Fix formatting of read_attribute_reprocess

I noticed that the start of read_attribute_reprocess had the wrong
formatting.  This patch fixes it.

gdb/ChangeLog
2020-03-28  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (read_attribute_reprocess): Fix formatting.

ld: Set COMMONPAGESIZE for i386 Solaris

Since COMMONPAGESIZE is set for other Solaris targets, including x86-64
Solaris, also set COMMONPAGESIZE for i386 Solaris to fix

FAIL: Build pr20995-2.so
FAIL: pr20995-2

PR 25732
* emulparams/elf_i386_ldso.sh (COMMONPAGESIZE): New.
* testsuite/ld-elf/shared.exp:Don't xfail pr20995-2 tests for
Solaris.

Re: Adjust objcopy_test

Last patch didn't manage to xfail spu due to clear_xfail *-*-*elf*.
Clearing *-*-*elf* dates back to a time when we had rather a lot more
setup_xfail patterns, so limiting it to hppa*-*-*elf*.  Also,
mips-*-irix ought to have been mips-*-irix* and I'm having second
thoughts about xfailing mips and hiding what looks like a problem: If
the mips target is supposed to emit names for local section symbols
and does so for objcopy, why isn't it doing the same for ld?  Also,
lots more mips targets would be subject to this test failing.  So I'm
backing out those xfails and leaving it to someone more knowledgeable
about mips.

* testsuite/binutils-all/objcopy.exp (objcopy_test): Only
clear_xfail hppa*-*-*elf*.  Revert mips xfails.

Adjust objcopy_test

xfails spu due to a note section getting a different vma, and some
mips targets that give section symbols a name string.  I added -p
for the executable test in an attempt to fix all the pe target fails,
but that doesn't preserve the date/time for some reason.

* testsuite/binutils-all/objcopy.exp (objcopy_test): Move xfails
from here to calls.  Remove "m8*-*-*" entry.  Don't xfail tic54x
but do xfail spu, mipstx39 and mips-sgi-irix for the executable
test.  Pass "-p" to objcopy for the executable test.

Automatic date update in version.in

Always fix system DLL paths for 32bit programs

GetModuleFileNameEx might also return the 64bit system directory for 32bit
programs even for a 32bit gdb:

(gdb) info sharedlibrary
From        To          Syms Read   Shared Object Library
0x779d0000  0x77b34d20  Yes (*)     C:\Windows\SysWOW64\ntdll.dll
0x76850000  0x7694ad9c  Yes (*)     C:\Windows\syswow64\kernel32.dll
0x75421000  0x75466a18  Yes (*)     C:\Windows\syswow64\KernelBase.dll
0x6fbe1000  0x6fcca1c0  Yes (*)     C:\Windows\system32\dbghelp.dll
0x76d31000  0x76ddb2c4  Yes (*)     C:\Windows\syswow64\msvcrt.dll

So this makes the path conversion for all 32bit programs.

gdb/ChangeLog:

2020-03-27  Hannes Domani  <ssbssa@yahoo.de>

* windows-nat.c (windows_add_all_dlls): Fix system dll paths.

ld: Xfail pr20995-2 tests for Solaris

Xfail pr20995-2 tests for Solaris since Solaris doesn't support RELRO.

PR 25732
* testsuite/ld-elf/shared.exp: Xfail pr20995-2 tests for Solaris.

ld: Skip some x86 IFUNC tests for Solaris

Since Solaris ld.so.1 doesn't support ifunc and never will, skip some
x86 IFUNC tests for Solaris.

PR 25732
* testsuite/ld-ifunc/ifunc-23a-x86.d: Add notarget for Solaris.
* testsuite/ld-ifunc/ifunc-24a-x86.d: Likewise.
* testsuite/ld-ifunc/ifunc-25a-x86.d: Likewise.

gdbsupport: Resolve shellcheck issues in create-version.sh script

Run shellcheck (version 0.4.7) on the create-version.sh script, and
resolve the issues it highlighter - they all seemed reasonable.

gdbsupport/ChangeLog:

* create-version.sh: Resolve issues highlighted by shellcheck.

Add testcase for PR 25662 invalid sh_offset for section

binutils/ChangeLog:

2020-03-27  Jozef Lawrynowicz  <jozef.l@mittosystems.com>

PR binutils/25662
* testsuite/binutils-all/objcopy.exp (objcopy_test): Add argument to
specify whether an object file or executable should be built and tested.
Change test names to report whether an object file or executable is
being tested.
* testsuite/binutils-all/pr25662.ld: New test.
* testsuite/binutils-all/pr25662.s: New test.

Re: readelf looping in process_archive

This patch fixes a leak of qualified_name caused by 4c83662712 and a
double free introduced by fd486f32d1.  Not breaking out of the loop
results in an error: "failed to seek to next archive header".  That's
slightly better than silently preventing the possibility of endless
loops.

* readelf.c (process_archive): Don't double free qualified_name.
Don't break out of loop with "negative" archive_file_size, just
set file offset to max.

Automatic date update in version.in

Support AT_BSDFLAGS on FreeBSD.

FreeBSD's kernel recently added a new ELF auxiliary vector entry
holding a mask of software features provided by the kernel.  This
change fixes 'info auxv' to report the name and description for this
vector entry instead of '???'.

include/ChangeLog:

* elf/common.h (AT_FREEBSD_BSDFLAGS): Define.

gdb/ChangeLog:

* fbsd-tdep.c (fbsd_print_auxv_entry): Handle AT_FREEBSD_BSDFLAGS.

Change two functions to be methods on struct attribute

This changes dwarf2_get_ref_die_offset and
dwarf2_get_attr_constant_value to be methods on struct attribute.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (handle_data_member_location, dwarf2_add_field)
(mark_common_block_symbol_computed, read_tag_string_type)
(attr_to_dynamic_prop, read_subrange_type): Update.
(dwarf2_get_ref_die_offset, dwarf2_get_attr_constant_value): Move
to be methods on struct attribute.
(skip_one_die, process_imported_unit_die, read_namespace_alias)
(read_call_site_scope, partial_die_info::read)
(partial_die_info::read, lookup_die_type, follow_die_ref):
Update.
* dwarf2/attribute.c (attribute::get_ref_die_offset): New method,
from dwarf2_get_ref_die_offset.
(attribute::constant_value): New method, from
dwarf2_get_attr_constant_value.
* dwarf2/attribute.h (struct attribute) <get_ref_die_offset>:
Declare method.
<constant_value>: New method.

Move DWARF-constant stringifying code to new file

This moves the DWARF debugging functions that stringify various
constants to a new file, dwarf2/stringify.c.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf_unit_type_name, dwarf_tag_name)
(dwarf_attr_name, dwarf_form_name, dwarf_bool_name)
(dwarf_type_encoding_name): Move to stringify.c.
* Makefile.in (COMMON_SFILES): Add dwarf2/stringify.c.
* dwarf2/stringify.c: New file.
* dwarf2/stringify.h: New file.

Rewrite new die_info methods

This rewrites the two new die_info methods to iterate over attributes
rather than to do two separate searches.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/die.h (struct die_info) <addr_base, ranges_base>:
Rewrite.

Change two more functions to be methods on die_info

This changes lookup_addr_base and lookup_ranges_base to be methods on
die_info.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/die.h (struct die_info) <addr_base, ranges_base>: New
methods.
* dwarf2/read.c (lookup_addr_base): Move to die.h.
(lookup_ranges_base): Likewise.
(read_cutu_die_from_dwo, read_full_die_1): Update.

Remove sibling_die

The sibling_die helper function does not seem to add much value,
considering that many other fields of die_info are directly accessed.
So, this removes it.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (read_import_statement, read_file_scope)
(read_type_unit_scope, inherit_abstract_dies, read_func_scope)
(read_lexical_block_scope, read_call_site_scope)
(dwarf2_get_subprogram_pc_bounds, get_scope_pc_bounds)
(handle_struct_member_die, process_structure_scope)
(update_enumeration_type_from_children)
(process_enumeration_scope, read_array_type, read_common_block)
(read_namespace, read_module, read_subroutine_type): Update.
(sibling_die): Remove.

Change dwarf2_attr_no_follow to be a method

This changes dwarf2_attr_no_follow to be a method on die_info.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (lookup_addr_base, lookup_ranges_base)
(build_type_psymtabs_reader, read_structure_type)
(read_enumeration_type, read_full_die_1): Update.
(dwarf2_attr_no_follow): Move to die.h.
* dwarf2/die.h (struct die_info) <attr>: New method.

Remove dwarf2_cu::base_known

This removes dwarf2_cu::base_known, changing base_address to be a
gdb::optional.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (struct dwarf2_cu) <base_known>: Remove.
<base_address>: Now an optional.
(dwarf2_find_base_address, dwarf2_rnglists_process)
(dwarf2_ranges_process, fill_in_loclist_baton)
(dwarf2_symbol_mark_computed): Update.

Move die_info to new header

This moves struct die_info to a new header, dwarf2/die.h.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (struct die_info): Move to die.h.
* dwarf2/die.h: New file.

Move more code to line-header.c

This moves some more code out of read.c and into line-header.c.
dwarf_decode_line_header is split into two -- the part remaining in
read.c handles interfacing to the dwarf2_cu; while the part in
line-header.c (more or less) purely handles the actual decoding.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/line-header.h (dwarf_decode_line_header): Declare.
* dwarf2/read.c
(dwarf2_statement_list_fits_in_line_number_section_complaint):
Move to line-header.c.
(read_checked_initial_length_and_offset, read_formatted_entries):
Likewise.
(dwarf_decode_line_header): Split into two.
* dwarf2/line-header.c
(dwarf2_statement_list_fits_in_line_number_section_complaint):
Move from read.c.
(read_checked_initial_length_and_offset, read_formatted_entries):
Likewise.
(dwarf_decode_line_header): New function, split from read.c.

Convert read_indirect_line_string to a method

This changes read_indirect_line_string to be a method on
dwarf2_per_objfile.  This makes it a bit simpler to share between
files.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.h (struct dwarf2_per_objfile) <read_line_string>:
Declare method.
* dwarf2/read.c (read_attribute_value): Update.
(dwarf2_per_objfile::read_line_string): Rename from
read_indirect_line_string.
(read_formatted_entries): Update.

Trivial fix in dwarf_decode_macro_bytes

One spot in dwarf_decode_macro_bytes could use the existing "objfile"
local variable.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/macro.c (dwarf_decode_macro_bytes): Use objfile local
variable.

Use a const dwarf2_section_info in macro reader

This changes the DWARF macro reader to use a const dwarf2_section_info.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/macro.h (dwarf_decode_macros): Make section parameter
const.
* dwarf2/macro.c (skip_form_bytes, skip_unknown_opcode)
(dwarf_decode_macro_bytes, dwarf_decode_macros): Make section
parameter const.

Use a const line_header in macro reader

This changes the DWARF macro reader to use a const line_header.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf_decode_macros): Make "lh" const.
* dwarf2/macro.h (dwarf_decode_macros): Constify "lh" parameter.
* dwarf2/macro.c (macro_start_file): Constify "lh" parameter.
(dwarf_decode_macro_bytes, dwarf_decode_macros): Likewise.

Make some line_header methods const

This changes a few line_header methods to be const.  In some cases, a
const overload is added.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/line-header.h (struct line_header) <is_valid_file_index,
file_names_size, file_full_name, file_file_name>: Use const.
<file_name_at, file_names>: Add const overload.
* dwarf2/line-header.c (line_header::file_file_name)
(line_header::file_full_name): Update.

Move code to new file dwarf2/macro.c

This moves some more code out of dwarf2/read.c, introducing new files
dwarf2/macro.c and dwarf2/macro.h.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf2_macro_malformed_definition_complaint)
(macro_start_file, consume_improper_spaces)
(parse_macro_definition, skip_form_bytes, skip_unknown_opcode)
(dwarf_parse_macro_header, dwarf_decode_macro_bytes)
(dwarf_decode_macros): Move to macro.c.
* dwarf2/macro.c: New file.
* dwarf2/macro.h: New file.
* Makefile.in (COMMON_SFILES): Add dwarf2/macro.c.

Add dwarf2_section_info::read_string method

This moves a string-reading function to be a method on
dwarf2_section_info, and then updates the users.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/section.h (struct dwarf2_section_info) <read_string>: New
method.
* dwarf2/section.c: New method.  From
read_indirect_string_at_offset_from.
* dwarf2/read.c (mapped_debug_names::namei_to_name): Update.
(read_indirect_string_at_offset_from): Move to section.c.
(read_indirect_string_at_offset): Rewrite.
(read_indirect_line_string_at_offset): Remove.
(read_indirect_string, read_indirect_line_string)
(dwarf_decode_macro_bytes): Update.

Convert dwarf2_section_buffer_overflow_complaint to a method

This changes dwarf2_section_buffer_overflow_complaint to be a method
on dwarf2_section_info.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/section.h (struct dwarf2_section_info)
<overload_complaint>: Declare.
(dwarf2_section_buffer_overflow_complaint): Don't declare.
* dwarf2/section.c (dwarf2_section_info::overflow_complaint):
Rename from dwarf2_section_buffer_overflow_complaint.
* dwarf2/read.c (skip_one_die, partial_die_info::read)
(skip_form_bytes, dwarf_decode_macro_bytes): Update.

Move dwarf2_section_buffer_overflow_complaint to dwarf2/section.c

This moves dwarf2_section_buffer_overflow_complaint to
dwarf2/section.c.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/section.h (dwarf2_section_buffer_overflow_complaint):
Declare.
* dwarf2/section.c (dwarf2_section_buffer_overflow_complaint):
Move from read.c.
* dwarf2/read.c (dwarf2_section_buffer_overflow_complaint): Move
to section.c.

Split dwarf_decode_macros into two overloads

This splits dwarf_decode_macros into two overloads -- one that's
suitable for splitting into a separate file, and one that finds the
correct section and should remain in dwarf2/read.c.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (dwarf_decode_macros): Split into two overloads.

Change dwarf_decode_macro_bytes calling convention

This changes dwarf_decode_macro_bytes to accept a buildsym_compunit
rather than a dwarf2_cu.  This enables some subsequent changes; and
also makes the function accept a "more specific" parameter.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (macro_start_file): Change "cu" parameter to
"builder".
(dwarf_decode_macro_bytes): Likewise.  Add dwarf2_per_objfile
parameter.
(dwarf_decode_macros): Update.

Add dwz.c and dwz_file::read_string

This changes read_indirect_string_from_dwz to be a method on the
dwz_file, and adds a new dwarf2/dwz.c file.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.c (read_attribute_value): Update.
(read_indirect_string_from_dwz): Move to dwz.c; change into
method.
(dwarf_decode_macro_bytes): Update.
* dwarf2/dwz.h (struct dwz_file) <read_string>: Declare method.
* dwarf2/dwz.c: New file.
* Makefile.in (COMMON_SFILES): Add dwz.c.

Introduce dwarf2/dwz.h

This moves "struct dwz_file" to a new header file, dwarf2/dwz.h.

gdb/ChangeLog
2020-03-26  Tom Tromey  <tom@tromey.com>

* dwarf2/read.h (struct dwz_file): Move to dwz.h.
* dwarf2/read.c: Add include.
* dwarf2/index-write.c: Add include.
* dwarf2/index-cache.c: Add include.
* dwarf2/dwz.h: New file.

Revert earlier delta adding bfd_coff_get_internal_extra_pe_aouthdr() function.

* cofflink.c (bfd_coff_get_internal_extra_pe_aouthdr): Delete.
* libbfd-in.h (bfd_coff_get_internal_extra_pe_aouthdr): Remove
prototype.
* libbfd.h: Regenerate.

Re: H8300 use of uninitialised value

This patch also had some problems.  Calculation of maxlen was wrong,
and the insn arg loop needed rearranging to work with a correct length.

* disassemble.h (opcodes_assert): Declare.
(OPCODES_ASSERT): Define.
* disassemble.c: Don't include assert.h.  Include opintl.h.
(opcodes_assert): New function.
* h8300-dis.c (bfd_h8_disassemble_init): Use OPCODES_ASSERT.
(bfd_h8_disassemble): Reduce size of data array.  Correctly
calculate maxlen.  Omit insn decoding when insn length exceeds
maxlen.  Exit from nibble loop when looking for E, before
accessing next data byte.  Move processing of E outside loop.
Replace tests of maxlen in loop with assertions.

Re: i386msdos uninitialised read

Another fix.

* i386msdos.c (msdos_object_p): Catch -1 return from bfd_bread.

Re: ARC: Use of uninitialised value

git commit beea5cc1bc fixed one use of an uninitialised value but
ignored another one a few lines later.

* arc-dis.c (find_format): Init needs_limm.  Simplify use of limm.

alpha-vms: Sanity check ETIR__C_CTL_DFLOC index

I doubt anyone will want to create more than 16M debug location
entries.  If there is no bound the object format allows for 32-bit
indices and of course fuzzers find that and attempt allocation of up
to a 16G byte array.  The patch also fixes potential integer overflows
in calculating the array size.

* vms-alpha.c (dst_define_location): Limit size of dst_ptr_offsets
array.
(_bfd_vms_slurp_object_records): Rename "err" to "ok".

Automatic date update in version.in

Fix error message in compile-object-load.c

I noticed that an error message in compile-object-load.c mentions the
wrong symbol name.  The loop just above the error is looking for
COMPILE_I_EXPR_VAL, but the error references COMPILE_I_EXPR_PTR_TYPE.

I'm checking this in as obvious.  I don't have a test case -- I
noticed it because another patch I'm working on caused this error to
be thrown, but that was due to regression in my patch.

gdb/ChangeLog
2020-03-25  Tom Tromey  <tom@tromey.com>

* compile/compile-object-load.c (get_out_value_type): Mention
correct symbol name in error message.

Fix WOW64 process system DLL paths

GetModuleFileNameEx returns for some DLLs of WOW64 processes
the path inside the 64bit system directory instead of the 32bit
syswow64 directory.

Problem happens e.g. with dbghelp.dll:

(gdb) start
Temporary breakpoint 1 at 0x415a00: file fiber.cpp, line 430.
Starting program: C:\src\tests\fiber.exe
warning: `C:\Windows\system32\dbghelp.dll': Shared library architecture i386:x86-64 is not compatible with target architecture i386.

Temporary breakpoint 1, main () at fiber.cpp:430
430     {
(gdb) info sharedlibrary
From        To          Syms Read   Shared Object Library
0x77070000  0x771d4d20  Yes (*)     C:\Windows\SysWOW64\ntdll.dll
0x74dc0000  0x74ebad9c  Yes (*)     C:\Windows\syswow64\kernel32.dll
0x75341000  0x75386a18  Yes (*)     C:\Windows\syswow64\KernelBase.dll
0x6f6a1000  0x6f7c48fc  Yes (*)     C:\Windows\system32\dbghelp.dll
0x74d01000  0x74dab2c4  Yes (*)     C:\Windows\syswow64\msvcrt.dll
(*): Shared library is missing debugging information.

This detects this situation and converts the DLL path to the
syswow64 equivalent.

gdb/ChangeLog:

2020-03-25  Hannes Domani  <ssbssa@yahoo.de>

* windows-nat.c (windows_add_all_dlls): Fix system dll paths.

readelf looping in process_archive

With a crafted "negative" ar_hdr.ar_size it is possible to make
readelf loop.  This patch catches the overflow in a file offset
calculation.

* readelf.c (process_archive): Prevent endless loop.

Add a new function to the BFD library to allow users access to the COFF internal_extra_pe_outhdr structure.

* cofflink.c (bfd_coff_get_internal_extra_pe_aouthdr): New
function.
* libbfd-in.h (bfd_coff_get_internal_extra_pe_aouthdr): Prototype.
* libbfd.h: Regenerate.

The "b" flag for COFF sections only unsets the LOAD attribute.  It should also clear the CONTENTS attribute so that named bss sections don't take up space in an object file.  This can be achieved by setting the 'bss' flag in seg_info.

* config/obj-coff.c (obj_coff_section): Set the bss flag on
sections with the "b" attribute.

[gdb] Print user/includes fields for maint commands

The type struct compunit_symtab contains two fields (disregarding field next)
that express relations with other compunit_symtabs: user and includes.

These fields are currently not printed with "maint info symtabs" and
"maint print symbols".

Fix this such that for "maint info symtabs" we print:
...
   { ((struct compunit_symtab *) 0x23e8450)
     debugformat DWARF 2
     producer (null)
     dirname (null)
     blockvector ((struct blockvector *) 0x23e8590)
+    user ((struct compunit_symtab *) 0x2336280)
+    ( includes
+      ((struct compunit_symtab *) 0x23e85e0)
+      ((struct compunit_symtab *) 0x23e8960)
+    )
         { symtab <unknown> ((struct symtab *) 0x23e85b0)
           fullname (null)
           linetable ((struct linetable *) 0x0)
         }
   }
...

And for "maint print symbols" we print:
...
-Symtab for file <unknown>
+Symtab for file <unknown> at 0x23e85b0
 Read from object file /data/gdb_versions/devel/a.out (0x233ccf0)
 Language: c

 Blockvector:

 block #000, object at 0x23e8530, 0 syms/buckets in 0x0..0x0
   block #001, object at 0x23e84d0 under 0x23e8530, 0 syms/buckets in 0x0..0x0

+Compunit user: 0x2336300
+Compunit include: 0x23e8900
+Compunit include: 0x23dd970
...
Note: for user and includes we don't list the actual compunit_symtab address,
but instead the corresponding symtab address, which allows us to find that
symtab elsewhere in the output (given that we also now print the address of
symtabs).

gdb/ChangeLog:

2020-03-25  Tom de Vries  <tdevries@suse.de>

* symtab.h (is_main_symtab_of_compunit_symtab): New function.
* symmisc.c (dump_symtab_1): Print user and includes fields.
(maintenance_info_symtabs): Same.

gdb/riscv: Apply NaN boxing when writing return values into registers

When setting up function parameters we already perform NaN boxing, as
required by the RISC-V ABI, however, we don't do this when writing
values into registers as part of setting up a return value.

This commit moves the NaN boxing code into a small helper function,
and then makes use of this function when setting up function
parameters, and also when setting up return values.

This should resolve this failure:

  FAIL: gdb.base/return-nodebug.exp: float: full width of the returned result

gdb/ChangeLog:

PR gdb/25489
* riscv-tdep.c (riscv_arg_info::c_offset): Update comment.
(riscv_regcache_cooked_write): New function.
(riscv_push_dummy_call): Use new function.
(riscv_return_value): Likewise.

arc: Use correct string when printing bfd DEBUG data

PRINT_DEBUG_RELOC_INFO_BEFORE() macro prints bunch of parameters
for debugging purposes.  Due to a seemingly copy/paste mistake,
the "input_section->vma" is printed under the field name
"symbol_section->vma".  This commit fixes that.

This fix is a courtesy of xiangzhai.

* elf32-arc.c (PRINT_DEBUG_RELOC_INFO_BEFORE): Use the
correct field name in the output string.

h8300-linux ld testsuite

This fixes lots of fails caused by h8300-linux not supporting -shared
and related options.  I've also fixed ld-h8300 tests to accept the
valid h8300-linux -m options .

pr22450.d is also fixed for avr, crx, ip2k, m68hc11 and xc16x, and the
new pr25708 test for hppa64.

* testsuite/ld-elf/compress1c.d: xfail h8300.
* testsuite/ld-elf/compressed1c.d: Likewise.
* testsuite/ld-elf/compressed1e.d: Likewise.
* testsuite/ld-elf/dynamic1.d: Likewise.
* testsuite/ld-elf/dynsym1.d: Likewise.
* testsuite/ld-elf/ehdr_start-shared.d: Likewise.
* testsuite/ld-elf/exclude3b.d: Likewise.
* testsuite/ld-elf/export-class.exp: Don't run without -shared support.
* testsuite/ld-elf/global1.d: xfail h8300.
* testsuite/ld-elf/hash.d: Likewise.
* testsuite/ld-elf/loadaddr1.d: Likewise.
* testsuite/ld-elf/loadaddr2.d: Likewise.
* testsuite/ld-elf/local1.d: Likewise.
* testsuite/ld-elf/mbind1b.d: Likewise.
* testsuite/ld-elf/mbind1c.d: Likewise.
* testsuite/ld-elf/now-1.d: Likewise.
* testsuite/ld-elf/now-2.d: Likewise.
* testsuite/ld-elf/now-3.d: Likewise.
* testsuite/ld-elf/now-4.d: Likewise.
* testsuite/ld-elf/pie.d: Likewise.
* testsuite/ld-elf/pr12975.d: Likewise.
* testsuite/ld-elf/pr13177.d: Likewise.
* testsuite/ld-elf/pr13195.d: Likewise.
* testsuite/ld-elf/pr16322.d: Likewise.
* testsuite/ld-elf/pr16498a.d: Likewise.
* testsuite/ld-elf/pr16498b.d: Likewise.
* testsuite/ld-elf/pr17615.d: Likewise.
* testsuite/ld-elf/pr19162.d: Likewise.
* testsuite/ld-elf/pr19539.d: Likewise.
* testsuite/ld-elf/pr19617a.d: Likewise.
* testsuite/ld-elf/pr19698.d: Likewise.
* testsuite/ld-elf/pr20513c.d: Likewise.
* testsuite/ld-elf/pr20513d.d: Likewise.
* testsuite/ld-elf/pr21389a.d: Likewise.
* testsuite/ld-elf/pr21389b.d: Likewise.
* testsuite/ld-elf/pr21389c.d: Likewise.
* testsuite/ld-elf/pr21562a.d: Likewise.
* testsuite/ld-elf/pr21562b.d: Likewise.
* testsuite/ld-elf/pr21562c.d: Likewise.
* testsuite/ld-elf/pr21562d.d: Likewise.
* testsuite/ld-elf/pr21562e.d: Likewise.
* testsuite/ld-elf/pr21562f.d: Likewise.
* testsuite/ld-elf/pr21562g.d: Likewise.
* testsuite/ld-elf/pr21562h.d: Likewise.
* testsuite/ld-elf/pr21562i.d: Likewise.
* testsuite/ld-elf/pr21562j.d: Likewise.
* testsuite/ld-elf/pr21562k.d: Likewise.
* testsuite/ld-elf/pr21562l.d: Likewise.
* testsuite/ld-elf/pr21562m.d: Likewise.
* testsuite/ld-elf/pr21562n.d: Likewise.
* testsuite/ld-elf/pr21903a.d: Likewise.
* testsuite/ld-elf/pr21903b.d: Likewise.
* testsuite/ld-elf/pr21903d.d: Likewise.
* testsuite/ld-elf/pr22269a.d: Likewise.
* testsuite/ld-elf/pr22269b.d: Likewise.
* testsuite/ld-elf/pr22393-1a.d: Likewise.
* testsuite/ld-elf/pr22393-1b.d: Likewise.
* testsuite/ld-elf/pr22393-1c.d: Likewise.
* testsuite/ld-elf/pr22393-1d.d: Likewise.
* testsuite/ld-elf/pr22393-1e.d: Likewise.
* testsuite/ld-elf/pr22393-1f.d: Likewise.
* testsuite/ld-elf/pr22423.d: Likewise.
* testsuite/ld-elf/pr22450.d: xfail avr, crx, h8300, ip2k, m68hc11,
and xc16x.
* testsuite/ld-elf/pr23658-1c.d: xfail h8300.
* testsuite/ld-elf/pr25708.d: xfail h8300 and hppa64.
* testsuite/ld-elf/rpath-1.d: xfail h8300.
* testsuite/ld-elf/rpath-2.d: Likewise.
* testsuite/ld-elf/runpath-1.d: Likewise.
* testsuite/ld-elf/runpath-2.d: Likewise.
* testsuite/ld-elf/sizeofb.d: Likewise.
* testsuite/ld-elf/startofb.d: Likewise.
* testsuite/ld-elf/strtab.d: Likewise.
* testsuite/ld-elf/textaddr1.d: Likewise.
* testsuite/ld-elf/textaddr2.d: Likewise.
* testsuite/ld-elf/textaddr4.d: Likewise.
* testsuite/ld-elf/textaddr5.d: Likewise.
* testsuite/ld-elf/textaddr6.d: Likewise.
* testsuite/ld-elf/unknown2.d: Likewise.
* testsuite/ld-undefined/entry-3.d: Likewise.
* testsuite/ld-undefined/entry-4.d: Likewise.
* testsuite/ld-h8300/h8300.exp: Pass appropriate ld -m option.
* testsuite/ld-h8300/gcsection.d: Remove ld -m option and relax
file format match.
* testsuite/ld-h8300/relax-2.d: Likewise.
* testsuite/ld-h8300/relax-3.d: Likewise.
* testsuite/ld-h8300/relax-4.d: Likewise.
* testsuite/ld-h8300/relax-5.d: Likewise.
* testsuite/ld-h8300/relax-6.d: Likewise.
* testsuite/ld-h8300/relax-7.d: Likewise.
* testsuite/ld-h8300/relax.d: Relax file format match.

PR25662, invalid sh_offset for first section in segment with phdrs

PR 25662
* elf.c (assign_file_positions_for_load_sections): Adjust offset
for SHT_NOBITS section if first in segment.

Automatic date update in version.in

bfd: Add a bfd_boolean argument to bfd_get_symbol_version_string

We can't call _bfd_elf_get_symbol_version_name from nm.c since it isn't
available for all target configurations.  This patch add a bfd_boolean
argument to bfd_get_symbol_version_string instead.

bfd/

PR binutils/25708
* elf-bfd.h (_bfd_elf_get_symbol_version_name): Renamed to ...
(_bfd_elf_get_symbol_version_string): This.
* elf.c (_bfd_elf_get_symbol_version_name): Renamed to ...
(_bfd_elf_get_symbol_version_string): This.
(bfd_elf_print_symbol): Pass TRUE to
_bfd_elf_get_symbol_version_string.
* libbfd-in.h (_bfd_nosymbols_get_symbol_version_string): Add a
bfd_boolean argument.
* syms.c (_bfd_nosymbols_get_symbol_version_string): Likewise.
* targets.c  (_bfd_get_symbol_version_string): Likewise.
(bfd_get_symbol_version_string): Likewise.
* bfd-in2.h: Regenerated.

binutils/

PR binutils/25708
* nm.c (print_symname): Replace _bfd_elf_get_symbol_version_name
with bfd_get_symbol_version_string.
(print_symbo): Pass TRUE to bfd_get_symbol_version_string.
* objdump.c (objdump_print_symname): Likewise.

Uninitialised memory read in z80-dis.c

objdump -d -m ez80-adl testcase:
 .byte 0x40,0xfd

* z80-dis.c (suffix): Init mybuf.

gdb: bool-ify follow_fork

Change parameters and return value of the various follow_fork
functions/methods from int to bool.

gdb/ChangeLog:

* fbsd-nat.c (fbsd_nat_target::follow_fork): Change bool to int.
* fbsd-nat.h (class fbsd_nat_target) <follow_fork>: Likewise.
* inf-ptrace.c (inf_ptrace_target::follow_fork): Likewise.
* inf-ptrace.h (struct inf_ptrace_target) <follow_fork>: Likewise.
* infrun.c (follow_fork): Likewise.
(follow_fork_inferior): Likewise.
* linux-nat.c (linux_nat_target::follow_fork): Likewise.
* linux-nat.h (class linux_nat_target): Likewise.
* remote.c (class remote_target) <follow_fork>: Likewise.
(remote_target::follow_fork): Likewise.
* target-delegates.c: Re-generate.
* target.c (default_follow_fork): Likewise.
(target_follow_fork): Likewise.
* target.h (struct target_ops) <follow_fork>: Likewise.
(target_follow_fork): Likewise.

Update changelog.

Add code to the BFD library to handle opening files with pathnames longer than MAX_PATH on Win32 systems.

PR 25713
* bfdio.c (_bfd_real_fopen): Add code to handle long filenames on
Win32 systems.

Fix assertion failure in the BFD library when linking with --emit-relocs enabled.

PR 25681
* elf.c (_bfd_elf_map_sections_to_segments): When looking for a
segment to use for PT_GNU_RELRO, ignore empty sections in a
segment's current list.

bfd: Change num_group to unsigned int

elf.c failed with to with GCC 10 as of

commit 906b3eb9df6c577d3f6e9c3ea5c9d7e4d1e90536
Author: Martin Liska <mliska@suse.cz>
Date:   Tue Mar 24 11:40:10 2020 +0100

    Improve endianess detection.

            PR lto/94249
            * plugin-api.h: Add more robust endianess detection.

binutils-gdb/bfd/elf.c: In function ‘setup_group’:
binutils-gdb/bfd/elf.c:740:35: error: overflow in conversion from ‘unsigned int’ to ‘int’ changes value from ‘num_group = 4294967295’ to ‘-1’ [-Werror=overflow]
  740 |     elf_tdata (abfd)->num_group = num_group = -1;
      |                                   ^~~~~~~~~
cc1: all warnings being treated as errors
make[2]: *** [Makefile:1608: elf.lo] Error 1

Change num_group in elf_obj_tdata to unsigned int to compile with GCC 10.

PR binutils/25717
* elf-bfd.h (elf_obj_tdata): Change num_group to unsigned int.

include: Sync plugin-api.h with GCC

Improve endianess detection.

PR lto/94249
* plugin-api.h: Add more robust endianess detection.

bfd: Display symbol version for nm -D

Extend _bfd_elf_get_symbol_version_string for nm -D to display symbol
version.  _bfd_elf_get_symbol_version_name is added to avoid updating
all XXX_get_symbol_version_string functions.

bfd/

PR binutils/25708
* elf-bfd.h (_bfd_elf_get_symbol_version_name): New.
* elf.c (_bfd_elf_get_symbol_version_name): New function.  Based
on the previous _bfd_elf_get_symbol_version_string.
(_bfd_elf_get_symbol_version_string): Use it.

binutils/

PR binutils/25708
* nm.c (SYM_NAME): Removed.
(print_symname): Add a pointer to struct extended_symbol_info
argument.  Call _bfd_elf_get_symbol_version_name to get symbol
version.
(print_symdef_entry): Pass NULL to print_symname.
(print_symbol_info_bsd): Update call to print_symname.
(print_symbol_info_sysv): Likewise.
(print_symbol_info_posix): Likewise.

ld/

PR binutils/25708
* testsuite/ld-elf/pr25708.d: New file.

[gdb] Print user for maint info psymtabs

The type struct partial_symtab contains two fields (disregarding field next)
that express relations with other symtabs: user and dependencies.

When using "maint print psymbols", we see both the dependencies and the user
fields:
...
Partial symtab for source file  (object 0x35ef270)
  ...
  Depends on 0 other partial symtabs.
  Shared partial symtab with user 0x35d5f40
...

But with "maint info psymtabs", we only see dependencies:
...
  { psymtab  ((struct partial_symtab *) 0x35ef270)
    ...
    dependencies (none)
  }
...

Add printing of the user field for "maint info psymtabs", such that we have:
...
   { psymtab  ((struct partial_symtab *) 0x35ef270)
     ...
+    user hello.c ((struct partial_symtab *) 0x35d5f40)
     dependencies (none)
   }
...

Tested on x86_64-linux.

gdb/ChangeLog:

2020-03-24  Tom de Vries  <tdevries@suse.de>

* psymtab.c (maintenance_info_psymtabs): Print user field.

Overlarge allocation in _bfd_generic_read_ar_hdr_mag

* archive.c (_bfd_generic_read_ar_hdr_mag): Sanity check extended
name size.  Use bfd_malloc rather than bfd_zmalloc, clearing just
struct areltdata.

readelf memory leak in process_mips_specific

* readelf.c (process_mips_specific): Free iconf on error path.

Automatic date update in version.in

Avoid trigger an assertion failure in the STABS parser by replacing the abort with an error return.

PR 25714
* wrstabs.c (stab_pop_type): Replace assertion with error return.
* write_stabs_in_sections_debugging_info: Likewise.
* stab_enum_type: Likewise.
* stab_modify_type: Likewise.
* stab_struct_field: Likewise.
* stab_end_struct_type: Likewise.
* stab_start_class_type: Likewise.
* stab_class_static_member: Likewise.
* stab_class_baseclass: Likewise.
* stab_class_start_method: Likewise.
* stab_class_method_var: Likewise.
* stab_class_end_method: Likewise.
* stab_end_class_type: Likewise.
* stab_typedef_type: Likewise.
* stab_start_function: Likewise.
* stab_end_block: Likewise.
* stab_lineno: Likewise.

Mention .tdata in comment in _bfd_elf_tls_setup()

This helps to find code areas which deal with the .tdata section.

bfd/

     * elflink.c (_bfd_elf_tls_setup): Mention .tdata in comment.

ECOFF archive uninitialised read

* ecoff.c (_bfd_ecoff_slurp_armap): Sanity check parsed_size and
symbol count.  Allocate an extra byte to ensure name strings
are terminated.  Sanity check name offsets.  Release memory on
error return.

i386msdos uninitialised read

Also reinstate ld i386aout for i386-msdos target, which doesn't build
otherwise.

bfd/
* i386msdos.c (msdos_object_p): Don't access e_lfanew when that
field hasn't been read.  Remove unnecessary casts.
ld/
* Makefile.am (ALL_EMULATION_SOURCES): Reinstate ei386aout.c.
Include ei386aout dep file.
* Makefile.in: Regenerate.
* po/BLD-POTFILES.in: Regenerate.

gdb/testsuite: Remove hard coded addresses from expected results

In commit:

  commit 6b8c53f2f1c0cf5bee46120d892d4c72571375eb
  Date:   Sat Feb 8 21:26:31 2020 +0000

      gdb/testsuite/fortran: Add mixed language stack test

The test incorrectly included two hard coded addresses in the expected
output, this commit replaces them with the $hex pattern.

gdb/testsuite/ChangeLog:

* gdb.fortran/mixed-lang-stack.exp: Replace two hard coded address
with $hex.

Automatic date update in version.in

XCOFF64 uninitialised read

Like git commit 67338173a4.

* coff64-rs6000.c (xcoff64_slurp_armap): Ensure size is large
enough to read number of symbols.

H8300 use of uninitialised value

* h8300-dis.c (bfd_h8_disassemble): Limit data[] access to that
successflly read from section.

ARC: Use of uninitialised value

* arc-dis.c (find_format): Use ISO C string concatenation rather
than line continuation within a string.  Don't access needs_limm
before testing opcode != NULL.

NS32K arg_bufs uninitialised

git commit d1e304bc27 was aimed at stopping uninitialised memory
access to the index_offset array.  Unfortunately that patch resulted
in a different array being uninitialised for all instructions with
more than two arguments.

* ns32k-dis.c (print_insn_arg): Update comment.
(print_insn_ns32k): Reduce size of index_offset array, and
initialize, passing -1 to print_insn_arg for args that are not
an index.  Don't exit arg loop early.  Abort on bad arg number.

s12z disassembler tidy

Don't ignore buffer memory read failure, or malloc failure.  Lots of
functions get a return status to pass these failures up the chain in
this patch.

opcodes/
* s12z-dis.c (abstract_read_memory): Don't print error on EOI.
* s12z-opc.c: Formatting.
(operands_f): Return an int.
(opr_n_bytes_p1): Return -1 on reaching buffer memory limit.
(opr_n_bytes2, bfextins_n_bytes, mul_n_bytes, bm_n_bytes),
(shift_n_bytes, mov_imm_opr_n_bytes, loop_prim_n_bytes),
(exg_sex_discrim): Likewise.
(create_immediate_operand, create_bitfield_operand),
(create_register_operand_with_size, create_register_all_operand),
(create_register_all16_operand, create_simple_memory_operand),
(create_memory_operand, create_memory_auto_operand): Don't
segfault on malloc failure.
(z_ext24_decode): Return an int status, negative on fail, zero
on success.
(x_imm1, imm1_decode, trap_decode, z_opr_decode, z_opr_decode2),
(imm1234, reg_s_imm, reg_s_opr, z_imm1234_8base, z_imm1234_0base),
(z_tfr, z_reg, reg_xy, lea_reg_xys_opr, lea_reg_xys, rel_15_7),
(decode_rel_15_7, cmp_xy, sub_d6_x_y, sub_d6_y_x),
(ld_18bit_decode, mul_decode, bm_decode, bm_rel_decode),
(mov_imm_opr, ld_18bit_decode, exg_sex_decode),
(loop_primitive_decode, shift_decode, psh_pul_decode),
(bit_field_decode): Similarly.
(z_decode_signed_value, decode_signed_value): Similarly.  Add arg
to return value, update callers.
(x_opr_decode_with_size): Check all reads, returning NULL on fail.
Don't segfault on NULL operand.
(decode_operation): Return OP_INVALID on first fail.
(decode_s12z): Check all reads, returning -1 on fail.
gas/
* testsuite/gas/s12z/truncated.d: Update expected output.

Automatic date update in version.in

include: Sync lto-symtab.h and plugin-api.h with GCC

Fix comma at end of enumerator list seen with -std=c++98.

* plugin-api.h (enum ld_plugin_symbol_type): Remove
comma after last value of an enum.
* lto-symtab.h (enum gcc_plugin_symbol_type): Likewise.

Automatic date update in version.in

Make dwarf2_evaluate_property parameter const

dwarf2_evaluate_property should not modify its "addr_stack"
parameter's contents.  This patch makes this part of the API, by
marking it const.

gdb/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* dwarf2/loc.h (dwarf2_evaluate_property): Make "addr_stack"
const.
* dwarf2/loc.c (dwarf2_evaluate_property): Make "addr_stack"
const.

[gdb/testsuite] Fix gdb.threads/omp-par-scope.exp XPASS

When running test-case gdb.threads/omp-par-scope.exp, I get this XPASS:
...
XPASS: gdb.threads/omp-par-scope.exp: nested_parallel: outer_threads: \
  outer stop: get valueof "num"
...
for test:
...
           set thread_num [get_valueof "" "num" "unknown"]
...

The intention of the test is to get the value of local variable num, which
has been set to:
...
    int num = omp_get_thread_num ();
...
but the actually printed value is 'num':
...
(gdb) print num^M
$76 = num^M
...

This is due to the fact that num is missing in the locals, so instead we find
the enum member 'num' of enum expression_operator in glibc/intl/plural-exp.h.

Fix this by getting the value using a new proc get_local_valueof, which uses
the "info locals" commands to get the value.

Tested on x86_64-linux, with gcc 7.5.0 (where the test xfails) and gcc
10.0.1 (where the test passes).

gdb: remove HAVE_DECL_PTRACE

I stumbled on this snippet in nat/gdb_ptrace.h:

    /* Some systems, in particular DEC OSF/1, Digital Unix, Compaq Tru64
       or whatever it's called these days, don't provide a prototype for
       ptrace.  Provide one to silence compiler warnings.  */

    #ifndef HAVE_DECL_PTRACE
    extern PTRACE_TYPE_RET ptrace();
    #endif

I believe this is unnecessary today and should be removed.  First, the
comment only mentions OSes we don't support (and to be honest, I had
never even heard of).

But most importantly, in C++, a declaration with empty parenthesis
declares a function that accepts no arguments, unlike in C.  So if this
declaration was really used, GDB wouldn't build, since all ptrace call
sites pass some arguments.  Since we haven't heard anything about this
causing some build failures since we have transitioned to C++, I
conclude that it's not used.

This patch removes it as well as the corresponding configure check.

gdb/ChangeLog:

* ptrace.m4: Don't check for ptrace declaration.
* config.in: Re-generate.
* configure: Re-generate.
* nat/gdb_ptrace.h: Don't declare ptrace if HAVE_DECL_PTRACE is
not defined.

gdbserver/ChangeLog:

* config.in: Re-generate.
* configure: Re-generate.

gdbsupport/ChangeLog:

* config.in: Re-generate.
* configure: Re-generate.

Update the return type of gdb_ptrace to be more flexible

Linux returns long from ptrace(2) and BSDs int.

gdb/ChangeLog:

       * amd64-bsd-nat.c (gdb_ptrace): Change return type from `int' to
       `PTRACE_TYPE_RET'.
       * i386-bsd-nat.c (gdb_ptrace): Likewise.
       * sparc-nat.c (gdb_ptrace): Likewise.
       * x86-bsd-nat.c (gdb_ptrace): Likewise.

Fix assert in c-exp.y

The "restrict" patch added some asserts to c-exp.y, but one spot was
copy-pasted and referred to the wrong table.  This was pointed out by
-fsanitize=address.  This patch fixes the bug.

gdb/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* c-exp.y (lex_one_token): Fix assert.

Avoid stringop-truncation errors

I configured with -fsanitize=address and built gdb.  linux-tdep.c and
ada-tasks.c failed to build due to some stringop-truncation errors,
e.g.:

In function ‘char* strncpy(char*, const char*, size_t)’,
    inlined from ‘int linux_fill_prpsinfo(elf_internal_linux_prpsinfo*)’ at ../../binutils-gdb/gdb/linux-tdep.c:1742:11,
    inlined from ‘char* linux_make_corefile_notes(gdbarch*, bfd*, int*)’ at ../../binutils-gdb/gdb/linux-tdep.c:1878:27:
/usr/include/bits/string_fortified.h:106:34: error: ‘char* __builtin_strncpy(char*, const char*, long unsigned int)’ specified bound 81 equals destination size [-Werror=stringop-truncation]

This patch fixes the problem by using "sizeof - 1" in the call to
strndup, as recommended in the GCC manual.  This doesn't make a
difference here because the next line, in all cases, sets the final
element to '\0' anyway.

gdb/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* ada-tasks.c (read_atcb): Use smaller length in strncpy call.
* linux-tdep.c (linux_fill_prpsinfo): Use smaller length in
strncpy call.

Fix column alignment in "maint info line-table"

Andrew Burgess pointed out on irc that "maint info line-table" doesn't
properly align the table headers.  This patch fixes the problem by
switching the table to use ui-out.

This required a small tweak to one test case, as ui-out will pad a
field using spaces, even at the end of a line.

gdb/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* symmisc.c (maintenance_print_one_line_table): Use ui_out.

gdb/testsuite/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* gdb.dwarf2/dw2-ranges-base.exp: Update regular expressions.

Fix Ada val_print removal regression

The removal of val_print caused a regression in the Ada code.  In one
scenario, a variant type would not be properly printed, because the
address of a component was lost.  This patch fixes the bug by changing
this API to be value-based.  This is cleaner and fixes the bug as a
side effect.

gdb/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* ada-valprint.c (print_variant_part): Remove parameters; switch
to value-based API.
(print_field_values): Likewise.
(ada_val_print_struct_union): Likewise.
(ada_value_print_1): Update.

gdb/testsuite/ChangeLog
2020-03-20  Tom Tromey  <tromey@adacore.com>

* gdb.ada/sub_variant/subv.adb: New file.
* gdb.ada/sub_variant.exp: New file.

Inherit ppc_nbsd_nat_target from nbsd_nat_target

gdb/ChangeLog:

* ppc-nbsd-nat.c (ppc_nbsd_nat_target): Inherit from
nbsd_nat_target instead of inf_ptrace_target.
* ppc-nbsd-nat.c: Include "nbsd-nat.h", as we are now using
nbsd_nat_target.

Add support for NetBSD threads in hppa-nbsd-nat.c

NetBSD ptrace(2) accepts thread id (LWP) as the 4th argument for threads.

gdb/ChangeLog:

* hppa-nbsd-nat.c (fetch_registers): New variable lwp and pass
it to the ptrace call.
* (store_registers): Likewise.

[gdb/testsuite] Fix timeouts in gdb.threads/step-over-*.exp

When running test-cases gdb.threads/step-over-lands-on-breakpoint.exp and
gdb.threads/step-over-trips-on-watchpoint.exp with target board
unix/-flto/-O0/-flto-partition=none/-ffat-lto-objects, we run into timeouts
due not being able to set a breakpoint and then trying to continue to that
breakpoint.

In total, we run into 186 timeouts, which roughly corresponds to half an hour:
...
$ grep "FAIL.*(timeout)" gdb.sum \
  | awk '{print $2}' \
  | sort \
  | uniq -c
     66 gdb.threads/step-over-lands-on-breakpoint.exp:
    120 gdb.threads/step-over-trips-on-watchpoint.exp:
...

Fix this by bailing out if the first break fails.

Tested on x86_64-linux, both with native and with target board mentioned above.

gdb/testsuite/ChangeLog:

2020-03-20  Tom de Vries  <tdevries@suse.de>

* gdb.threads/step-over-lands-on-breakpoint.exp (do_test): Bail out if
first break fails.
* gdb.threads/step-over-trips-on-watchpoint.exp: (do_test): Same.

Add support for the xdef and xref pseudo-ops to the Z80 assembler.

PR 25690
* config/tc-z80.c (md_pseudo_table): Add xdef anf xref pseudo ops.
* doc/c-z80.texi: Update documentation.

Add support for NetBSD threads in ppc-nbsd-nat.c

NetBSD ptrace(2) accepts thread id (LWP) as the 4th argument for threads.

gdb/ChangeLog:

* ppc-nbsd-nat.c (fetch_registers): New variable lwp and pass
        it to the ptrace call.
        * (store_registers): Likewise.

Fix compile time warnings about the possibility of using uninitialised fields in the Elf_Internal_Chdr structure when building readelf.

* readelf.c (get_compression_header): Add ATTRIBUTE_WARN_UNUSED_RESULT.
(process_section_headers): Check the return value from
get_compression_header.
(dump_section_as_strings): Likewise.
(dump_section_as_bytes): Likewise.
(load_specific_debug_section): Likewise.

plugin: Don't invoke LTO-wrapper

Don't invoke LTO-wrapper since the LTO wrapper approach is not only
slow but also unreliable.  For GCC 10 or newer, LDPT_ADD_SYMBOLS_V2
will be used.

bfd/

* configure.ac (HAVE_EXECUTABLE_SUFFIX): Removed.
(EXECUTABLE_SUFFIX): Likewise.
* config.in: Regenerated.
* configure: Likewise.
* plugin.c (bfd_plugin_close_and_cleanup): Defined as
_bfd_generic_close_and_cleanup.
(plugin_list_entry): Remove resolution_file, resolution_option,
real_bfd, real_nsyms, real_syms, lto_nsyms, lto_syms, gcc,
lto_wrapper, gcc_env and initialized,
(need_lto_wrapper_p): Removed.
(get_lto_wrapper): Likewise.
(setup_lto_wrapper_env): Likewise.
(register_all_symbols_read): Likewise.
(egister_cleanup): Likewise.
(get_symbols): Likewise.
(add_input_file): Likewise.
(bfd_plugin_set_program_name): Remove need_lto_wrapper.
(add_symbols): Updated.
(try_claim): Likewise.
(try_load_plugin): Likewise.
(bfd_plugin_canonicalize_symtab): Likewise.
* plugin.h (bfd_plugin_set_program_name): Remove int argument.
(plugin_data_struct): Remove real_bfd, real_nsyms and real_syms.

binutils/

* ar.c (main): Update bfd_plugin_set_program_name call.
* nm.c (main): Likewise.

ld/

* testsuite/ld-plugin/lto.exp (lto_link_tests): Run PR ld/25355
test only for GCC 10 or newer.

plugin: Use LDPT_ADD_SYMBOLS_V2 to get symbol type

Since LTO plugin may generate more than one ltrans.o file from one input
IR object as LTO wrapper ignores -flto-partition=none:

lto-wrapper.c:608:

   604          /* Drop arguments that we want to take from the link line.  */
   605          case OPT_flto_:
   606          case OPT_flto:
   607          case OPT_flto_partition_:
   608            continue;

the LTO wrapper approach is not only slow but also unreliable.  Since
the LTO plugin API has been extended to add LDPT_ADD_SYMBOLS_V2 with
symbol type and section kind, we can use LDPT_ADD_SYMBOLS_V2 to get
symbol type, instead of invoking the LTO wrapper.

PR binutils/25640
* plugin.c (plugin_list_entry): Add has_symbol_type.
(add_symbols_v2): New function.
(bfd_plugin_open_input): Don't invoke LTO wrapper if LTO plugin
provides symbol type.
(try_load_plugin): Add LDPT_ADD_SYMBOLS_V2.
(bfd_plugin_canonicalize_symtab): Use LTO plugin symbol type if
available.

Silence warnings due to plugin API change

* testplug.c (parse_symdefstr): Use %hhi to read sym->def, and
clear new fields.
* testplug2.c (parse_symdefstr): Likewise.
* testplug3.c (parse_symdefstr): Likewise.
* testplug4.c (parse_symdefstr): Likewise.

XCOFF uninitialized read

* coff-rs6000.c (_bfd_xcoff_slurp_armap): Ensure size is large
enough to read number of symbols.

metag uninitialized memory read

* metag-dis.c (print_insn_metag): Don't ignore status from
read_memory_func.

NDS32 disassembly of odd sized sections

* nds32-dis.c (print_insn_nds32): Remove unnecessary casts.
Initialize parts of buffer not written when handling a possible
2-byte insn at end of section.  Don't attempt decoding of such
an insn by the 4-byte machinery.

PowerPC disassembly of odd sized sections

We shouldn't really decode a 2-byte left-over at the end of a section
as if the section contains two more bytes of zeros.  Not that it
matters very much, but this patch tidies the corner case.

* ppc-dis.c (print_insn_powerpc): Only clear needed bytes of
partially filled buffer.  Prevent lookup of 4-byte insns when
only VLE 2-byte insns are possible due to section size.  Print
".word" rather than ".long" for 2-byte leftovers.