Comparing test results between
--target_board=native-gdbserver
--target_board=native-stdio-gdbserver
I noticed that gdb.base/bigcore.exp is failing with native-stdio-gdbserver:
Running src/gdb/testsuite/gdb.base/bigcore.exp ...
FAIL: gdb.base/bigcore.exp: continue (timeout)
...
The problem is that:
1. When debugging with "target remote | CMD", the inferior's
stdout/stderr streams are connected to a pipe.
2. The bigcore.c program prints a lot to the screen before it
reaches the breakpoint location that the "continue" shown above
wants to reach.
3. GDB is not flushing the inferior's output pipe while the inferior
is running.
4. The pipe becomes full.
5. The inferior thus deadlocks.
The bug is #3 above, which is what this commit fixes. A new test is
added, that specifically exercises this scenario. The test fails
before the fix, and passes after, and gdb.base/bigcore.exp also starts
passing.
gdb/ChangeLog:
2017-10-19 Pedro Alves <palves@redhat.com>
* ser-base.c (ser_base_read_error_fd): Delete the file handler if
async.
(handle_error_fd): New function.
(ser_base_async): Add/delete an event loop file handler for
error_fd.
gdb/testsuite/ChangeLog:
2017-10-19 Pedro Alves <palves@redhat.com>
* gdb.base/long-inferior-output.c: New file.
* gdb.base/long-inferior-output.exp: New file.
There doesn't seem to be a good reason we're reading the file one
chunk at a time.
gdb/ChangeLog:
2017-10-19 Pedro Alves <palves@redhat.com>
* xml-support.c (xml_fetch_content_from_file): Don't read in
chunks. Instead use fseek to determine the file's size, and read
it in one go.
Copied from x86, check bfd_link_executable, instead of bfd_link_pic,
for TLS transition check. Not sure if it works correctly. All usages
of bfd_link_pic should be audited.
PR ld/22263
* elfxx-tilegx.c (tilegx_elf_tls_transition): Replace
bfd_link_pic with !bfd_link_executable, !bfd_link_pic with
bfd_link_executable for TLS check.
(tilegx_elf_check_relocs): Likewise.
(allocate_dynrelocs): Likewise.
(tilegx_elf_relocate_section): Likewise.
Copied from x86, check bfd_link_executable, instead of bfd_link_pic,
for TLS transition check. Not sure if it works correctly. All usages
of bfd_link_pic should be audited.
PR ld/22263
* elf32-tilepro.c (tilepro_elf_tls_transition): Replace
bfd_link_pic with !bfd_link_executable, !bfd_link_pic with
bfd_link_executable for TLS check.
(tilepro_elf_check_relocs): Likewise.
(allocate_dynrelocs): Likewise.
(tilepro_elf_relocate_section): Likewise.
Copied from x86, check bfd_link_executable, instead of bfd_link_pic,
for TLS transition check. Not sure if it works correctly. All usages
of bfd_link_pic should be audited.
PR ld/22263
* elfxx-sparc.c (sparc_elf_tls_transition): Replace
bfd_link_pic with !bfd_link_executable, !bfd_link_pic with
bfd_link_executable for TLS check.
(_bfd_sparc_elf_check_relocs): Likewise.
(allocate_dynrelocs): Likewise.
(_bfd_sparc_elf_relocate_section): Likewise.
This fixes various issues with the fill-1 testcase causing fails on a
couple of targets.
gas/ChangeLog:
2017-10-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* testsuite/gas/all/fill-1.s: Use normal labels. Change .text to
.data. Pick different values. Use .dc.w instead of .word.
* testsuite/gas/all/fill-1.d: New objdump output check.
* testsuite/gas/all/gas.exp: Use run_dump_test to execute fill-1
testcase.
TLS_GD_CALL relocations implicitly reference __tls_get_addr. Since
elf_gc_mark_hook is called before check_relocs now, we need to call
_bfd_generic_link_add_one_symbol to mark __tls_get_addr for garbage
collection.
* elf32-tilepro.c (tilepro_elf_gc_mark_hook): Call
_bfd_generic_link_add_one_symbol to mark __tls_get_addr.
* elfxx-tilegx.c (tilegx_elf_gc_mark_hook): Likewise.
There are individual comments that explain why each test isn't
supported, but the vast majority of them are due to RISC-V's aggressive
linker relaxation. The SLEB test cases should eventually be supported,
but the remaining ones probably won't ever be.
2017-10-18 Palmer Dabbelt <palmer@dabbelt.com>
* testsuite/gas/all/align.d: Mark as unsupported on RISC-V.
testsuite/gas/all/relax.d: Likewise.
testsuite/gas/all/sleb128-2.d: Likewise.
testsuite/gas/all/sleb128-4.d: Likewise.
testsuite/gas/all/sleb128-5.d: Likewise.
testsuite/gas/all/sleb128-7.d: Likewise.
testsuite/gas/elf/section11.d: Likewise.
testsuite/gas/all/gas.exp (diff1.s): Likewise.
Consider a conversion operator such as:
operator foo const* const* ();
There are two small parser problems, highlighted by this test:
(gdb) p operator foo const* const*
There is no field named operatorfoo const* const *
GDB is looking up the symbol "operatorfoo const* const*" -- it is missing a
space between the keyword "operator" and the type name "foo const* const*".
Additionally, this input of the user-defined type needs to be canonicalized
so that different "spellings" of the type are recognized:
(gdb) p operator const foo* const *
There is no field named operator const foo* const *
gdb/ChangeLog:
* c-exp.y (oper): Canonicalize conversion operators of user-defined
types.
Add whitespace to front of type name.
gdb/testsuite/ChangeLog:
* gdb.cp/cpexprs.cc (base) <operator fluff const* const*>: New
method.
(main): Call it.
* gdb.cp/cpexprs.exp: Add new conversion operator to test matrix.
Add additional user-defined conversion operator tests.
A previous patch called gdb_assert_not_reached whenever reading
the accessibility of a nested typedef definition. Wisely, Pedro has asked me
not do this.
This patch changes the previous one so that it issues a complaint instead.
gdb/ChangeLog:
* dwarf2read.c (dwarf2_add_typedef): Issue a complaint on unhandled
DW_AT_accessibility.
The print_core_note change here fixes the PR, the rest is making
readelf a little more bombproof against maliciously crafted binaries.
PR 22303
* readelf.c (print_core_note): Ensure "count" sanity check
calculation doesn't overflow.
(process_notes_at): Perform note namesz and descsz checks
using unsigned comparisons against data remaining. Catch
alignment overflow of namesz and descsz too. Don't allocate a
temp for terminating "name" when there is space available
before descdata.
Relocations against indirect and warning symbols should be applied to real
symbols.
* elf32-microblaze.c (microblaze_elf_check_relocs): Check for
indirect and warning symbols.
Fixes a thinko. Given code that puts variables into the TOC (a bad
idea, but some see the TOC as a small data section) this bug could
result in an attempt to optimize a sequence that should not be
optimized.
* powerpc.cc (Target_powerpc::Scan::local): Correct dst_off
calculation for TOC16 relocs.
(Target_powerpc::Scan::global): Likewise.
I've noticed now that due to a last-minute change, commit 739b3f1d8f
("Make native gdbserver boards no longer be "remote" (in DejaGnu
terms)") managed to miss loading "local-board" in the
native-stdio-gdbserver board...
gdb/testsuite/ChangeLog:
2017-10-17 Pedro Alves <palves@redhat.com>
* boards/native-stdio-gdbserver.exp: Load "local-board".
In https://sourceware.org/ml/binutils/2017-10/msg00160.html, Nick said I
should update MAINTAINERS to mention that I'm the dwarf-mode.el
maintainer. So, I'm checking this in.
2017-10-17 Tom Tromey <tom@tromey.com>
* MAINTAINERS: Add myself as dwarf-mode.el maintainer.
In my multi-target branch, I had managed to break GDB exiting
successfuly in response to "quit" or SIGHUP/SIGTERM when:
- you're debugging with "target extended-remote",
- have more than one inferior loaded in gdb, some running, and at
least one not running, and,
- quit gdb with the inferior that is not running yet selected.
The testsuite still passed cleanly anyway. I only noticed because I
was left with a bunch of core dumps in the gdb/testsuite/ directory --
the testsuite infrastructure closes GDB's pty after running each
testcase, which results in GDB getting a SIGHUP and should make GDB
exit gracefully. If GDB crashes at that point though, there's no
indication about it in gdb.sum/gdb.log.
This commit adds a multitude of tests exercising quitting GDB with
live inferiors, some of which would have caught the problem.
gdb/testsuite/ChangeLog:
2017-10-17 Pedro Alves <palves@redhat.com>
* gdb.base/quit-live.c: New file.
* gdb.base/quit-live.exp: New file.
This changes the remaining spots in disasm.c to use the RAII ui-out
emitters, removing a few cleanups. This also fixes a regression that
Simon pointed out.
2017-10-17 Tom Tromey <tom@tromey.com>
* disasm.c (do_mixed_source_and_assembly_deprecated): Use
gdb::optional, ui_out_emit_list, ui_out_emit_tuple.
(do_mixed_source_and_assembly): Likewise.
When building I got:
../../binutils-gdb/gdb/regcache.c:935:24: error: the address of ‘ssize_t read(int, void*, size_t)’ will never be NULL [-Werror=address]
This happens because "read" used to be a parameter to this function,
which was then removed; but the assertion wasn't updated.
I don't think the assertion is relevant any more, to this removes it.
I'm checking it in as obvious.
2017-10-17 Tom Tromey <tom@tromey.com>
* regcache.c (regcache::xfer_part): Remove assertion.
When adding an unbounded increment to a pointer, you can't just check
against the end of the buffer but also must check that overflow
doesn't result in "negative" pointer movement. Pointer comparisons
are signed. Better, check the increment against the space left using
an unsigned comparison.
PR 22307
* elf-properties.c (_bfd_elf_parse_gnu_properties): Compare datasz
against size left rather than comparing pointers. Reorganise loop.
Fixes a double-free regression introduced by commit b7b030adc4
("Return unique_xmalloc_ptr from target_read_stralloc"):
gdb.sum:
Running src/gdb/testsuite/gdb.base/catch-syscall.exp ...
ERROR: Process no longer exists
Valgrind shows:
(gdb) catch syscall
==3687== Thread 1:
==3687== Invalid free() / delete / delete[] / realloc()
==3687== at 0x4C29CF0: free (vg_replace_malloc.c:530)
==3687== by 0x610862: xfree(void*) (common-utils.c:101)
==3687== by 0x440D5D: gdb::xfree_deleter<char>::operator()(char*) const (gdb_unique_ptr.h:34)
==3687== by 0x446CC6: std::unique_ptr<char, gdb::xfree_deleter<char> >::reset(char*) (unique_ptr.h:344)
==3687== by 0x81BE50: xml_fetch_content_from_file(char const*, void*) (xml-support.c:1042)
==3687== by 0x81DA86: xml_init_syscalls_info(char const*) (xml-syscall.c:366)
==3687== by 0x81DBDD: init_syscalls_info(gdbarch*) (xml-syscall.c:398)
==3687== by 0x81E131: get_syscall_by_number(gdbarch*, int, syscall*) (xml-syscall.c:599)
==3687== by 0x5BE86F: catch_syscall_command_1(char*, int, cmd_list_element*) (break-catch-syscall.c:481)
==3687== by 0x4B46B1: do_sfunc(cmd_list_element*, char*, int) (cli-decode.c:138)
==3687== by 0x4B76B8: cmd_func(cmd_list_element*, char*, int) (cli-decode.c:1952)
==3687== by 0x7E91C7: execute_command(char*, int) (top.c:615)
==3687== Address 0x14332ae0 is 0 bytes inside a block of size 4,096 free'd
==3687== at 0x4C2AB8B: realloc (vg_replace_malloc.c:785)
==3687== by 0x610792: xrealloc (common-utils.c:62)
==3687== by 0x81BE3E: xml_fetch_content_from_file(char const*, void*) (xml-support.c:1042)
==3687== by 0x81DA86: xml_init_syscalls_info(char const*) (xml-syscall.c:366)
==3687== by 0x81DBDD: init_syscalls_info(gdbarch*) (xml-syscall.c:398)
==3687== by 0x81E131: get_syscall_by_number(gdbarch*, int, syscall*) (xml-syscall.c:599)
==3687== by 0x5BE86F: catch_syscall_command_1(char*, int, cmd_list_element*) (break-catch-syscall.c:481)
==3687== by 0x4B46B1: do_sfunc(cmd_list_element*, char*, int) (cli-decode.c:138)
==3687== by 0x4B76B8: cmd_func(cmd_list_element*, char*, int) (cli-decode.c:1952)
==3687== by 0x7E91C7: execute_command(char*, int) (top.c:615)
==3687== by 0x6A422D: command_handler(char*) (event-top.c:583)
==3687== by 0x6A45F2: command_line_handler(char*) (event-top.c:773)
[...]
The problem is that if xrealloc decides it needs a new memory block,
it frees the previous block/pointer, and then text.reset() frees it
again.
gdb/ChangeLog:
2017-10-17 Pedro Alves <palves@redhat.com>
* xml-support.c (xml_fetch_content_from_file): Call
unique_ptr::release() instead unique_ptr::get() when passing
through xrealloc.
Since xfer_part is already a class method, and only
{raw,cooked}_{read,write} are passed to it. We can remove these two
arguments, but add a bool argument is_raw, indicating raw registers or
cooked registers are accessed.
gdb:
2017-10-17 Yao Qi <yao.qi@linaro.org>
* regcache.c (regcache::xfer_part): Remove parameters read and
write, add parameter is_raw. All callers are updated.
gdb/gdbserver:
2017-10-17 Yao Qi <yao.qi@linaro.org>
* configure.srv: Rename arm-linux.o with arch/arm-linux.o.
Rename arm-get-next-pcs.o with arch/arm-get-next-pcs.o.
Similar to f38307f5 (Replicate src dir in build dir), this patch change
configure and Makefile to generate object files in arch/ directory.
gdb/gdbserver:
2017-10-17 Yao Qi <yao.qi@linaro.org>
* Makefile.in (CONFIG_SRC_SUBDIR): New variable.
(clean): Remove .o files in CONFIG_SRC_SUBDIR.
(distclean): Remove DEPDIR in CONFIG_SRC_SUBDIR.
(arch-i386.o, arch-amd64.o): Remove rules.
(arch/%.o): New rule.
Update POSTCOMPILE and COMPILE.pre.
* configure.ac: Invoke AC_CONFIG_COMMANDS.
* configure: Re-generated.
* configure.srv: Replace arch-i386.o with arch/i386.o.
Replace arch-amd64.o with arch/amd64.o.
-z text applies to all dynamic binaries, not just shared libraries.
A lot of the other options needed attention too.
* ld.texinfo (-z): Combine negative options with corresponding
positive option. Sort the table. Expand and correct "combreloc",
"common", "common-page-size", "interpose", "loadfltr",
"max-page-size", "muldefs", "nodefaultlib", "nodelete", "nodlopen",
"nodump", "noextern-protected-data", "now", "origin", and "text".
2017-10-16 Sandra Loosemore <sandra@codesourcery.com>
Henry Wong <henry@stuffedcow.net>
gas/
* config/tc-nios2.c (nios2_translate_pseudo_insn): Check for
correct number of arguments.
(md_assemble): Handle failure of nios2_translate_pseudo_insn.
* testsuite/gas/nios2/illegal_pseudoinst.l: New file.
* testsuite/gas/nios2/illegal_pseudoinst.s: New file.
* testsuite/gas/nios2/nios2.exp: Add illegal_pseudoinst test.
We currently do not record access information for typedefs defined inside
classes. Consider:
struct foo
{
typedef int PUBLIC;
private:
typedef int PRIVATE;
PRIVATE b;
};
(gdb) ptype foo
type = struct foo {
private:
PRIVATE b;
typedef int PRIVATE;
typedef int PUBLIC;
}
This patch fixes this:
(gdb) ptype foo
type = struct foo {
private:
PRIVATE b;
typedef int PRIVATE;
public:
typedef int PUBLIC;
}
gdb/ChangeLog:
* c-typeprint.c (enum access_specifier): Moved here from
c_type_print_base.
(output_access_specifier): New function.
(c_type_print_base): Consider typedefs when assessing
whether access labels are needed.
Use output_access_specifier as needed.
Output access specifier for typedefs, if needed.
* dwarf2read.c (dwarf2_add_typedef): Record DW_AT_accessibility.
* gdbtypes.h (struct typedef_field) <is_protected, is_private>: New
fields.
(TYPE_TYPEDEF_FIELD_PROTECTED, TYPE_TYPEDEF_FIELD_PRIVATE): New
accessor macros.
gdb/testsuite/ChangeLog:
* gdb.cp/classes.cc (class_with_typedefs, class_with_public_typedef)
(class_with_protected_typedef, class_with_private_typedef)
(struct_with_public_typedef, struct_with_protected_typedef)
(struct_with_private_typedef): New classes/structs.
* gdb.cp/classes.exp (test_ptype_class_objects): Add tests for
typedefs and access specifiers.
Change target_fileio_read_stralloc to return unique_xmalloc_ptr and
fix up the callers. This removes a number of cleanups.
ChangeLog
2017-10-16 Tom Tromey <tom@tromey.com>
* linux-tdep.c (linux_info_proc, linux_find_memory_regions_full)
(linux_fill_prpsinfo, linux_vsyscall_range_raw): Update.
* target.c (target_fileio_read_stralloc): Update.
* sparc64-tdep.c (adi_is_addr_mapped): Update.
* target.h (target_fileio_read_stralloc): Return
unique_xmalloc_ptr.
This removes a few cleanups in remote.c using the usual techniques:
std::vector, unique_xmalloc_ptr, and gdb::def_vector.
ChangeLog
2017-10-16 Tom Tromey <tom@tromey.com>
* remote.c (remote_register_number_and_offset): Use std::vector.
(remote_set_syscall_catchpoint): Use gdb::unique_xmalloc_ptr.
(putpkt_binary): Use gdb::def_vector.
(compare_sections_command): Use gdb::byte_vector.
This removes a cleanup from ppc-linux-nat.c, by using
unique_xmalloc_ptr. It also slightly simplifies the code by using
XDUP rather than XNEW and memcpy.
ChangeLog
2017-10-16 Tom Tromey <tom@tromey.com>
* ppc-linux-nat.c (hwdebug_insert_point): Use
gdb::unique_xmalloc_ptr, XDUP.
This removes some cleanups from parse_probes by using std::string; and
removes some unnecessary cleanups from elsewhere in probe.c.
ChangeLog
2017-10-16 Tom Tromey <tom@tromey.com>
* probe.c (parse_probes): Use std::string.
(info_probes_for_ops, enable_probes_command)
(disable_probes_command): Remove cleanups.
Change end_symtab_get_static_block to use std::vector. This removes a
cleanup.
ChangeLog
2017-10-16 Tom Tromey <tom@tromey.com>
* buildsym.c (block_compar): Remove.
(end_symtab_get_static_block): Use std::vector.
This commit finally clears the "isremote" flag in the native-gdbserver
and native-stdio-gdbserver boards. The goal is to make all "native"
boards be considered not remote in DejaGnu terms, like the
native-extended-gdbserver board is too.
DejaGnu automatically considers boards remote if their names don't
match the local hostname. That means that native-gdbserver and
native-extended-gdbserver are considered remote by default by DejaGnu,
even though they run locally. native-extended-gdbserver, however,
overrides its isremote flag to force it to be not remote. So we are
in that weird state where native-gdbserver is considered remote, and
native-extended-gdbserver is considered not remote.
A recent set of commits fixed all the problems (and some more) exposed
by testing with --target_board=native-gdbserver and
--target_board=native-stdio-gdbserver with isremote forced off on
x86-64 GNU/Linux. I believe we're good to go now.
The native-stdio-gdbserver.exp/remote-stdio-gdbserver.exp boards
required deep non-obvious modifications unfortunately... The problem
is that if a board is not remote, then DejaGnu doesn't call
${board}_spawn / ${board}_exec at all, and the
native-stdio-gdbserver.exp board relies on those procedures being
called. To fix that, this commit redesigns how the stdio boards hook
into the testing framework to spawn gdbserver. IMO, this is a good
change anyway, because the way its done currently is a bit of a hack,
and the result turns out to be simpler, even. With this commit, they
now no longer load the "gdbserver" generic config, and hook at the
mi_gdb_target_load/gdb_reload level instead, making them more like
traditional board files.
To share code between native-stdio-gdbserver.exp and
remote-stdio-gdbserver.exp, a new shared stdio-gdbserver-base.exp file
is created.
Instead of having each native board clear isremote manually, boards
source the new "local-board.exp" file.
This also adds a new section to testsuite/README file discussing
local/remote/native, so that we can easily refer to it.
gdb/testsuite/ChangeLog:
2017-10-16 Pedro Alves <palves@redhat.com>
Simon Marchi <simon.marchi@polymtl.ca>
* README (Local vs Remote vs Native): New section.
* boards/local-board.exp: New file, with bits factored out from
...
* boards/native-extended-gdbserver.exp: ... here. Load
"local-board".
* boards/native-gdbserver.exp: Load "local-board".
(${board}_spawn, ${board}_exec): Delete.
* boards/native-stdio-gdbserver.exp: Most contents factored out to
...
* boards/stdio-gdbserver-base.exp: ... this new file.
* boards/native-stdio-gdbserver.exp: Reimplement, by loading
"stdio-gdbserver-base" and defining a get_target_remote_pipe_cmd
procedure.
* boards/remote-stdio-gdbserver.exp: Load stdio-gdbserver-base
instead of native-stdio-gdbserver. Don't set gdb_server_prog nor
stdio_gdbserver_command.
(${board}_get_remote_address, ${board}_get_comm_port)
(${board}_download, ${board}_upload): Delete.
(get_target_remote_pipe_cmd): New.
Use proc_with_prefix to avoid having to call with_test_prefix with a
duplicate of the proc name. The diff is mostly lines being re-indented.
gdb/testsuite/ChangeLog:
* gdb.python/py-breakpoint.exp (test_bkpt_basic,
test_bkpt_deletion, test_bkpt_cond_and_cmds,
test_bkpt_invisible, test_watchpoints, test_bkpt_internal,
test_bkpt_eval_funcs, test_bkpt_temporary, test_bkpt_address,
test_bkpt_pending, test_bkpt_events): Use proc_with_prefix,
remove with_test_prefix.
This patch replaces the last usages of VEC(mem_range_s) with
std::vector<mem_range>. This allows getting rid of a few cleanups and
of the DEF_VEC_O(mem_range_s).
I added a test for normalize_mem_ranges to make sure I didn't break
anything there.
Regtested on the buildbot.
gdb/ChangeLog:
* memrange.h (struct mem_range): Define operator< and operator==.
(mem_range_s): Remove.
(DEF_VEC_O (mem_range_s)): Remove.
(normalize_mem_ranges): Change parameter type to std::vector.
* memrange.c (compare_mem_ranges): Remove.
(normalize_mem_ranges): Change parameter type to std::vector,
adjust to vector change.
* exec.c (section_table_available_memory): Return vector, remove
parameter.
(section_table_read_available_memory): Adjust to std::vector
change.
* remote.c (remote_read_bytes): Adjust to std::vector
change.
* tracepoint.h (traceframe_available_memory): Change parameter
type to std::vector.
* tracepoint.c (traceframe_available_memory): Change parameter
type to std::vector, adjust.
* gdb/mi/mi-main.c (mi_cmd_trace_frame_collected): Adjust to
std::vector change.
* gdb/Makefile.in (SUBDIR_UNITTESTS_SRCS): Add
unittests/memrange-selftests.c.
(SUBDIR_UNITTESTS_OBS): Add memrange-selftests.o.
* gdb/unittests/memrange-selftests.c: New file.
This commit works around a GCC 6.3.1 bug several people are hitting:
https://sourceware.org/ml/gdb-patches/2017-09/msg00270.htmlhttps://sourceware.org/ml/gdb-patches/2017-10/msg00418.html
It manifests like this:
../../../binutils-gdb/gdb/probe.c:68:12: error: types may not be defined in a for-range-declaration [-Werror]
for (struct probe *probe : probes)
^~~~~~
Fix it by renaming the range-for named variables to something different
from their type's name.
gdb/ChangeLog:
2017-10-16 Pedro Alves <palves@redhat.com>
* elfread.c (probe_key_free): Rename range-for variable.
* probe.c (parse_probes_in_pspace, find_probes_in_objfile)
(find_probe_by_pc, collect_probes): Rename range-for variable.
