Jan Beulich [Fri, 27 Dec 2019 08:39:17 +0000 (09:39 +0100)]
x86-64: correct / adjust prefix emission
First and foremost REX must come last. Next JumpInterSegment branches
can't possibly have a REX prefix, as they're consistently CpuNo64. And
finally make BND prefix handling in output_branch() consistent with that
of other prefixes in the same function, and make its placement among
prefixes consistent with output_jump() (which, oddly enough, still isn't
the supposedly canonical order specified by the *_PREFIX definitions).
Jan Beulich [Fri, 27 Dec 2019 08:38:34 +0000 (09:38 +0100)]
x86-64: fix Intel64 handling of branch with data16 prefix
The expectation of x86-64-branch-3 for "call" / "jmp" with an obvious
direct destination to translate to an indirect _far_ branch is plain
wrong. The operand size prefix should have no effect at all on the
interpretation of the operand. The main underlying issue here is that
the Intel64 templates of the direct branches don't include Disp16, yet
various assumptions exist that it would always be there when there's
also Disp32/Disp32S, toggled by the operand size prefix (which is
being ignored by direct branches in Intel64 mode).
Along these lines it was also wrong to base the displacement width
decision solely on the operand size prefix: REX.W cancels this effect
and hence needs taking into consideration, too.
A disassembler change is needed here as well: XBEGIN was wrongly treated
the same as direct CALL/JMP, which isn't the case - the operand size
prefix does affect displacement size there, it's merely ignored when it
comes to updating [ER]IP.
Jan Beulich [Fri, 27 Dec 2019 08:22:03 +0000 (09:22 +0100)]
x86: consolidate Disp<NN> handling a little
In memory operand addressing, which forms of displacement are permitted
besides Disp8 is pretty clearly limited
- outside of 64-bit mode, Disp16 or Disp32 only, depending on address
size (MPX being special in not allowing Disp16),
- in 64-bit mode, Disp32s or Disp64 without address size override, and
solely Disp32 with one.
Adjust assembler and i386-gen to match this, observing that templates
already get adjusted before trying to match them against input depending
on the presence of an address size prefix.
This adjustment logic gets extended to all cases, as certain DispNN
values should also be dropped when there's no such prefix. In fact
behavior of the assembler, perhaps besides the exact diagnostics wording,
should not differ between there being templates applicable to 64-bit and
non-64-bit at the same time, or there being fully separate sets of
templates, with their DispNN settings already reduced accordingly.
This adjustment logic further gets guarded such that there wouldn't be
and Disp<N> conversion based on address size prefix when this prefix
doesn't control the width of the displacement (on branches other than
absolute ones).
These adjustments then also allow folding two MOV templates, which had
been split between 64-bit and non-64-bits variants so far.
Once in this area also
- drop the bogus DispNN from JumpByte templates, leaving just the
correct Disp8 there (compensated by i386_finalize_displacement()
now setting Disp8 on their operands),
- add the missing Disp32S to XBEGIN.
Note that the changes make it necessary to temporarily mark a test as
XFAIL; this will get taken care of by a subsequent patch. The failing
parts are entirely bogus and will get replaced.
This also renames it to make it clearer that this is not a cheap
function (to compute_and_set_names). Also renames name to m_name
to make the implementation of the renamed function more readable.
Most of the places that access sym->m_name directly were also changed
to call linkage_name () instead, to make it clearer which name they
are accessing.
gdb/ChangeLog:
2019-12-26 Christian Biesinger <cbiesinger@google.com>
Alan Modra [Tue, 26 Nov 2019 06:19:44 +0000 (16:49 +1030)]
Avoid ubsan bug complaining about &p->field
I reckon it's quite OK to write &p->field in C when p might be NULL,
and lots of old C programmers probably agree with me. However, ubsan
disagrees and so do some people I respect. I suspect C++ influence is
to blame for the ubsan behaviour. See
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92634. So far no one has
educated me as to why I'm wrong to claim that there isn't anything in
the C standard to say that p->field is always (*p).field. Note 79
doesn't quite do that because it doesn't cover null pointers. If
there was such an equivalence then you could claim &p->field has a
null pointer reference when p is NULL, even though no C compiler would
ever dereference p.
Anyway, to silence ubsan I'm going to apply the following though I
prefer to avoid casts when possible. And I'm using (void *)
deliberately because this is C, not C++!
Alan Modra [Thu, 26 Dec 2019 01:55:31 +0000 (12:25 +1030)]
asan: som: heap-buffer-overflow
Triggered by overflow of size calulation resulting in a too small
buffer. The testcase found one of the som_bfd_count_ar_symbols
problems.
* som.c (setup_sections): Don't overflow space_strings_size. Use
bfd_malloc2 to catch overflow of size calculation.
(som_prep_for_fixups): Use bfd_zalloc2 to catch overflow of size
calculation.
(som_build_and_write_symbol_table): Similarly use bfd_zmalloc2.
(som_slurp_symbol_table): Similarly use bfd_zmalloc2, bfd_malloc2,
and bfd_zalloc2.
(bfd_som_attach_aux_hdr): Use size_t vars for string length.
(som_bfd_count_ar_symbols): Use bfd_malloc2 to catch overflow of
size calculation. Use size_t vars for length and catch overflow.
(som_slurp_armap): Use bfd_alloc2 to catch overflow of size
calculation.
(som_bfd_ar_write_symbol_stuff): Similarly use bfd_zmalloc2 and
bfd_malloc2. Perform size calculations in bfd_size_type.
Eli Zaretskii [Mon, 23 Dec 2019 14:28:32 +0000 (16:28 +0200)]
Fix compilation of Readline on mingw.org's MinGW
readline/ChangeLog
2019-12-23 Eli Zaretskii <eliz@gnu.org>
* posixstat.h (S_IRGRP, S_IWGRP, S_IXGRP, S_IROTH, S_IWOTH)
(S_IXOTH, S_IRWXG, S_IRWXO): Define if undefined, even if S_IRWXU
is defined, because non-Posix systems may defined only the user
bits.
Alan Modra [Mon, 23 Dec 2019 07:36:23 +0000 (18:06 +1030)]
asan: vms-alpha: heap-buffer-overflow
Two buffer overflows, and some over restrictive length checks.
* vms-alpha.c (add_symbol): Add "max" parameter. Error on string
length larger than max.
(_bfd_vms_slurp_egsd): Ensure record is at least large enough to
read string length byte, error if not. Pass size to add_symbol.
(_bfd_vms_slurp_etir): Don't read past end of buffer when reading
type and length. Allow read of last byte in buffer.
Alan Modra [Mon, 23 Dec 2019 07:32:44 +0000 (18:02 +1030)]
ubsan: d30v: left shift cannot be represented in type 'long long'
* d30v-dis.c (extract_value): Make num param a uint64_t, constify
oper. Use unsigned vars.
(print_insn): Make num var uint64_t. Constify oper and remove now
unnecessary casts on extract_value calls.
(print_insn_d30v): Use unsigned vars. Adjust printf formats.
Alan Modra [Mon, 23 Dec 2019 07:31:34 +0000 (18:01 +1030)]
Revise sleb128 and uleb128 reader
This patch catches and reports errors when reading leb128 values,
addressing a FIXME in read_leb128.
* dwarf.h (read_leb128): Update prototype.
(report_leb_status): New inline function.
(SKIP_ULEB, SKIP_SLEB, READ_ULEB, READ_SLEB): Define.
* dwarf.c: Use above macros throughout file. Formatting.
(read_leb128): Reorder params. Add status return param.
Don't stop reading until finding terminator or end of data.
Detect loss of significant bits. Sign extend only on
terminating byte.
(read_sleb128, read_uleb128): Delete functions.
(SKIP_ULEB, SKIP_SLEB, READ_ULEB, READ_SLEB): Delete macros.
(read_and_print_leb128): Rewrite.
(process_extended_line_op): Return a size_t. Use size_t vars.
Adjust to suit new macros. Add proper name size to "data" when
processing DW_LNE_define_file.
(process_abbrev_section): Adjust to suit new macros.
(decode_location_expression, skip_attr_bytes): Likewise.
(get_type_signedness): Likewise.
(read_and_display_attr_value): Likewise. Consolidate block code.
(process_debug_info): Adjust to suit new macros.
(display_formatted_table, display_debug_lines_raw): Likewise.
(display_debug_lines_decoded): Likewise. Properly check for end
of DW_LNS_extended_op.
(display_debug_macinfo): Adjust to suit new macros.
(get_line_filename_and_dirname, display_debug_macro): Likewise.
(display_view_pair_list): Likewise. Don't back off when hitting
end of data.
(display_loc_list): Adjust to suit new macros.
(display_loclists_list, display_loc_list_dwo): Likewise.
(display_debug_rnglists_list, read_cie): Likewise.
(display_debug_frames): Likewise.
* readelf.c: Use new ULEB macros throughout file.
(read_uleb128): Delete.
(decode_arm_unwind_bytecode): Use read_leb128.
(decode_tic6x_unwind_bytecode): Likewise.
(display_tag_value): Adjust to suit new macros.
(display_arc_attribute, display_arm_attribute): Likewise.
(display_gnu_attribute, display_power_gnu_attribute): Likewise.
(display_s390_gnu_attribute, display_sparc_gnu_attribute): Likewise.
(display_mips_gnu_attribute, display_tic6x_attribute): Likewise.
(display_msp430x_attribute, display_msp430_gnu_attribute): Likewise.
(display_riscv_attribute, process_attributes): Likewise.
George Barrett [Sun, 15 Dec 2019 00:12:09 +0000 (11:12 +1100)]
Fix disabling of solib probes when LD_AUDITing
The SVR4 solib event handler determines whether an event is related to a
non-base link namespace by comparing the event's debug struct address
to the debug struct address of the initial program image. However, this
can fail when using LD_AUDIT as audit libraries are loaded before the
loader has initialised the initial program image's debug struct. When
the event handler fails to find the debug struct, the probe-based
debugger interface is disabled and a warning is flagged to the user.
This commit adds a fallback test to help determine whether an event is
for a foreign link namespace when the debug struct isn't available.
gdb/ChangeLog:
2019-12-15 George Barrett <bob@bob131.so>
* solib-svr4.c (svr4_handle_solib_event): Add fallback link
namespace test for when the debug struct isn't available.
Weimin Pan [Fri, 20 Dec 2019 22:30:17 +0000 (22:30 +0000)]
Address Tom Tromey's comments on the CTF reader.
* Use the type-safe registry for ctf_file_key;
* Drop "typedef" when defining "struct ctf_context";
* Use ANOFFSET with SECT_OFF_TEXT to get the text base address;
gdb/ChangeLog
2019-12-20 Weimin Pan <weimin.pan@oracle.com>
* ctfread.c (ctf_file_key): Change type to objfile_key.
(struct ctf_context): Remove typedef.
(get_objfile_text_range): Use ANOFFSET to get text base.
Use enum bitfield for the calling_convention attribute of a subroutine
This is a refactoring. Instead of a plain unsigned value, use an enum
bitfield.
gdb/ChangeLog:
2019-12-20 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
* dwarf2read.c (is_valid_DW_AT_calling_convention_for_subroutine):
New function.
(read_subroutine_type): Validate the parsed
DW_AT_calling_convention value before assigning it to a
subroutine's calling_convention attribute.
* gdbtypes.h (struct func_type) <calling_convention>: Use
an enum bitfield as its type, instead of plain unsigned.
testsuite, cp: increase the coverage of testing pass-by-ref arguments
Extend testcases for GDB's infcall of call-by-value functions that
take aggregate values as parameters. In particular, existing test has
been substantially extended with class definitions whose definitions
of copy constructor, destructor, and move constructor functions are a
combination of
(1) explicitly defined by the user,
(2) defaulted inside the class declaration,
(3) defaulted outside the class declaration,
(4) deleted
(5) not defined in the source.
For each combination, a small and a large class is generated as well
as a derived class and a container class. Additionally, the following
manually-written cases are provided:
- a dynamic class (i.e. class with a virtual method)
- classes that contain an array field
- a class whose copy ctor is inlined
- a class whose destructor is deleted
- classes with multiple copy and/or move ctors
Test cases check whether GDB makes the right decision to pass an
object by value or implicitly by reference, whether really a copy of
the argument is passed, and whether the copy constructor and
destructor of the clone of the argument are invoked properly.
The input program pass-by-ref.cc is generated in the test's output
directory. The input program pass-by-ref-2.cc is manually-written.
Tests have been verified on the X86_64 architecture with
GCC 7.4.0, 8.2.0, and 9.2.1.
gdb/testsuite/ChangeLog:
2019-12-20 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
* gdb.cp/pass-by-ref.cc: Delete. Generated in the output
directory instead.
* gdb.cp/pass-by-ref.exp: Extend with more cases.
* gdb.cp/pass-by-ref-2.cc: New file.
* gdb.cp/pass-by-ref-2.exp: New file.
If an aggregate argument is implicitly pass-by-reference, allocate a
temporary object on the stack, initialize it via the copy constructor
(if exists) or trivially by memcpy'ing. Pass the reference of the
temporary to the callee function. After the callee returns, invoke
the destructor of the temporary.
gdb/ChangeLog:
2019-12-20 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
PR gdb/25054
* infcall.c (call_function_by_hand_dummy): Update the argument-
passing section for call-by-value parameters.
(struct destructor_info): New struct.
(call_destructors): New auxiliary function.
infcall, c++: collect more pass-by-reference information
Walk through a given type to collect information about whether the
type is copy constructible, destructible, trivially copyable,
trivially copy constructible, trivially destructible. The previous
algorithm returned only a boolean result about whether the type is
trivially copyable. This patch computes more info. Additionally, it
utilizes DWARF attributes that were previously not taken into account;
namely, DW_AT_deleted, DW_AT_defaulted, and DW_AT_calling_convention.
gdb/ChangeLog:
2019-12-20 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
* gnu-v3-abi.c (enum definition_style): New enum type.
(get_def_style): New function.
(is_user_provided_def): New function.
(is_implicit_def): New function.
(is_copy_or_move_constructor_type): New function.
(is_copy_constructor_type): New function.
(is_move_constructor_type): New function.
(gnuv3_pass_by_reference): Collect language_pass_by_ref_info
for a given type.
infcall, c++: allow more info to be computed for pass-by-reference values
In C++, call-by-value arguments that cannot be trivially copied are
implicitly passed by reference. When making an infcall, GDB needs to
find out if an argument is pass-by-reference or not, so that the
correct semantics can be followed. This patch enriches the
information computed by the language ops for pass-by-reference
arguments. Instead of a plain binary result, the computed information
now includes whether the argument is
gdb: recognize new DWARF attributes: defaulted, deleted, calling conv.
Extend GDB's internal representation of types to include the
DW_AT_calling_convention, DW_AT_defaulted, and DW_AT_deleted attributes
that were introduced in DWARF5.
These attributes will be helpful in a future patch about infcall'ing
functions with call-by-value parameters. GDB will use the attributes
to decide whether the type of a call-by-value parameter is implicitly
pass-by-reference.
gdb/ChangeLog:
2019-12-20 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
* dwarf2read.c (dwarf2_add_member_fn): Read the DW_AT_defaulted
and DW_AT_deleted attributes of a function.
(read_structure_type): Read the DW_AT_calling_convention attribute
of a type.
(is_valid_DW_AT_defaulted): New function.
(is_valid_DW_AT_calling_convention_for_type): New function.
* gdbtypes.h: Include dwarf2.h.
(struct fn_field)<defaulted>: New field to store the
DW_AT_defaulted attribute.
(struct fn_field)<is_deleted>: New field to store the
DW_AT_deleted attribute.
(struct cplus_struct_type)<calling_convention>: New field to store
the DW_AT_calling_convention attribute.
(TYPE_FN_FIELD_DEFAULTED): New macro.
(TYPE_FN_FIELD_DELETED): New macro.
(TYPE_CPLUS_CALLING_CONVENTION): New macro.
* gdbtypes.c (dump_fn_fieldlists): Update for the changes made
to the .h file.
(print_cplus_stuff): Likewise.
Tom Tromey [Wed, 13 Nov 2019 23:47:58 +0000 (16:47 -0700)]
Don't call tui_show_source from tui_ui_out
This removes the call to tui_show_source from tui_ui_out. This always
seemed like a hack, and now that the TUI is using the proper
observers, it seems not to be needed.
The rest of the logic remains, unfortunately, because it is needed to
suppress some gdb output in the TUI case. We could probably find a
nicer way to do this (maybe a ui_out_flag), but I haven't attempted
this.
This was the last caller of tui_show_source, so this is removed as
well.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 23:30:52 +0000 (16:30 -0700)]
Fix "list" command in the TUI
PR tui/18932 notes that "list" no longer works in the TUI. At some
point in the past, it switched the TUI source window to show the
specified source; but now this source briefly flashes before the TUI
reverts to showing the current stack frame's source.
This patch fixes this bug by introducing a new observer that notices
when the user selected context has changed. Then, the existing
before-prompt observer is updated to request the correct update:
either one based on the current stack frame, or one based on the
user's source symtab_and_line.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
PR tui/18932:
* tui/tui-hooks.c (tui_refresh_frame_and_register_information):
Rename parameters. Handle the not-from-stack-frame case.
(from_stack, from_source_symtab): New globals.
(tui_before_prompt, tui_normal_stop): Update.
(tui_context_changed, tui_symtab_changed): New functions.
(tui_attach_detach_observers): Attach new observers.
gdb/testsuite/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 22:54:47 +0000 (15:54 -0700)]
Don't call set_current_source_symtab_and_line from TUI
update_source_window_as_is calls set_current_source_symtab_and_line,
but I don't think there is any reason it should be doing this. This
patch removes the call.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 22:41:08 +0000 (15:41 -0700)]
Change set_locator_info to take a symtab_and_line
This changes set_locator_info to take a symtab_and_line, rather than
the individual components.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
* tui/tui-stack.h (struct tui_locator_window) <set_locator_info>:
Take a symtab_and_line.
* tui/tui-stack.c (tui_locator_window::set_locator_info): Take a
symtab_and_line.
(tui_show_frame_info): Update.
Tom Tromey [Wed, 13 Nov 2019 07:16:39 +0000 (00:16 -0700)]
Remove a call to update_exec_info
tui_show_frame_info calls update_exec_info after calling
erase_source_content, but there's no need to do this, as
erase_source_content already clears the exec info.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 01:04:01 +0000 (18:04 -0700)]
Simplify tui_update_source_windows_with_line
This changes tui_update_source_windows_with_line to take a
symtab_and_line, rather than separate parameters, and then updates the
caller.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
* tui/tui.c (tui_show_source): Update.
* tui/tui-winsource.h (tui_update_source_windows_with_line): Update.
* tui/tui-winsource.c (tui_update_source_windows_with_line): Take
a symtab_symbol_info, not a separate symtab and line. Simplify.
Tom Tromey [Wed, 13 Nov 2019 00:59:06 +0000 (17:59 -0700)]
Simplify tui_update_source_windows_with_addr
After the previous changes, tui_update_source_windows_with_addr simply
updates each source-like window separately, passing the same data to
each. So, it can be simplified by using a loop instead.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 00:56:42 +0000 (17:56 -0700)]
Use symtab_and_line when updating TUI windows
This changes a few TUI source window methods to take a symtab_and_line
rather than separate symtab and tui_line_or_address parameters. A
symtab_and_line already incorporates the same information, so this
seemed simpler. Also, it helps avoid the problem that the source and
disassembly windows need different information -- both forms are
present in the SAL.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
* tui/tui-winsource.h (struct tui_source_window_base)
<set_contents, update_source_window_as_is, update_source_window>:
Take a sal, not a separate symtab and tui_line_or_address.
* tui/tui-winsource.c (tui_source_window_base::update_source_window)
(tui_source_window_base::update_source_window_as_is): Take a sal,
not a separate symtab and tui_line_or_address.
(tui_update_source_windows_with_addr)
(tui_update_source_windows_with_line)
(tui_source_window_base::rerender)
(tui_source_window_base::refill): Update.
* tui/tui-source.h (struct tui_source_window) <set_contents>: Take
a sal, not a separate symtab and tui_line_or_address.
* tui/tui-source.c (tui_source_window::set_contents): Take a sal,
not a separate symtab and tui_line_or_address.
(tui_source_window::maybe_update): Update.
* tui/tui-disasm.h (struct tui_disasm_window) <set_contents>: Take
a sal, not a separate symtab and tui_line_or_address.
* tui/tui-disasm.c (tui_disasm_window::set_contents): Take a sal,
not a separate symtab and tui_line_or_address.
(tui_disasm_window::do_scroll_vertical)
(tui_disasm_window::maybe_update): Update.
Tom Tromey [Wed, 13 Nov 2019 00:40:33 +0000 (17:40 -0700)]
Use start_line_or_addr in TUI windows
A few spots in the TUI source and disassembly windows referred to
content[0], where start_line_or_addr is equivalent. This patch makes
this substitution.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
* tui/tui-winsource.c (tui_source_window_base::refill): Use
start_line_or_addr.
* tui/tui-source.c (tui_source_window::do_scroll_vertical): Use
start_line_or_addr.
* tui/tui-disasm.c (tui_disasm_window::do_scroll_vertical): Use
start_line_or_addr.
Tom Tromey [Wed, 13 Nov 2019 00:26:50 +0000 (17:26 -0700)]
Change tui_source_window_base::set_contents to return bool
This changes tui_source_window_base::set_contents to return bool,
rather than tui_status. It also changes one implementation of
set_contents to use early returns rather than a variable, which IMO
makes it easier to follow.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 00:19:59 +0000 (17:19 -0700)]
Remove tui_show_disassem
tui_show_disassem is just a wrapper for the update_source_window
method, and it only has a single caller. This removes the function
and inlines the logic into that caller.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 13 Nov 2019 00:15:12 +0000 (17:15 -0700)]
Remove some unnecessary focus switches
A couple of lower-level utility functions can change the TUI focus.
This seems incorrect to me -- focus switches should only be done
either by explicit user request, or ass a side effect of changing the
layout.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
tui_source_window_base::maybe_update takes a symtab_and_line, plus a
separate line number and PC. Because a symtab_and_line already holds
a line number and a PC, it is possible to remove these extra
parameters.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
* tui/tui-winsource.h (struct tui_source_window_base)
<maybe_update>: Remove line_no and addr parameters.
* tui/tui-stack.c (tui_show_frame_info): Set PC on sal. Update.
* tui/tui-source.h (struct tui_source_window) <maybe_update>:
Update.
* tui/tui-source.c (tui_source_window::maybe_update): Remove
line_no and addr parameters.
* tui/tui-disasm.h (struct tui_disasm_window) <maybe_update>:
Update.
* tui/tui-disasm.c (tui_disasm_window::maybe_update): Remove
line_no and addr parameters.
Tom Tromey [Wed, 13 Nov 2019 00:02:49 +0000 (17:02 -0700)]
Remove some TUI asserts
This removes a few asserts from the TUI. These asserts aren't useful,
because they simply check an invariant that's already ensured by the
type system.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 6 Nov 2019 23:09:28 +0000 (16:09 -0700)]
Remove tui_show_disassem_and_update_source
tui_show_disassem_and_update_source only has a single caller. This
patch simplifies that caller, by having it call tui_show_disassem, and
then removes tui_show_disassem_and_update_source.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Tom Tromey [Wed, 20 Nov 2019 23:39:44 +0000 (16:39 -0700)]
Make isearch change readline prompt in TUI
PR tui/23619 points out that isearch changes the prompt in the CLI gdb
(and in Bash) -- but not in the TUI. This turns out to be easily
fixed by removing tui_rl_saved_prompt and instead using the prompt
that readline computes.
This is stored in rl_display_prompt, which according to git was added
in readline 6.2.
gdb/ChangeLog
2019-12-20 Tom Tromey <tom@tromey.com>
Alan Modra [Fri, 20 Dec 2019 07:26:35 +0000 (17:56 +1030)]
PR25281, sh disassembler abort
PR 25281
* sh-dis.c (print_insn_ddt): Properly check validity of MOVX_NOPY
and MOVY_NOPX insns. For invalid cases include 0xf000 in the word
printed. Print .word in more cases.
Bernd Edlinger [Fri, 13 Dec 2019 16:21:21 +0000 (17:21 +0100)]
Fix build with GNU Make 3.81
GNU Make 3.81 is apparently confused when the same
source file is processed by a pattern rule and an
explicit rule at the same time with different output file.
The pattern %.o: ../%.c and alloc-ipa.o: ../alloc.c
both have the source ../alloc.c but two independent
object files alloc.o and alloc-ipa.o, so
while building gdbserver I see the following message:
make[4]: Circular alloc-ipa.o <- ../alloc.c dependency dropped.
CXX alloc-ipa.o
g++: warning: '-x c++' after last input file has no effect
g++: fatal error: no input files
compilation terminated.
In the make debug output I see the pattern is first correct:
So indeed now $< is empty, and the build step fails.
This happens only when alloc.o needs to be built, when alloc.o
was already built, the build succeeds, but it takes often
several attempts until the build succeeds.
By rewriting the alloc-ipa.c: ../alloc.c rule into a pattern
rule, the problem goes away.
While already at it, this patch removes also the
$(WARN_CFLAGS_NO_FORMAT) from the build rule, which is just a
copy/paste thing that is not necessary for alloc.c at all.
Make the literal argument to pow a double, not an integer
Since pow takes doubles, pass 2.0 instead of 2 to pow ().
Conveniently, this fixes the ambiguous call to pow on Solaris 11
with gcc 5.5 (gcc211 on the compile farm), which has a "using std::pow"
directive in a system header, which brings in float/double/long double
overloads. Fixes the build on Solaris with enable-targets=all.
gdb/ChangeLog:
2019-12-19 Christian Biesinger <cbiesinger@google.com>
* score-tdep.c (score7_analyze_prologue): Pass 2.0 instead of
2 to pow ().
Cast the log10 argument to double to disambiguate it
On Solaris 11 with gcc 5.5.0 (gcc211 on the compile farm), math.h has a
using std::log10; directive. This is unfortunate because std::log10 has
overloads for float/double/long double. To disambiguate this call,
cast the argument to double to fix the build.
gdb/ChangeLog:
2019-12-19 Christian Biesinger <cbiesinger@google.com>
* tui/tui-source.c (tui_source_window::set_contents): Cast argument of
log10 to double to fix Solaris 11 with gcc 5.5.
Rename "sun" variable to avoid conflicts on Solaris
A Solaris system header has a #define for "sun". This renames
that variable to avoid the conflict, fixing a build error with
--enable-targets=all on Solaris.
gdb/ChangeLog:
2019-12-19 Christian Biesinger <cbiesinger@google.com>
* fbsd-tdep.c (fbsd_info_proc_files_entry): Rename local var
"sun" to "saddr_un".
Tom Tromey [Wed, 18 Dec 2019 15:50:57 +0000 (08:50 -0700)]
Add install-strip to sim/
PR build/24572 notes that "make install-strip" fails. For me, it
works in every directory except "sim", so this patch adds
install-strip targets to the Makefiles that appear there.
sim/ChangeLog
2019-12-19 Tom Tromey <tromey@adacore.com>
PR build/24572:
* Makefile.in (install-strip): New target.
sim/common/ChangeLog
2019-12-19 Tom Tromey <tromey@adacore.com>
PR build/24572:
* Makefile.in (install-strip): New target.
sim/igen/ChangeLog
2019-12-19 Tom Tromey <tromey@adacore.com>
PR build/24572:
* Makefile.in (install-strip): New target.
sim/ppc/ChangeLog
2019-12-19 Tom Tromey <tromey@adacore.com>
PR build/24572:
* Makefile.in (install-strip): New target.
sim/testsuite/ChangeLog
2019-12-19 Tom Tromey <tromey@adacore.com>
PR build/24572:
* Makefile.in (install-strip): New target.
Tom Tromey [Tue, 10 Dec 2019 18:44:36 +0000 (11:44 -0700)]
Handle CRLF when reading XML on Windows
xml-support.c uses FOPEN_RT, but then reads the entire contents of the
file and verifies that the number of bytes read matches the length.
This can fail on Windows, where the read will translate line
terminators.
This patch fixes the bug by changing xml-support.c to use FOPEN_RB.
This works because expat correctly handles \r\n line terminators.
gdb/ChangeLog
2019-12-11 Tom Tromey <tromey@adacore.com>
* xml-support.c (xml_fetch_content_from_file): Use FOPEN_RB.
gdb/testsuite/ChangeLog
2019-12-11 Tom Tromey <tromey@adacore.com>
* coff-rs6000.c (_bfd_xcoff_slurp_armap): Don't overflow when
checking symbol count against section size. Guard against strlen
running off end of buffer by allocating one more byte and zeroing.
* coff64-rs6000.c (xcoff64_slurp_armap): Likewise.
Alan Modra [Thu, 19 Dec 2019 05:08:39 +0000 (15:38 +1030)]
vax decoding of indexed addressing mode
This patch prevents print_insn_mode recursing into another index mode
byte, which if repeated enough times will overflow private.the_buffer
and scribble over other memory.
* vax-dis.c (print_insn_mode): Stop index mode recursion.
Tom Tromey [Wed, 18 Dec 2019 18:45:51 +0000 (11:45 -0700)]
Fix pthread_setname_np build error
My earlier patch to fix the pthread_setname_np build error on macOS
was incorrect. While the macOS man page claims that
pthread_setname_np returns void, in <pthread.h> it is actually
declared returning "int". I knew this earlier, but must have made
some mistake when preparing the patch for submission (perhaps when
removing the templates?).
This patch re-fixes the bug. I'm also applying it to the 9.1 branch.
Tested by building on macOS High Sierra.
gdb/ChangeLog
2019-12-18 Tom Tromey <tromey@adacore.com>
PR build/25268:
* gdbsupport/thread-pool.c (set_thread_name): Expect "int" return
type on macOS. Add comment.
Simon Marchi [Wed, 18 Dec 2019 18:27:18 +0000 (13:27 -0500)]
Fix indentation (and clang warning) in c-lang.c
I see this warning when building with clang:
CXX c-lang.o
/home/smarchi/src/binutils-gdb/gdb/c-lang.c:314:7: error: misleading indentation; statement is not part of the previous 'if' [-Werror,-Wmisleading-indentation]
*length = i * width;
^
/home/smarchi/src/binutils-gdb/gdb/c-lang.c:308:4: note: previous statement is here
if (extract_unsigned_integer (contents + i * width,
^
It took me a while to notice that some lines in that area have a
spurious space before the tabs, at the beginning of the ling. I'm not
sure how clang translates that to misleading indentation, but making the
indentation correct gets rid of the error.
There are many more instances of this in the code base (`grep -P '^ \t'
*.c`), if others think it's a good idea, it would be pretty easy to fix
them all up in one shot.
gdb/ChangeLog:
* c-lang.c (c_get_string, asm_language_defn): Remove space
before tab.
Tom Tromey [Wed, 18 Dec 2019 15:03:01 +0000 (08:03 -0700)]
Fix build failure on macOS
PR build/25250 notes that the gdb 9 pre-release fails to build on
macOS, due to a name clash between field_kind::STRING and the STRING
token in ada-exp.y. I am not sure (I couldn't reproduce this myself),
but presumably this is due to differences caused by the version of
bison in use there.
This patch works around the problem by renaming the field_kind
enumerator. I chose to rename this one because it is used in
relatively few places -- it's just an implementation detail of the
style code.
This version also renames field_kind::SIGNED for consistency.
Let me know what you think. I intend to check this in on the gdb 9
branch as well.
gdb/ChangeLog
2019-12-18 Tom Tromey <tromey@adacore.com>
/home/smarchi/src/binutils-gdb/gdb/top.c:1549:5: error: misleading indentation; statement is not part of the previous 'if' [-Werror,-Wmisleading-indentation]
fprintf_filtered (stream, _("\n\
^
/home/smarchi/src/binutils-gdb/gdb/top.c:1543:3: note: previous statement is here
if (SYSTEM_GDBINIT_DIR[0])
^
This looks like a legitimate warning, the fprintf_filtered is too much
indented. Fix it, and at the same time add a bit of whitespace to make
this function easier to read.
Alan Modra [Wed, 18 Dec 2019 05:07:44 +0000 (15:37 +1030)]
More signed overflow fixes
The arc fix in create_map avoiding signed overflow by casting an
unsigned char to unsigned int before shifting, shows one of the
dangers of blinding doing that. The problem in this case was that the
variable storing the value, newAuxRegister->address, was a long.
Using the unsigned cast meant that the 32-bit value was zero extended
when long is 64 bits. Previously we had a sign extension. Net result
was that comparisons in arcExtMap_auxRegName didn't match. Of course,
I could have cast the 32-bit unsigned value back to signed before
storing in a long, but it's neater to just use an unsigned int for the
address.
=================================================================
==2829136==ERROR: AddressSanitizer: new-delete-type-mismatch on 0x608000009a20 in thread T0:
object passed to delete has wrong type:
size of the allocated type: 88 bytes;
size of the deallocated type: 24 bytes.
#0 0x7f470fe2507e in operator delete(void*, unsigned long) /build/gcc/src/gcc/libsanitizer/asan/asan_new_delete.cc:177
#1 0x55f88c75700d in std::default_delete<tui_layout_base>::operator()(tui_layout_base*) const /usr/include/c++/9.2.0/bits/unique_ptr.h:81
#2 0x55f88c756328 in std::unique_ptr<tui_layout_base, std::default_delete<tui_layout_base> >::~unique_ptr() /usr/include/c++/9.2.0/bits/unique_ptr.h:284
#3 0x7f470ee536a6 in __run_exit_handlers (/usr/lib/libc.so.6+0x3e6a6)
#4 0x7f470ee5385d in __GI_exit (/usr/lib/libc.so.6+0x3e85d)
#5 0x55f88c69f2ac in quit_force(int*, int) /home/simark/src/binutils-gdb/gdb/top.c:1766
#6 0x55f88becc29a in captured_main_1 /home/simark/src/binutils-gdb/gdb/main.c:1183
#7 0x55f88becc814 in captured_main /home/simark/src/binutils-gdb/gdb/main.c:1192
#8 0x55f88becc8a9 in gdb_main(captured_main_args*) /home/simark/src/binutils-gdb/gdb/main.c:1217
#9 0x55f88b3159cd in main /home/simark/src/binutils-gdb/gdb/gdb.c:32
#10 0x7f470ee3c152 in __libc_start_main (/usr/lib/libc.so.6+0x27152)
#11 0x55f88b31579d in _start (/home/simark/build/binutils-gdb/gdb/gdb+0x11fb79d)
0x608000009a20 is located 0 bytes inside of 88-byte region [0x608000009a20,0x608000009a78)
allocated by thread T0 here:
#0 0x7f470fe238f8 in operator new(unsigned long) /build/gcc/src/gcc/libsanitizer/asan/asan_new_delete.cc:104
#1 0x55f88c750906 in tui_layout_split::clone() const /home/simark/src/binutils-gdb/gdb/tui/tui-layout.c:515
#2 0x55f88c74e60e in show_layout /home/simark/src/binutils-gdb/gdb/tui/tui-layout.c:90
#3 0x55f88c74e7db in tui_set_layout(tui_layout_type) /home/simark/src/binutils-gdb/gdb/tui/tui-layout.c:116
#4 0x55f88c782f4f in tui_enable() /home/simark/src/binutils-gdb/gdb/tui/tui.c:481
#5 0x55f88c74eeb2 in tui_layout_command /home/simark/src/binutils-gdb/gdb/tui/tui-layout.c:286
#6 0x55f88b6f969b in do_const_cfunc /home/simark/src/binutils-gdb/gdb/cli/cli-decode.c:107
#7 0x55f88b701859 in cmd_func(cmd_list_element*, char const*, int) /home/simark/src/binutils-gdb/gdb/cli/cli-decode.c:1952
#8 0x55f88c69b455 in execute_command(char const*, int) /home/simark/src/binutils-gdb/gdb/top.c:652
#9 0x55f88bec9026 in catch_command_errors /home/simark/src/binutils-gdb/gdb/main.c:400
#10 0x55f88becc1f2 in captured_main_1 /home/simark/src/binutils-gdb/gdb/main.c:1167
#11 0x55f88becc814 in captured_main /home/simark/src/binutils-gdb/gdb/main.c:1192
#12 0x55f88becc8a9 in gdb_main(captured_main_args*) /home/simark/src/binutils-gdb/gdb/main.c:1217
#13 0x55f88b3159cd in main /home/simark/src/binutils-gdb/gdb/gdb.c:32
#14 0x7f470ee3c152 in __libc_start_main (/usr/lib/libc.so.6+0x27152)
The problem is that the tui_layout_base is missing a virtual destructor.
We allocate a derived object (tui_layout_split), but delete it through a
tui_layout_base pointer. Since the tui_layout_base destructor is not
virtual, the derived (tui_layout_split) destructor is not called, only
the base destructor.
That code is not in gdb-9-branch, so I don't think this patch is
relevant for the stable branch.
Note that this is caught as a diagnostic with clang:
In file included from /home/simark/src/binutils-gdb/gdb/tui/tui-layout.c:22:
In file included from /home/simark/src/binutils-gdb/gdb/defs.h:28:
In file included from /home/simark/src/binutils-gdb/gdb/gdbsupport/common-defs.h:133:
In file included from /home/simark/src/binutils-gdb/gdb/gdbsupport/common-exceptions.h:25:
In file included from /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/9.2.0/../../../../include/c++/9.2.0/memory:80:
/usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/9.2.0/../../../../include/c++/9.2.0/bits/unique_ptr.h:81:2: error: delete called on 'tui_layout_base' that is abstract but has non-virtual destructor [-Werror,-Wdelete-abstract-non-virtual-dtor]
delete __ptr;
^
/usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/9.2.0/../../../../include/c++/9.2.0/bits/unique_ptr.h:284:4: note: in instantiation of member function 'std::default_delete<tui_layout_base>::operator()' requested here
get_deleter()(std::move(__ptr));
^
/home/simark/src/binutils-gdb/gdb/tui/tui-layout.c:54:41: note: in instantiation of member function 'std::unique_ptr<tui_layout_base, std::default_delete<tui_layout_base> >::~unique_ptr' requested here
static std::unique_ptr<tui_layout_base> applied_layout;
^
1 error generated.
GCC has the similar -Wdelete-non-virtual-dtor, enabled by -Wall, but it
doesn't show up because warnings are inhibited for system headers, where
std::unique_ptr is defined. There is a bug about it here:
Bernd Edlinger [Sun, 15 Dec 2019 10:05:47 +0000 (11:05 +0100)]
Fix skip.exp test failure observed with gcc-9.2.0
We need to step a second time with this gcc version.
The first step jumps back to main before entering foo.
Previously the control flow was from bar directly to foo.
Further ananlysis suggests, that this change in behavior started
with gcc-8.1.0 when -gcolumn-info was enabled by default.
The option -gcolumn-info was first implemented in gcc-7.1.0 but
default-disabled, so you can get the altered behavior already with
gcc-7 if you manually enable -gcolumn-info.
Previously there was just one point where line 30 (of skip.c) started:
[0x00000032] Advance Line by 27 to 28
[0x00000034] Copy
[0x00000035] Special opcode 63: advance Address by 4 to 0x4004cb and Line by 2 to 30
[0x00000036] Advance PC by constant 17 to 0x4004dc
[0x00000037] Special opcode 7: advance Address by 0 to 0x4004dc and Line by 2 to 32
But with -gcolumn-info enabled, we have line 30 three times with different column:
[0x00000034] Advance Line by 27 to 28
[0x00000036] Copy
[0x00000037] Set column to 9
[0x00000039] Special opcode 63: advance Address by 4 to 0x4004c6 and Line by 2 to 30
[0x0000003a] Set column to 17
[0x0000003c] Special opcode 75: advance Address by 5 to 0x4004cb and Line by 0 to 30
[0x0000003d] Set column to 3
[0x0000003f] Special opcode 75: advance Address by 5 to 0x4004d0 and Line by 0 to 30
[0x00000040] Special opcode 105: advance Address by 7 to 0x4004d7 and Line by 2 to 32
That could probably be filtered in dwarf2read.c to keep the old behavior, but
the new behavior makes still sense, even if we cannot really make use of the
column in the line number info for now.