC_FRAG_INIT): Likewise.
* config/tc-ns32k.h (TC_FRAG_INIT): Likewise.
* config/tc-rl78.h (TC_FRAG_INIT): Likewise.
* config/tc-rx.h (TC_FRAG_INIT): Likewise.
* config/tc-score.h (TC_FRAG_INIT): Likewise.
* config/tc-tic54x.h (TC_FRAG_INIT): Likewise.
* config/tc-tic6x.h (TC_FRAG_INIT): Likewise.
* config/tc-xtensa.h (TC_FRAG_INIT): Likewise.
* config/tc-i386.h (MAX_MEM_FOR_RS_ALIGN_CODE): Set to
(alignment ? ((1 << alignment) - 1) : 1)
(i386_tc_frag_data): Add max_bytes.
(TC_FRAG_INIT): Add and track max_bytes.
(HANDLE_ALIGN): Replace MAX_MEM_FOR_RS_ALIGN_CODE with
fragP->tc_frag_data.max_bytes.
* doc/internals.texi: Update TC_FRAG_TYPE with max_bytes.
Valgrind detects a bunch of leaks in several tests, such as:
==22905== 40 (24 direct, 16 indirect) bytes in 1 blocks are definitely lost in loss record 531 of 3,268
==22905== at 0x4C2C4CC: operator new(unsigned long) (vg_replace_malloc.c:344)
==22905== by 0x5893AD: get_type_stack() (parse.c:1509)
==22905== by 0x3F4EAD: c_yyparse() (c-exp.y:1223)
==22905== by 0x3F71BC: c_parse(parser_state*) (c-exp.y:3308)
==22905== by 0x588CEA: parse_exp_in_context_1(char const**, unsigned long, block const*, int, int, int*) [clone .constprop.89] (parse.c:1205)
==22905== by 0x588FA1: parse_exp_in_context (parse.c:1108)
==22905== by 0x588FA1: parse_exp_1 (parse.c:1099)
==22905== by 0x588FA1: parse_expression(char const*) (parse.c:1247)
...
==22395== 456 (168 direct, 288 indirect) bytes in 7 blocks are definitely lost in loss record 2,658 of 2,978
==22395== at 0x4C2C4CC: operator new(unsigned long) (vg_replace_malloc.c:344)
==22395== by 0x5893AD: get_type_stack() (parse.c:1509)
==22395== by 0x3F4ECF: c_yyparse() (c-exp.y:1230)
==22395== by 0x3F71BC: c_parse(parser_state*) (c-exp.y:3308)
==22395== by 0x588CEA: parse_exp_in_context_1(char const**, unsigned long, block const*, int, int, int*) [clone .constprop.89] (parse.c:1205)
==22395== by 0x588FA1: parse_exp_in_context (parse.c:1108)
==22395== by 0x588FA1: parse_exp_1 (parse.c:1099)
==22395== by 0x588FA1: parse_expression(char const*) (parse.c:1247)
==22395== by 0x67BB9D: whatis_exp(char const*, int) (typeprint.c:515)
...
==22395== VALGRIND_GDB_ERROR_BEGIN
==22395== 144 (24 direct, 120 indirect) bytes in 1 blocks are definitely lost in loss record 1,016 of 2,978
==22395== at 0x4C2C4CC: operator new(unsigned long) (vg_replace_malloc.c:344)
==22395== by 0x5893AD: get_type_stack() (parse.c:1509)
==22395== by 0x3F4E8A: c_yyparse() (c-exp.y:1217)
==22395== by 0x3F71BC: c_parse(parser_state*) (c-exp.y:3308)
==22395== by 0x588CEA: parse_exp_in_context_1(char const**, unsigned long, block const*, int, int, int*) [clone .constprop.89] (parse.c:1205)
==22395== by 0x588FA1: parse_exp_in_context (parse.c:1108)
==22395== by 0x588FA1: parse_exp_1 (parse.c:1099)
==22395== by 0x588FA1: parse_expression(char const*) (parse.c:1247)
==22395== by 0x67BB9D: whatis_exp(char const*, int) (typeprint.c:515)
...
Fix these by storing the allocated type_stack in the cpstate->type_stacks
vector.
Tested on debian/amd64, natively and under valgrind.
gdb/ChangeLog
2019-02-10 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* c-exp.y (direct_abs_decl): Use emplace_back to record the
type_stack.
Consider the following variable, which is a string whose value
is not known at compile time, because it is the return value
from a function call (Get_Name):
A : String := Get_Name;
If one tries to create a varobj for that variable, everything works
as expected:
| (gdb) -var-create a * a
| ^done,name="a",numchild="19",value="[19] \"Some kind of string\"",type="<ref> array (1 .. 19) of character",thread-id="1",has_more="0"
However, try then to request an update, regardless of whether the string
has changed or not, and we get a crash:
| -var-update a
| ~"/[...]/gdb/varobj.c:1379: internal-error: bool install_new_value(varobj*, value*, bool): Assertion `!value_lazy (var->value.get ())' failed.\nA problem internal to GDB has been detected,\nfurther debugging may prove unreliable.\nQuit this debugging session? (y or n) "
When the varobj gets created (-var-create), the expression is evaluated
and transformed into a value. The debugging information describes our
variables as a reference to an array of characters, so our value has
the corresponding type. We then call varobj.c::install_new_value
to store that value inside our varobj, and we see that this function
pretty starts by determining weither our varobj is changeable, via:
| changeable = varobj_value_is_changeable_p (var);
(where 'var' is the varobj we are building, and where the function
varobj_value_is_changeable_p simply dispatches to the Ada version
of this routine: ada_value_is_changeable_p).
At this point, the varobj doesn't have a value, yet, but it does
have a type which was provided by varobj_create a little bit before
install_new_value was called. So ada_value_is_changeable_p uses that
to determine whether or not our type is changeable.
Since our type is a reference to an array, and that the value of
such objects is displayed as if there weren't a reference, it means
that our object is changeable -- in other words, if an element of
the string changes, then the "value" field of the varobj will change
accordingly. But unfortunately, ada_value_is_changeable_p mistakenly
returns false, because it is missing the handling of reference types.
As a consequence of this, install_new_value doesn't feel it is
necessary to fetch the value's contents, as explained by the following
comment inside that function:
/* The new value might be lazy. If the type is changeable,
that is we'll be comparing values of this type, fetch the
value now. Otherwise, on the next update the old value
will be lazy, which means we've lost that old value. */
This means that a lazy value gets installed inside our varobj
as a result of the mistake in ada_value_is_changeable_p.
Another important detail is that, after determining whether
our varobj is changeable or not, it then purposefully removes
the reference layer from our value:
/* We are not interested in the address of references, and given
that in C++ a reference is not rebindable, it cannot
meaningfully change. So, get hold of the real value. */
if (value)
value = coerce_ref (value);
The consequence of those two facts on shows up only later, when
the user requests an update (-var-update). When doing so, GDB
evaluates the expression again into a value which is once more
a reference to a string, and then calls install_new_value again
to install the new value and report any changes. This time around,
the call to...
| changeable = varobj_value_is_changeable_p (var);
... now gets a varobj which has a value, and one which had the reference
layer removed! So, this time, we classify the varobj correctly, and
say it is changeable. And because it is changeable, we then go into
the section of code in install_new_value which checks for changes,
where we need the varobj's value to not be lazy, as explained by
the comment we quoted above. That's what the assertion was about.
This patch fixes the issues by teaching ada_value_is_changeable_p
to ignore reference layers when evaluating a given varobj's type.
gdb/ChangeLog:
* ada-varobj.c (ada_value_is_changeable_p): Add handling of
TYPE_CODE_REF types.
gdb/testsuite/ChangeLog:
* gdb.ada/mi_ref_changeable: New testcase.
Prior to this patch, this testcase reports 2 unresolved tests
(due to GDB hitting the internal error). With this patch, all
tests in this testcase pass.
Tested on x86_64-linux, no regression.
Recent gcc commit b4371b277f1e ("[ARC] Enable init_array support")
inhibits DT_"INIT,FINI} in favor of DT_{INIT,FINI}ARRAY.
Even prior to that, it seems ARC port is the only one with this
special DT_INIT/FINI handling in linker emulation. Removing it
doesn't seem to change any uClibc/glibc testsuite results,
so this can RIP anyways.
bfd/
2019-02-01 Vineet Gupta <vgupta@synopsys.com>
* elf32-arc.c: Delete init_str, fini_str
ld/
2019-02-01 Vineet Gupta <vgupta@synopsys.com>
* emultempl/arclinux.em : Delete special INIT/FINI handling.
Joel Brobecker
GDB Administrator
Tom Tromey
Saagar Jha
Weimin Pan
Thomas Schwinge
David Michael
KONRAD Frederic
Simon Marchi
John Baldwin
Philippe Waroquiers
Nick Clifton
tromey
Alan Hayward
H.J. Lu
Claudiu Zissulescu