With this in place we can retire bfd_boolean one project at a time.
bfd/
* bfd-in.h: Include stdbool.h.
(bfd_boolean): Define as bool
* bfd-in2.h: Regenerate.
This fixes the issue that startswith depends on strncpy being
declared, and not all projects using bfd.h include string.h before
bfd.h. I've also deleted some macros that don't find much use
anywhere.
bfd/
* bfd-in.h: Include string.h.
(LITMEMCPY, LITSTRCPY): Delete.
* bfd-in2.h: Regenerate.
binutils/
* prdbg.c (pr_function_type): Replace LITSTTCPY with strcpy.
New in v2:
- Disable sharing only for -readnow objfiles, not all objfiles.
As described in PR 27541, we hit an internal error when loading a binary
the standard way and then loading it with the -readnow option:
$ ./gdb -nx -q --data-directory=data-directory ~/a.out -ex "set confirm off" -ex "file -readnow ~/a.out"
Reading symbols from /home/simark/a.out...
Reading symbols from ~/a.out...
/home/simark/src/binutils-gdb/gdb/dwarf2/read.c:8098: internal-error: void create_all_comp_units(dwarf2_per_objfile*): Assertion `per_objfile->per_bfd->all_comp_units.empty ()' failed.
This is a recurring problem that exposes a design issue in the DWARF
per-BFD sharing feature. Things work well when loading a binary with
the same method (with/without index, with/without readnow) twice in a
row. But they don't work so well when loading a binary with different
methods. See this previous fix, for example:
efb763a5ea ("gdb: check for partial symtab presence in dwarf2_initialize_objfile")
That one handled the case where the first load is normal (uses partial
symbols) and the second load uses an index.
The problem is that when loading an objfile with a method A, we create a
dwarf2_per_bfd and some dwarf2_per_cu_data and initialize them with the
data belonging to that method. When loading another obfile sharing the
same BFD but with a different method B, it's not clear how to re-use the
dwarf2_per_bfd/dwarf2_per_cu_data previously created, because they
contain the data specific to method A.
I think the most sensible fix would be to not share a dwarf2_per_bfd
between two objfiles loaded with different methods. That means that two
objfiles sharing the same BFD and loaded the same way would share a
dwarf2_per_bfd. Two objfiles sharing the same BFD but loaded with
different methods would use two different dwarf2_per_bfd structures.
However, this isn't a trivial change. So to fix the known issue quickly
(including in the gdb 10 branch), this patch just disables all
dwarf2_per_bfd sharing for objfiles using READNOW.
Generalize the gdb.base/index-cache-load-twice.exp test to test all
the possible combinations of loading a file with partial symtabs, index
and readnow. Move it to gdb.dwarf2, since it really exercises features
of the DWARF reader.
gdb/ChangeLog:
PR gdb/27541
* dwarf2/read.c (dwarf2_has_info): Don't share dwarf2_per_bfd
with objfiles using READNOW.
gdb/testsuite/ChangeLog:
PR gdb/27541
* gdb.base/index-cache-load-twice.exp: Remove.
* gdb.base/index-cache-load-twice.c: Remove.
* gdb.dwarf2/per-bfd-sharing.exp: New.
* gdb.dwarf2/per-bfd-sharing.c: New.
Change-Id: I9ffcf1e136f3e75242f70e4e58e4ba1fd3083389
Since commit 27012aba8a "Remove Irix 6 workaround from DWARF abbrev reader"
we have:
...
(gdb) file dw2-cu-size^M
Reading symbols from dw2-cu-size...^M
DW_FORM_strp pointing outside of .debug_str section [in module dw2-cu-size]^M
(No debugging symbols found in dw2-cu-size)^M
(gdb) ptype noloc^M
No symbol table is loaded. Use the "file" command.^M
(gdb) FAIL: gdb.dwarf2/dw2-cu-size.exp: ptype noloc
...
The problem is a missing .debug_abbrev terminator in dw2-cu-size.S, which
causes the .debug_abbrev contribution to be merged with the next:
...
Number TAG (0x9b)
1 DW_TAG_compile_unit [has children]
DW_AT_name DW_FORM_string
DW_AT_producer DW_FORM_string
DW_AT_language DW_FORM_data1
DW_AT value: 0 DW_FORM value: 0
2 DW_TAG_variable [no children]
DW_AT_name DW_FORM_string
DW_AT_type DW_FORM_ref4
DW_AT_external DW_FORM_flag
DW_AT value: 0 DW_FORM value: 0
3 DW_TAG_base_type [no children]
DW_AT_name DW_FORM_string
DW_AT_byte_size DW_FORM_data1
DW_AT_encoding DW_FORM_data1
DW_AT value: 0 DW_FORM value: 0
4 DW_TAG_const_type [no children]
DW_AT_type DW_FORM_ref_udata
DW_AT value: 0 DW_FORM value: 0
1 DW_TAG_compile_unit [has children]
DW_AT_producer DW_FORM_strp
DW_AT_language DW_FORM_data1
DW_AT_name DW_FORM_strp
DW_AT_comp_dir DW_FORM_strp
DW_AT_low_pc DW_FORM_addr
DW_AT_high_pc DW_FORM_data8
DW_AT_stmt_list DW_FORM_sec_offset
DW_AT value: 0 DW_FORM value: 0
...
and consequently, abbreviation code 1 no longer refers to a unique entry.
The eventually causes us to access the first attribute of this DIE:
...
<0><124>: Abbrev Number: 1 (DW_TAG_compile_unit)
<125> DW_AT_name : file1.txt
<12f> DW_AT_producer : GNU C 3.3.3
<13b> DW_AT_language : 1 (ANSI C)
...
which has form DW_FORM_string, using DW_FORM_strp.
Fix this by adding the missing .debug_abbrev terminator in dw2-cu-size.S.
gdb/testsuite/ChangeLog:
2021-03-30 Tom de Vries <tdevries@suse.de>
PR testsuite/27604
* gdb.dwarf2/dw2-cu-size.S: Add missing .debug_abbrev terminator.
During reviews, I changed the success/failure variables from int to bool, but
missed updating the code in a couple spots. Given the logic inversion, the
gdbserver code fails instead of succeeding.
Fixed with the following patch. Seems fairly obvious, so I'll push it soon.
gdbserver/ChangeLog:
2021-03-30 Luis Machado <luis.machado@linaro.org>
* server.cc (handle_general_set, handle_query): Update variable
to bool and fix verification logic.
The former two are unused anyway. And having such constants isn't very
helpful either, when they live in a place where updating the register
table wouldn't even allow noticing the need to adjust these constants.
st(1) ... st(7) will never be looked up in the hash table, so there's no
point inserting the entries. It's also not really necessary to do a 2nd
hash lookup after parsing the register number, nor is there a real
reason for having both st and st(0) entries. Plus we can easily do away
with the need for st to be first in the table.
There's no need for the extra level of indirection and the extra storage
needed for the pointer, pointing from one piece of static data to
another. Key checking of rounding being in effect off of the type field
of the structure instead.
There's no need for the extra level of indirection and the extra storage
needed for the pointer, pointing from one piece of static data to
another. Key checking of broadcast being in effect off of the type field
of the structure instead.
There's no need for the extra level of indirection and the extra storage
needed for the pointer, pointing from one piece of static data to
another. Key checking of masking being in effect off of the register
field of the structure instead.
All callers pass unsigned values (in some cases by virtue of passing
non-negative literal numbers).
This in turn requires struct {Mask,RC,Broadcast}_Operation's "operand"
fields to become unsigned, in turn allowing to reduce the amount of
casting needed (the two new casts that are necessary cast _to_
"unsigned" instead of _from_, as that's the form that'll never case
undefined behavior).
Seen after converting bfd_boolean to bool.
mmix +FAIL: ld-mmix/zeroehmmo
./ld-new -L/home/alan/src/binutils-gdb/ld/testsuite/ld-mmix -m mmo -Ttext 0xa00 -T /home/alan/src/binutils-gdb/ld/testsuite/ld-mmix/zeroeh.ld -o tmpdir/dump tmpdir/x.o tmpdir/y.o
/home/alan/src/binutils-gdb/bfd/linker.c:2294:8: runtime error: load of value 253, which is not a valid value for type '_Bool'
* elflink.c (elf_link_add_object_symbols): Don't set h->indx
unless is_elf_hash_table.
This patch supports linking powerpc64 glibc with gold, specifically
the __tls_get_addr call in elf/dl-sym.c. That call lacks marker
relocations tying it to the arg setup instructions, but the arg setup
insns are also contructed lacking the usual relocations on a Global
Dynamic TLS code sequence. So there is no chance that anything in
that sequence might be wrongly edited by the linker.
In fact, the aim of linking glibc could have been supported by simply
omitting the error whenever TLS optimisation is disabled, as it is
when linking a shared library. The patch goes further than that,
disabling TLS GD and LD sequence optimisation on a per-object basis
for object files lacking marker relocs.
PR gold/27625
* powerpc.cc (Powerpc_relobj): Add no_tls_marker_, tls_marker_,
and tls_opt_error_ variables and accessors.
(Target_powerpc::Scan::local, global): Call set_tls_marker and
set_no_tls_marker for GD and LD code sequence relocations.
(Target_powerpc::Relocate::relocate): Downgrade the "lacks marker
reloc" error to a warning when safe to do so, and omit the error
entirely if not optimising TLS sequences. Do not optimise GD and
LD sequences for objects lacking marker relocs.
(Target_powerpc::relocate_relocs): Heed no_tls_marker here too.
I noticed that language_info is only ever called with a value of '1'.
This patch removes the parameter.
2021-03-29 Tom Tromey <tromey@adacore.com>
* top.c (check_frame_language_change): Update.
* language.c (language_info): Remove parameter.
* language.h (language_info): Remove parameter.
On aarch64-linux, I noticed the compile command didn't work at all. It
always gave the following error:
aarch64-linux-gnu-g++: error: : No such file or directory
Turns out we're passing an empty argv entry to GCC (because aarch64 doesn't
have a -m64 option), and GCC's behavior is to think that is a file it needs
to open. One can reproduce it like so:
gcc "" "" "" ""
gcc: error: : No such file or directory
gcc: error: : No such file or directory
gcc: error: : No such file or directory
gcc: error: : No such file or directory
gcc: fatal error: no input files
compilation terminated.
The solution is to check for an empty string and skip adding that to argv.
Regression tested on aarch64-linux/Ubuntu 18.04/20.04.
gdb/ChangeLog:
2021-03-29 Luis Machado <luis.machado@linaro.org>
* compile/compile.c (get_args): Don't add empty argv entries.
It was reported to me that on Ubuntu 14.04 (fairly old) the documentation
fails to build with the following:
gdb/doc/gdb.texinfo:10888: warning: node `Memory' is up for `Memory Tagging' in sectioning but not in menu
gdb/doc/gdb.texinfo:10693: node `Memory' lacks menu item for `Memory Tagging' despite being its Up target
Makefile:491: recipe for target 'gdb.info' failed
make[3]: *** [gdb.info] Error 1
This doesn't seem to happen on Ubuntu 18.04/20.04, but it does make sense.
Fix this by turning @subsection into a @section and adding the
"Memory Tagging" entry to the menu.
gdb/doc/ChangeLog:
2021-03-29 Luis Machado <luis.machado@linaro.org>
* gdb.textinfo (Memory Tagging): Make it a @section.
This test causes several timeouts for Clang, taking too long time to
finish. The reason is, for an infinite loop of the form
while(1); /* suppose this is line 30. */
Clang generates code that looks like
0x00000000004004d4 <+4>: jmp 0x4004d9 <loop+9>
0x00000000004004d9 <+9>: jmp 0x4004d9 <loop+9>
So, the real loop is the instruction at address 0x4004d9. But a
breakpoint that's defined at the loop line (assume line 30 in this
case) is inserted at address 0x4004d4.
(gdb) break 30
Breakpoint 1 at 0x4004d4: file test.c, line 30.
Therefore, continuing a thread that was spinning on the loop does not hit
the breakpoint. The bug is reported at
https://bugs.llvm.org/show_bug.cgi?id=49614
Tweak the infinite loop to spin on a variable to avoid this bug. The
test is unrelated to the bug.
gdb/testsuite/ChangeLog:
2021-03-29 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
* gdb.mi/user-selected-context-sync.exp: Spin on a variable in
the infinite loop to avoid a Clang bug.
Since c8fbd44a01 (gdb: remove
target_is_pushed free function), procfs.c compilation is broken, which
went unnoticed for lack of functioning buildbots:
/vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c: In member function 'virtual void procfs_target::attach(const char*, int)':
/vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:1772:8: error: 'inf' was not declared in this scope; did you mean 'info'?
1772 | if (!inf->target_is_pushed (this))
| ^~~
| info
/vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c: In member function 'virtual void procfs_target::create_inferior(const char*, const string&, char**, int)':
/vol/src/gnu/gdb/hg/master/dist/gdb/procfs.c:2865:8: error: 'inf' was not declared in this scope; did you mean 'info'?
2865 | if (!inf->target_is_pushed (this))
| ^~~
| info
Fixed by defining inf. Tested on amd64-pc-solaris2.11 and
sparcv9-sun-solaris2.11.
2021-03-29 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
gdb:
* procfs.c (procfs_target::attach): Define inf.
Use it.
(procfs_target::create_inferior): Likewise.
For a long time there hasn't been a need anymore to keep together all
templates with identical mnemonics. Move the MOVQ and MOVABS ones next
to their MOV counterparts. Move the string forms of CMPSD and MOVSD next
to their CMPS / MOVS counterparts. Re-arrange what so fgar was the SSE3
section.
This makes reasonably obvious that MONITOR/MWAIT aren't suitable to
cover by CpuSSE3, but adjusting this is left for another time.
In commit 79dec6b7ba ("x86-64: optimize certain commutative
VEX-encoded insns") I missed the fact that there being subtraction
involved here doesn't matter, as absolute differences get summed up.
This way not only the overall (source) table size shrinks by quite a
bit and the risk of related templates going out of sync with one another
gets lowered, but also (dis)similarities between neighboring templates
become easier to spot.
Note that for certain SSE2AVX templates this results in benign attribute
changes:
- LDMXCSR and STMXCSR: NoAVX gets set,
- MOVMSKPS, PMOVMSKB, PEXTR{B,W} (register destination), and PINSR{B,W}
(register source): IgnoreSize and NoRex64 get set,
- CVT{DQ,PS}2PD, CVTSD2SS, MOVMSKPD, MOVDDUP, PMOV{S,Z}X{BW,WD,DQ}, and
ROUNDSD: NoRex64 gets set,
- CVTSS2SD, INSERTPS, PEXTRW (memory destination), PINSRW (memory
source), and PMOV{S,Z}X{BD,WQ,BQ}: IgnoreSize gets set.
Similarly the "normal" (non-SSE2AVX)
- non-64-bit CVTS{I,S}2SD forms get NoRex64 set,
- CMP{EQ,ORD,NEQ,UNORD}{P,S}{S,D} forms get C set,
all again in a benign way.
The remaining differences in the generated table are due to re-ordering
of entries in the course of being folded into templates.
The table entries are more natural to read (and slightly shorter) when
the prefixes, like is the case for VEX/XOP/EVEX-encoded entries, are
specified as part of the opcode. This is particularly noticable for
side-by-side legacy and SSE2AVX entries.
An implication is that we now need to use "unsigned long long" for the
initially parsed opcode in i386-gen. I don't expect this to be an issue.
Now that all base opcodes are only at most 2 bytes in size, shrink its
template field to just as much. By also shrinking extension_opcode and
operands to just what they really need, we can free up an entire 32-bit
slot (plus 4 left bits past the bitfields themselves).
At present this alters sizeof(struct insn_template) only for 32-bit
builds. In 64-bit builds it instead leaves a padding hole that will
allow to buffer future growth of other fields (opcode_modifier,
cpu_flags, operand_types[]).
Just like is already done for VEX/XOP/EVEX encoded insns, record the
encoding space information in the respective opcode modifier field. Do
this again without changing the source table, but rather by deriving the
values from their existing source representation.
* objdump.c (process_links): Use type int.
* readelf.c (request_dump): Don't increment do_dump, set it.
* windint.h (target_is_bigendian): Use type bfd_boolean.
* windmc.c (target_is_bigendian): Likewise.
* windres.c (target_is_bigendian): Likewise.
elf_backend_link_output_symbol_hook and elf_link_output_symstrtab may
return 2 when a symbol is to be discarded. Update places that use
bfd_boolean rather than int for these functions.
* elflink.c (elf_link_output_symstrtab): Make flinfo parameter
a void pointer.
(bfd_elf_final_link): Delete out_sym_func typedef and don't cast
elf_link_output_symstrtab when calling output_arch_syms and
output_arch_local_syms.
* elf-bfd.h (struct elf_backend_data <elf_backend_output_arch_syms,
elf_backend_output_arch_local_syms>): Change return type of func
arg to match elf_link_output_symstrtab.
* elf-vxworks.h (elf_vxworks_link_output_symbol_hook): Correct
return type.
* elf32-nds32.c (nds32_elf_output_symbol_hook): Correct return type.
(nds32_elf_output_arch_syms): Correct func return type.
Now that the quick functions are separate from the object file format,
there's no need to have elfread.c push a new entry on the objfile 'qf'
list. Instead, this detail can be pushed into the DWARF reader. That
is what this patch implements.
I wasn't sure whether lazy reading still makes sense or not. It's
still only used by ELF, and only in certain situations (like vfork, I
think). It may not be carrying its weight, so we may want to consider
removing this in the future.
Also, I'm unclear on why the various indices are only used for ELF.
This seems sub-optimal. However, I haven't tried to address that
here.
gdb/ChangeLog
2021-03-28 Tom Tromey <tom@tromey.com>
* elfread.c (can_lazily_read_symbols): Move to dwarf2/read.c.
(elf_symfile_read): Simplify.
* dwarf2/read.c (struct lazy_dwarf_reader): Move from elfread.c.
(make_lazy_dwarf_reader): New function.
(make_dwarf_gdb_index, make_dwarf_debug_names): Now static.
(dwarf2_initialize_objfile): Return void. Remove index_kind
parameter. Push on 'qf' list.
* dwarf2/public.h (dwarf2_initialize_objfile): Change return
type. Remove 'index_kind' parameter.
(make_dwarf_gdb_index, make_dwarf_debug_names): Don't declare.
An earlier patch neglected to delete a forward declaration of
elf_sym_fns_lazy_psyms. This is no longer defined. This patch
removes it.
gdb/ChangeLog
2021-03-27 Tom Tromey <tom@tromey.com>
* elfread.c (elf_sym_fns_lazy_psyms): Don't declare.
I noticed that I forgot to make a change in my series to make it
possible to attach multiple debug readers to an objfile. In one spot,
elf_symfile_read still clears the 'qf' list. However, this should
have been removed toward the end of that series.
This patch fixes the offending spot. Tested on x86-64 Fedora 32.
gdb/ChangeLog
2021-03-27 Tom Tromey <tom@tromey.com>
* elfread.c (elf_symfile_read): Don't clear 'qf'.
Resolve some duplicate test name warnings in gdb.arch/powerpc-*.exp
tests by either extending the existing test names, or providing a new
test name.
gdb/testsuite/ChangeLog:
* gdb.arch/powerpc-disassembler-options.exp: Extend some test
names for uniqueness.
* gdb.arch/powerpc-fpscr-gcore.exp: Add more test names for
uniqueness.
While working on gdb-add-index.sh, it appeared that it uses the non
POSIX 'local' keyword. Instead of using local to allow variable
shadowing, I rename the local one to avoid name conflicts altogether.
This commit gets rid of the following shellcheck warning:
In gdb-add-index.sh line 63:
local file="$1"
^--------^ SC2039: In POSIX sh, 'local' is undefined.
gdb/ChangeLog:
* contrib/gdb-add-index.sh: Avoid variable shadowing and get
rid of 'local'.
This changes quick_symbol_functions::map_symbol_filenames to use a
function_view, and updates all the uses. It also changes the final
parameter to 'bool'. A couple of spots are further updated to use
operator() rather than a lambda.
gdb/ChangeLog
2021-03-26 Tom Tromey <tom@tromey.com>
* symtab.c (struct output_source_filename_data): Add 'output'
method and operator().
(output_source_filename_data::output): Rename from
output_source_filename.
(output_partial_symbol_filename): Remove.
(info_sources_command): Update.
(struct add_partial_filename_data): Add operator().
(add_partial_filename_data::operator()): Rename from
maybe_add_partial_symtab_filename.
(make_source_files_completion_list): Update.
* symfile.c (quick_symbol_functions): Update.
* symfile-debug.c (objfile::map_symbol_filenames): Update.
* quick-symbol.h (symbol_filename_ftype): Change type of 'fun' and
'need_fullname'. Remove 'data' parameter.
(struct quick_symbol_functions) <map_symbol_filenames>: Likewise.
* psymtab.c (psymbol_functions::map_symbol_filenames): Update.
* psympriv.h (struct psymbol_functions) <map_symbol_filenames>:
Change type of 'fun' and 'need_fullname'. Remove 'data'
parameter.
* objfiles.h (struct objfile) <map_symbol_filenames>: Change type
of 'fun' and 'need_fullname'. Remove 'data' parameter.
* mi/mi-cmd-file.c (print_partial_file_name): Remove 'ignore'
parameter.
(mi_cmd_file_list_exec_source_files): Update.
* dwarf2/read.c
(dwarf2_base_index_functions::map_symbol_filenames): Update.
I noticed that ada-lang.c creates a lambda to call
aux_add_nonlocal_symbols. However, this code can be simplified a bit
by changing match_data to implement operator(), and then simply
passing the object as the callback. That is what this patch
implements.
gdb/ChangeLog
2021-03-26 Tom Tromey <tom@tromey.com>
* ada-lang.c (struct match_data): Add operator().
(match_data::operator()): Rename from aux_add_nonlocal_symbols.
(callback): Remove 'callback'.
I noticed that psymbol_functions::expand_symtabs_matching calls
make_ignore_params once per psymtab that is matched. This seems
possibly expensive, so this patch hoists the call out of the loop.
gdb/ChangeLog
2021-03-26 Tom Tromey <tom@tromey.com>
* psymtab.c (psymbol_functions::expand_symtabs_matching): Only
call make_ignore_params once.
Currently the psymtab variant of expand_symtabs_matching has this
check:
/* We skip shared psymtabs because file-matching doesn't apply
to them; but we search them later in the loop. */
if (ps->user != NULL)
continue;
In a larger series I'm working on, it's convenient to remove this
check. And, I noticed that a similar check is not done for
expand_symtabs_with_fullname. So, it made sense to me to remove the
check here as well.
gdb/ChangeLog
2021-03-26 Tom Tromey <tom@tromey.com>
* psymtab.c (psymbol_functions::expand_symtabs_matching): Remove
"user" check.
