There is code in bfd/elf-eh-frame.c and ld/emultempl/elf32.em that
checks for the presence of eh_frame info by testing for a section
named .eh_frame sized more than 8 bytes. The size test is to exclude
a zero terminator. A similar check in elf64-ppc.c wrongly just tested
for non-zero size before creating the linker generated .eh_frame
describing plt call and other linkage stubs. The intention was to not
generate that info unless there was some user .eh_frame. (No user
.eh_frame implies the user doesn't care about exception handling.)
Because the test in elf64-ppc.c was wrong, ld generated the stub
.eh_frame just on finding a zero .eh_frame terminator in crtend.o, but
didn't generate the corresponding .eh_frame_hdr.
* elf64-ppc.c (ppc64_elf_size_stubs): Correct test for user
.eh_frame info.
Test names should not end with parentheses, since the buildbot strips
those.
gdb/testsuite/ChangeLog:
* gdb.opt/inline-locals.exp: Remove trailing parentheses from
test names.
Commit dd90581873 ("Place .shstrtab section after .symtab and .strtab,
thus restoring monotonically incre... ") adjusted section numbers, but
forgot to adjust sh_link references from relocation and group section
table entries.
Additionally some other (perhaps subsequent) change appears to have
added .rel.* and .rela.* sections to their respective groups, which
requires some further adjustments to group-2.d. I assume this additional
breakage wasn't noticed because the test was already failing at that
time.
This makes the gas testsuite complete successfully again for me in a
cross build on ix86-linux; there continue to be quite a few ld failures.
Linking EFI executables from ELF object files can result in corrupted
COFF section flags if the section's alignment is too high. Issue a
diagnostic in that case, erroring out if this is not a final link, and
make sure only in-range values get written to the output image.
While doing this also make tic80 use the generic alignment macros
instead of custom #ifdef-ary.
No testsuite regressions for the range of COFF/PE targets that actually
cross-build in the first place on x86-64-linux.
Just like is the case for xsave{s,c}64 and xrstors64 already. I wonder
though why xsave{s,c} and xrstors don't allow for the q suffix, other
than the other insns without the "64" suffix do.
... rather than silently dropping it altogether.
i386_finalize_displacement() expects baseindex to already be set, so
the respective statement needs to be moved up. This then also allows a
subsequent conditional to be simplified.
For this to not regress on 32-bit addressing, break out address size
guessing from i386_index_check(), invoking the new function earlier so
that i386_finalize_displacement() has i.prefix[ADDR_PREFIX] available.
i386_addressing_mode () in turn needs i.base_reg / i.index_reg set
earlier.
Extract GNU_RELRO region check into a new funtion and pass a pointer to
seg_align_type to lang_find_relro_sections_1 so that they can also be
used for text-only LOAD segment.
* ldlang.c (lang_size_sections_1): Extract GNU_RELRO region check
into ...
(ldlang_check_relro_region): New function.
(lang_find_relro_sections_1): Add an argument for pointer to
seg_align_type and replace expld.dataseg with the pointer.
(lang_find_relro_sections): Pass address of expld.dataseg to
lang_find_relro_sections_1.
Break lang_size_sections down into separate functions so that they can
also be used for text-only LOAD segment. lang_size_relro_segment will
call lang_size_relro_segment_1 and lang_size_segment for both GNU_RELRO
segment and text-only LOAD segment.
* ldlang.c (lang_size_segment): New function.
(lang_size_relro_segment_1): Likewise.
(lang_size_relro_segment): Likewise.
(lang_size_sections): Rewrite to call lang_size_relro_segment.
Extract DATA_SEGMENT_END/DATA_SEGMENT_ALIGN/DATA_SEGMENT_RELRO_END cases
for GNU_RELRO segment into separate functions so that they can also be
used for text-only LOAD segment.
* ldexp.c (fold_unary): Extract the DATA_SEGMENT_END case to ...
(fold_segment_end): New function.
(fold_binary): Extract the DATA_SEGMENT_ALIGN case to ...
(fold_segment_align): New function.
(fold_binary): Extract the DATA_SEGMENT_RELRO_END case to ...
(fold_segment_relro_end): New function.
Update GNU_RELRO date type so that they can also be used for text-only
LOAD segment.
* ldexp.h (phase_enum): Rename exp_dataseg_none,
exp_dataseg_align_seen, exp_dataseg_relro_seen,
exp_dataseg_end_seen, exp_dataseg_relro_adjust,
exp_dataseg_adjust and exp_dataseg_done to exp_seg_none,
exp_seg_align_seen, exp_seg_relro_seen, exp_seg_end_seen,
exp_seg_relro_adjust, exp_seg_adjust and exp_seg_done.
(relro_enum): Rename exp_dataseg_relro_none,
exp_dataseg_relro_start and exp_dataseg_relro_end to
exp_seg_relro_none, exp_seg_relro_start and exp_seg_relro_end.
(seg_align_type): New struct type.
(ldexp_control): Use seg_align_type.
* ldexp.c (fold_unary): Updated.
(fold_binary): Likewise.
* ldlang.c (strip_excluded_output_sections): Likewise.
(lang_size_sections_1): Likewise.
(lang_size_sections): Likewise.
It was reasonable to mark PT_PHDR segment with PF_X for compatibility
with UnixWare and Solaris linkers 20 years ago. But it is inappropriate
today when the primary OS of GNU ld is Linux. This patch removes PF_X
from PT_PHDR segment as gold does.
Tested natively on Linux/x86 as well as crosss-binutils for alpha-linux,
ia64-linux, powerpc64-linux, powerpc-linux, s390-linux, s390x-linux,
sparc64-linux and sparc-linux.
bfd/
PR ld/22423
* elf.c (_bfd_elf_map_sections_to_segments): Remove PF_X from
PT_PHDR segment.
ld/
PR ld/22423
* testsuite/ld-alpha/tlsbin.rd: Replace "R E " with "R +" for
PT_PHDR segment.
* testsuite/ld-alpha/tlsbinr.rd: Likewise.
* testsuite/ld-ia64/tlsbin.rd: Likewise.
* testsuite/ld-powerpc/tlsexe.r: Likewise.
* testsuite/ld-powerpc/tlsexe32.r: Likewise.
* testsuite/ld-powerpc/tlsexetoc.r: Likewise.
* testsuite/ld-s390/tlsbin.rd: Likewise.
* testsuite/ld-s390/tlsbin_64.rd: Likewise.
* testsuite/ld-sparc/tlssunbin32.rd: Likewise.
* testsuite/ld-sparc/tlssunbin64.rd: Likewise.
* testsuite/ld-elf/pr22423.d: New test.
OTHER_PLT_SECTIONS contains sections which should be placed right after
.plt section.
* emulparams/elf32_x86_64.sh (TINY_READONLY_SECTION): Renamed
to ...
(OTHER_PLT_SECTIONS): This.
* emulparams/elf_i386.sh: Likewise.
* emulparams/elf_iamcu.sh: Likewise.
* emulparams/elf_x86_64.sh: Likewise.
* scripttempl/elf.sc: Place ${OTHER_PLT_SECTIONS} just after
.plt.
The existing code allowed common and indirect symbols to pass an
assert, but then read root.u.def which is not valid for anything
besides defined and weakdef symbols. The garbage root.u.def.section
pointer read can't possibly match pointers stored at the same location
for indirect and common symbols, so we would always have left
u.weakdef NULL.
* elflink.c (elf_link_add_object_symbols): Ignore anything but
defined and defweak symbols when setting up weakdefs.
This patch was aimed at a FIXME in elf32-hppa.c, the ludicrous and
confusing fact that non_got_ref after adjust_dynamic_relocs in that
backend means precisely the inverse of what it means before
adjust_dynamic_relocs. Before, when non_got_ref is set it means there
are dynamic relocs, after, if non_got_ref is clear it means "keep
dynamic relocs" and later, "has dynamic relocs". There is a reason
why it was done that way.. Some symbols that may have dynamic
relocations pre-allocated in check_relocs turn out to not be dynamic,
and then are not seen by the backend adjust_dynamic_symbols. We want
those symbols to lose their dynamic relocs when non-pic, so it's handy
that non_got_ref means the opposite after adjust_dynamic_relocs. But
it's really confusing.
Most other targets, like ppc32, don't always set non_got_ref on
non-GOT references that have dynamic relocations. This is because the
primary purpose of non_got_ref before adjust_dynamic_relocs is to flag
symbols that might need to be copied to .dynbss, and there are
relocation types that may require dyn_relocs but clearly cannot have
symbols copied into .dynbss, for example, TLS relocations.
Why do we need a flag after adjust_dynamic_relocs to say "keep
dynamic relocations"? Well, you can discard most unwanted dyn_relocs
in the backend adjust_dynamic_relocs, and for those symbols that
aren't seen by the backend adjust_dynamic_relocs, in
allocate_dynrelocs based on a flag set by adjust_dynamic relocs,
dynamic_adjusted. That doesn't solve all our difficulties though.
relocate_section needs to know whether a symbol has dyn_relocs, and
many targets transfer dyn_relocs to a weakdef if the symbol has one.
The transfer means relocate_section can't test dyn_relocs itself and
the weakdef field has been overwritten by that time. So non_got_ref
is used to flag "this symbol has dynamic relocations" for
relocate_section.
Confused still? Well, let's hope the comments I've added help clarify
things.. The patch also fixes a case where we might wrongly emit
dynamic relocations in an executable for common and undefined symbols.
* elf32-hppa.c (elf32_hppa_adjust_dynamic_symbol): Set non_got_ref
to keep dyn_relocs, clear to discard. Comment.
(allocate_dynrelocs): Always clear non_got_ref when clearing
dyn_relocs in non-pic case. Invert non_got_ref test. Also test
dynamic_adjusted and ELF_COMMON_DEF_P. Move code deleting
dyn_relocs on undefined syms to handle for non-pic too.
(elf32_hppa_relocate_section): Simplify test for non-pic dyn relocs.
* elf32-ppc.c (ppc_elf_adjust_dynamic_symbol): Set non_got_ref
to keep dyn_relocs, clear to discard. Comment.
(allocate_dynrelocs): Always clear non_got_ref when clearing
dyn_relocs in non-pic case. Invert non_got_ref test. Also test
dynamic_adjusted and ELF_COMMON_DEF_P. Move code deleting
dyn_relocs on undefined syms to handle for non-pic too.
(ppc_elf_relocate_section): Simplify test for non-pic dyn relocs.
* elf64-ppc.c (ppc64_elf_adjust_dynamic_symbol): Discard
dyn_relocs here. Don't bother setting non_got_ref. Comment.
(allocate_dynrelocs): Delete special handling of non-pic ELFv2
ifuncs. Move code deleting dyn_relocs on undefined symbols to
handle for non-pic too. Don't test non_got_ref. Do test
dynamic_adjusted and ELF_COMMON_DEF_P.
'make check-read1 TESTS="gdb.tui/tui-completion.exp"' exposes this test race:
(gdb) PASS: gdb.tui/completion.exp: set max-completions unlimited
layout ^G
asm next prev regs split src
(gdb) FAIL: gdb.tui/completion.exp: completion of layout names: tab completion
Quit
(gdb) PASS: gdb.tui/completion.exp: completion of layout names: quit command input
focus ^G
cmd next prev src
(gdb) FAIL: gdb.tui/completion.exp: completion of focus command: tab completion
Quit
This is caused by expecting "$gdb_prompt layout $".
gdb_test_multiple's internal prompt regexp can match first if expect's
internal buffer is filled with partial output. Fix that by splitting
the gdb_test_multiple in question in two. Since the same problem/code
appears twice in the file, factor out a common procedure.
gdb/testsuite/ChangeLog:
2017-11-09 Pedro Alves <palves@redhat.com>
* gdb.tui/tui-completion.exp (test_tab_completion): New procedure,
factored out from ...
(top level): ... here, and adjusted to avoid expecting beyond the
prompt in a single gdb_test_multiple.
This commit fixes this same problem in several places:
(gdb) PASS: gdb.multi/multi-attach.exp: backtrace 2
kill
Kill the program being debugged? (y or n) y
(gdb) FAIL: gdb.multi/multi-attach.exp: kill inferior 2 (got interactive prompt)
This is just another case of the gdb_test_multiple's internal "got
interactive prompt" pattern matching because the testcase misses
matching enough.
gdb/testsuite/ChangeLog:
2017-11-09 Pedro Alves <palves@redhat.com>
* gdb.multi/multi-attach.exp ("kill" test): Match the whole query
output.
* gdb.server/ext-attach.exp ("kill" test): Likewise.
* gdb.server/ext-restart.exp ("kill" test): Likewise.
* gdb.server/ext-run.exp ("kill" test): Likewise.
* gdb.server/ext-wrapper.exp ("kill" test): Likewise.
With:
$ make check-read1 TESTS="gdb.cp/cpcompletion.exp"
we get (from gdb.log):
(gdb) complete break Foo::
break Foo::Foo()
break Foo::Foofoo()
break Foo::get_foo()
break Foo::set_foo(int)
break Foo::~Foo()
(gdb) FAIL: gdb.cp/cpcompletion.exp: complete class methods (Foo not found)
The problem is that the
"break ${class}::\[A-Za-z0-9_~\]+"
regexp patches partial input, like:
break Foo::F
break Foo::Fo
break Foo::Foo
etc.
Fix that by expecting each whole line.
gdb/testsuite/ChangeLog:
2017-11-09 Pedro Alves <palves@redhat.com>
* gdb.cp/cpcompletion.exp (test_class_complete): Tighten regex to
match till end of line.
Testing with:
$ make check-read1 TESTS="gdb.base/memattr.exp"
Exposes a testcase bug that can result in racy fails:
info mem
Using user-defined memory regions.
Num Enb Low Addr High Addr Attrs
1 y 0x0000000000601060 0x0000000000601160 wo nocache
2 y 0x0000000000601180 0x0000000000601280 ro nocache
4 y 0x0000000000601280 0x0000000000601380 rw nocache
3 y 0x0000000000601380 0x0000000000601480 rw nocache
5 y 0x0000000000601480 0x0000000000601580 rw nocache
(gdb) FAIL: gdb.base/memattr.exp: info mem (1)
The problem is that:
"Attrs\[^\n\r]*.."
matches:
"Attrs \r"
when the output buffer is filled with partial output like this:
"info mem\r\nUsing user-defined memory regions.\r\nNum Enb Low Addr High Addr Attrs \r"
gdb/testsuite/ChangeLog:
2017-11-09 Pedro Alves <palves@redhat.com>
* gdb.base/memattr.exp: Tighten regexes to match the end line.
Testing with:
$ make check-read1 TESTS="gdb.base/completion.exp"
Exposes a testcase bug that can result in racy fails:
FAIL: gdb.base/completion.exp: command-name completion limiting using tab character
ERROR: Undefined command "".
FAIL: gdb.base/completion.exp: symbol-name completion limiting using tab character
FAIL: gdb.base/completion.exp: symbol-name completion limiting using complete command
testsuite/gdb.log shows:
(gdb) PASS: gdb.base/completion.exp: set max-completions 5
p^G
passcount path print print-object printf
*** List may be truncated, max-completions reached. ***
(gdb) FAIL: gdb.base/completion.exp: command-name completion limiting using tab character
pcomplete p
Undefined command: "pcomplete". Try "help".
(gdb) ERROR: Undefined command "".
The problem is that the expect buffer can get filled with partial
output that ends in the gdb prompt, and so the default FAIL inside
gdb_test_multiple matches.
Fix that by splitting the gdb_test_multiple in two stages. Since that
is done in more than one place in the testcase, move the otherwise
duplicate code to helper procedures.
gdb/testsuite/ChangeLog:
2017-11-09 Pedro Alves <palves@redhat.com>
* gdb.base/completion.exp (ignore_and_resync, test_tab_complete):
New procedures, factored out from ...
(top level): ... here, and adjusted to avoid expecting beyond the
prompt in one go.
Testing with:
$ make check-read1 TESTS="gdb.asm/asm-source.exp"
Exposes a testcase bug that can result in racy fails:
(gdb) PASS: gdb.asm/asm-source.exp: next over foo3
return
Make selected stack frame return now? (y or n) y
n
#0 main () at /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.asm/asmsrc1.s:53
53 gdbasm_exit0
(gdb) FAIL: gdb.asm/asm-source.exp: return from foo2 (got interactive prompt)
n
The problem is that the "return now\?.*" regex can match partial
output like this:
"Make selected stack frame return no"
and then we send the 'y' too early, and then the next time around we
hit gdb_test_multiple's internal "got interactive prompt" regex.
Also, note we match "return no" instead of "return now" because the
regex is missing one quote level.
gdb/testsuite/ChangeLog:
2017-11-09 Pedro Alves <palves@redhat.com>
* gdb.asm/asm-source.exp ("kill" test): Match the whole query
output. Fix '?' match.
gdb/ChangeLog:
* ada-lang.c: Fix some typos in the general command documenting
how Ada expressions are being evaluated and how their result
is printed.
This patch simplifies the psymbol_hash function, by changing it not to
examine the contents of the symbol's name. This change just mirrors
what psymbol_compare already does -- it is checking for name equality,
which is ok because symbol names are interned in symbol_set_names.
This change speeds up psymbol reading. "gdb -nx -batch gdb"
previously took ~1.8 seconds on my machine, and with this patch it now
takes ~1.7 seconds.
gdb/ChangeLog
2017-11-09 Tom Tromey <tom@tromey.com>
* psymtab.c (psymbol_hash): Do not hash string contents.
(psymbol_compare): Add comment.
This speeds up dict_hash a bit, by moving the "TKB" check into the
switch in the loop.
For "gdb -nx -readnow -batch gdb", this improves the time from ~9.8s
before to ~8.5s afterward.
gdb/ChangeLog
2017-11-09 Tom Tromey <tom@tromey.com>
* dictionary.c (dict_hash): Move "TKB" check into the "switch".
When debugging an Ada program, and inserting a watchpoint tracking
a local variable, the watchpoint doesn't get automatically deleted
upon leaving that variable's scope. This watchpoint then starts
creating problems later on, when trying to resume the program's
execution from a location outside of the watchpoint's scope:
(gdb) c
Continuing.
Breakpoint 2, foo_p708_025 () at foo_p708_025.adb:7
7 Do_Nothing (Val);
(gdb) n
No frame is currently executing in block pck.get_val.
Command aborted.
(gdb) c
Continuing.
No frame is currently executing in block pck.get_val.
Command aborted.
The expected output is the following:
- The program's execution after the first continue should stop
as soon as we reach the end of the watchpoint's scope, and
the debugger should be deleting it.
- Then we can continue until reaching breakpoint 2 above;
- After which we should be able to do next/continue as usual.
The reason the watchpoint is not automatically deleted at scope exit
is because the watchpoint is not marked as being scope-specific
(b->exp_valid_block is equal NULL), and this is because the
symbol lookup for our local variable failed to set the innermost_block
global variable during the lookup.
More precisely, if we look at watch_command_1, we do the following:
innermost_block = NULL;
[...]
exp = parse_exp_1 (&arg, 0, 0, 0);
[...]
exp_valid_block = innermost_block;
Currently, innermost_block stays NULL after the call to parse_exp_1.
Digging further, this innermost_block is typically set during symbol
lookup when the symbol is considered to have a frame-relative address.
For instance, in c-exp.y, we see some code like the following:
if (symbol_read_needs_frame (sym.symbol))
{
if (innermost_block == 0
|| contained_in (sym.block,
innermost_block))
innermost_block = sym.block;
}
We actually have the exact same mechanism in ada-exp.y, except
that it vhas accidently been turned off. See write_var_from_sym,
where we start with:
if (orig_left_context == NULL && symbol_read_needs_frame (sym))
{
if (innermost_block == 0
|| contained_in (block, innermost_block))
innermost_block = block;
}
In this case, orig_left_context is a parameter, and looking at
the point of call in write_var_or_type, we see:
if (nsyms == 1)
{
write_var_from_sym (par_state, block, syms[0].block,
syms[0].symbol);
In the call above, the paramater we are interested in is "block",
which is a parameter for write_var_or_type as well, except we
explicitly override its value at the beginning when found to be NULL:
if (block == NULL)
block = expression_context_block;
So the block we pass to write_var_from_sym is not NULL, and
we therefore don't set innermost_block, which leads to the watchpoint
no longer being marked as scope-specific.
The handling of orig_left_context in write_var_from_sym was there
to handle the case where a user writes an expression where the symbol
is qualified with a scope (Eg: "function::variable"). But it appears
that handling this is specifically here is no longer necessary,
so this patch simply removes that parameter and the associated check,
and then updates all the points of calls.
Interestingly, this also affects GDB/MI, and in particular varobjs,
because local variables are now properly reported as having a block,
which causes the associated varob to have a "thread-id" field.
This patch also adjusts a couple of Ada/gdb-mi tests.
gdb/ChangeLog:
* ada-exp.y (write_var_from_sym): Remove parameter
"orig_left_context". Update all callers.
gdb/testsuite/ChangeLog:
* gdb.ada/scoped_watch: New testcase.
* gdb.ada/watch_arg.exp: Adjust expected behavior to the behavior
which is actually correct.
* gdb.ada/mi_interface.exp: Add missing thread-id in expected varobj.
* gdb.ada/mi_var_array.exp: Add missing thread-id in expected varobj.
Currently, encode_actions_rsp returns two malloc'ed arrays of malloc'ed
strings (char *) by pointer. Change this to use
std::vector<std::string>. This eliminates some cleanups in remote.c.
Regtested on the buildbot.
gdb/ChangeLog:
* tracepoint.h (class collection_list) <stringify>: Return
std::vector<std::string>.
(encode_actions_rsp): Change parameters to
std::vector<std::string> *.
* tracepoint.c (collection_list::stringify): Return
std::vector<std::string> and adjust accordingly.
(encode_actions_rsp): Changee parameters to
std::vector<std::string> and adjust accordingly.
* remote.c (free_actions_list),
free_actions_list_cleanup_wrapper): Remove.
(remote_download_tracepoint): Adjust to std::vector.
Since special symbol must be defined in a regular object, definition
from a dynamic object should be ignored. If special symbol has the
hidden or internal visibility, reference from a dynamic object should
also be ignored. Also __start and __stop symbols in a dynamic object
shouldn't be preempted.
PR gold/22291
* layout.cc (Layout::define_section_symbols): Use STV_PROTECTED
for __start and __stop symbols.
* symtab.cc (Symbol_table::define_special_symbol): Add an
argument, visibility. Ignore definition and reference from
a dynamic object, depending on visibility.
(Symbol_table::do_define_in_output_data): Pass visibility to
define_special_symbol.
(Symbol_table::do_define_in_output_segment): Likewise.
(Symbol_table::do_define_as_constant): Likewise.
(Symbol_table::add_undefined_symbol_from_command_line): Pass
STV_DEFAULT to define_special_symbol.
* symtab.h (Symbol_table::define_special_symbol): Add an
argument, visibility.
Jan Beulich
Alan Modra
Simon Marchi
GDB Administrator
H.J. Lu
Andreas Schwab
Jim Wilson
Pedro Alves
Joel Brobecker
Tamar Christina
Tom Tromey
Nick Clifton
Simon Marchi