Ref: https://sourceware.org/ml/gdb-patches/2015-08/msg00675.html
If multiprocess extensions are off (because specific gdbserver port
doesn't support them), then when gdbserver doesn't have a thread
selected yet, and GDB sends Hg packet to select one, gdbserver
crashes. That's because extracting the desired thread id out of the
packet that GDB sent depends on the current thread to fill in the
missing process id ... Fix this by getting the process id from the
first (and only) process in the processes list instead.
The GNU/Linux port doesn't trip on this because it always runs with
multiprocess extensions enabled. To make it easier to catch such
regressions going forward, this commit also adds a new smoke test that
spawns gdbserver, connects to it and runs to main with the
multiprocess extensions force-disabled.
gdb/gdbserver/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* inferiors.c (get_first_process): New function.
* inferiors.h (get_first_process): New declaration.
* remote-utils.c (read_ptid): Default to the first process in the
list, instead of to the current thread's process.
gdb/testsuite/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* gdb.server/connect-without-multi-process.c: New file.
* gdb.server/connect-without-multi-process.exp: New file.
Being able to force-disable the RSP multiprocess extensions is useful
for testing.
gdb/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* NEWS (New commands): Mention set/show remote
multiprocess-extensions-packet.
* remote.c (remote_query_supported): Only tell the server to use
the multiprocess extensions if the user hasn't force-disabled them
with "set remote multiprocess-extensions-packet off".
gdb/doc/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* gdb.texinfo (Remote Configuration): Document the "set/show
remote multiprocess-extensions-packet" commands.
After the last gnulib import (Dec 2012), gnulib upstream started
replacing mingw's 'struct timeval' with a version with 64-bit time_t,
for POSIX compliance:
commit f8e84098084b3b53bc6943a5542af1f607ffd477
Author: Bruno Haible <bruno@clisp.org>
Date: Sat Jan 28 18:12:10 2012 +0100
sys_time: Override 'struct timeval' on some native Windows platforms.
See:
https://lists.gnu.org/archive/html/bug-gnulib/2012-01/msg00372.html
However, that results in conflicts with native Winsock2's 'select':
select()'s argument
http://sourceforge.net/p/mingw-w64/mailman/message/29610438/
... and libiberty's timeval-utils.h timeval_add/timeval_sub, at the
least.
We don't really need the POSIX compliance, so this patch prepares us
to simply not use gnulib's 'struct timeval' replacement once a more
recent gnulib is imported, thus preserving the current behavior, by
adding a sys/time.h wrapper header that undefs gnulib's replacements,
and including that everywhere instead.
The SIZE -> OSIZE change is necessary because newer gnulib's
sys/time.h also includes windows.h/winsock2.h, which defines a
conflicting SIZE symbol.
Cross build-tested mingw-w64 32-bit and 64-bit.
Regtested on x86_64 Fedora 20.
gdb/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* Makefile.in (HFILES_NO_SRCDIR): Add common/gdb_sys_time.h.
* common/gdb_sys_time.h: New file.
* event-loop.c: Include gdb_sys_time.h instead of sys/time.h.
* gdb_select.h: Likewise.
* gdb_usleep.c: Likewise.
* maint.c: Likewise.
* mi/mi-main.c: Likewise.
* mi/mi-parse.h: Likewise.
* remote-fileio.c: Likewise.
* remote-m32r-sdi.c: Likewise.
* remote.c: Likewise.
* ser-base.c: Likewise.
* ser-pipe.c: Likewise.
* ser-tcp.c: Likewise.
* ser-unix.c: Likewise.
* symfile.c: Likewise.
* symfile.c: Likewise. Rename OSIZE to SIZE throughout.
* target-memory.c: Include gdb_sys_time.h instead of sys/time.h.
* utils.c: Likewise.
gdb/gdbserver/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* debug.c: Include gdb_sys_time.h instead of sys/time.h.
* event-loop.c: Likewise.
* remote-utils.c: Likewise.
* tracepoint.c: Likewise.
Ref: https://sourceware.org/ml/gdb-patches/2015-08/msg00675.html
gdbserver/spu-low.c: In function 'spu_request_interrupt':
gdbserver/spu-low.c:639: error: incompatible type for argument 1 of 'ptid_get_lwp'
gdb/gdbserver/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* spu-low.c (spu_request_interrupt): Use lwpid_of instead of
ptid_get_lwp.
This makes z an int for gdb/testsuite/gdb.opt/inline-markers.c.
gdb/testsuite/ChangeLog:
2015-08-24 Luis Machado <lgustavo@codesourcery.com>
* gdb.opt/inline-markers.c: Make z int.
This fixes a typo in gdb/testsuite/gdb.opt/inline-markers.c, making
z a volatile variable.
gdb/testsuite/ChangeLog:
2015-08-24 Luis Machado <lgustavo@codesourcery.com>
* gdb.opt/inline-markers.c: Make z volatile.
While doing some powerpc Linux tests on a ppc 476 board using GCC 5.2, i
noticed inline-bt.exp, inline-cmds.exp and inline-locals.exp failing.
FAIL: gdb.opt/inline-bt.exp: continue to bar (1)
FAIL: gdb.opt/inline-bt.exp: backtrace from bar (1)
FAIL: gdb.opt/inline-bt.exp: continue to bar (2)
FAIL: gdb.opt/inline-bt.exp: backtrace from bar (2)
FAIL: gdb.opt/inline-bt.exp: continue to bar (3)
FAIL: gdb.opt/inline-bt.exp: backtrace from bar (3)
FAIL: gdb.opt/inline-cmds.exp: continue to bar (1)
FAIL: gdb.opt/inline-cmds.exp: backtrace from bar (1)
FAIL: gdb.opt/inline-cmds.exp: continue to bar (2)
FAIL: gdb.opt/inline-cmds.exp: backtrace from bar (2)
FAIL: gdb.opt/inline-cmds.exp: continue to marker
FAIL: gdb.opt/inline-cmds.exp: backtrace from marker
FAIL: gdb.opt/inline-cmds.exp: step into finish marker
FAIL: gdb.opt/inline-locals.exp: continue to bar (1)
FAIL: gdb.opt/inline-locals.exp: continue to bar (2)
FAIL: gdb.opt/inline-locals.exp: backtrace from bar (2)
FAIL: gdb.opt/inline-locals.exp: continue to bar (3)
FAIL: gdb.opt/inline-locals.exp: backtrace from bar (3)
They failed because the breakpoint supposedly inserted at bar was actually
inserted at noinline.
(gdb) break inline-markers.c:20^M
Breakpoint 2 at 0x1000079c: file gdb/testsuite/gdb.opt/inline-markers.c, line 20.^M
(gdb) continue^M
Continuing.^M
^M
Breakpoint 2, noinline () at gdb/testsuite/gdb.opt/inline-markers.c:35^M
35 inlined_fn (); /* inlined */^M
As we can see, line 20 is really inside bar, not noinline:
18 void bar(void)
19 {
20 x += y; /* set breakpoint 1 here */
21 }
Further investigation shows that this is really due to GCC 5's new
ICF pass (-fipa-icf), now enabled by default at -O2, which folds bar
and marker into noinline, where the call to inlined_fn was inlined.
This breaks the testcase since it expects to stop at specific spots.
I thought about two possible fixes for this issue.
- Disable the ICF pass manually when building the binary (-fno-ipa-icf).
This has the advantage of not having to touch the testcase sources themselves,
but the disadvantage of having to add conditional blocks to test the GCC
version. If we ever change GCC's default, we will have to adjust the
conditional block again to match GCC's behavior.
- Modify the testcase sources to make the identical functions unique.
This solution doesn't touch the testcase itself, but changes the source
code slightly in order to make bar, marker and inlined_fn unique. This
causes GCC's ICF pass to ignore these functions and not fold them into
a common identical function.
I'm good with either of them, but i'm more inclined to go with the second
one.
The attached patch implements this by adding the new global variable z, set
to 0, that gets added in different ways to marker and inlined_fn. Since it
is 0, it doesn't affect any possible value checks that we may wish to do
in the future (we currently only check for values changed by bar).
Ok?
ps: I also noticed GDB doesn't do a great job at stating that the breakpoint
was actually inserted at a different source line than previously requested,
so this sounds like a bug that should be fixed, if it is not just wrong
DWARF information (did not investigate it further).
gdb/testsuite/ChangeLog:
2015-08-24 Luis Machado <lgustavo@codesourcery.com>
* gdb.opt/inline-bt.c: New volatile global z.
* gdb.opt/inline-cmds.c: Likewise.
* gdb.opt/inline-locals.c: Likewise.
* gdb.opt/inline-markers.c: New extern global z.
(marker): Use z.
(inline_fn): Likewise.
Support for target dbug/picobug/dink32/m32r/mon2000/ppcbug was just
removed, but support for ARM RDI, Sparclet, Sparclite, Z8000, target
r3900, target array, target sds, target op50n and target w89k had
already been removed many years ago. Drop it all in one go.
gdb/doc/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* gdb.texinfo (Embedded Processors) <ARM>: Remove references to RDI.
<M32R>: Remove references to M32R/D.
<M68K>: Remove references to target dbug.
<MIPS Embedded>: Remove references to target r3900 and target
array.
<PowerPC Embedded>: Remove references to target dink32 and target
ppcbug, target sds
<PA, Sparclet, Sparclite, Z8000>: Delete nodes.
Ref: https://sourceware.org/ml/gdb/2015-07/msg00011.html
All of these targets use gdb/monitor.c, which has bit rotted
years ago (I'd guess around ~6), and nobody seems to
have noticed:
| target | source |
|----------------+----------------------|
| target dbug | gdb/dbug-rom.c |
| target picobug | gdb/microblaze-rom.c |
| target dink32 | gdb/dink32-rom.c |
| target m32r | gdb/m32r-rom.c |
| target mon2000 | gdb/m32r-rom.c |
| target ppcbug | gdb/ppcbug-rom.c |
This deletes them, along with finally removing monitor.c.
A manual update will be done separately.
gdb/ChangeLog:
2015-08-24 Pedro Alves <palves@redhat.com>
* NEWS: Mention removed support for the various ROM monitors.
* Makefile.in (ALL_TARGET_OBS): Remove dbug-rom.o, dink32-rom.o,
ppcbug-rom.o, m32r-rom.o, dsrec.o and monitor.o from gdb_target_obs.
* configure.tgt (h8300-*-*): Remove monitor.o and m32r-rom.o from
gdb_target_obs.
(m68*-*-*): Remove monitor.o dbug-rom.o and dsrec.o from
gdb_target_obs.
(microblaze*-linux-*): Remove microblaze-rom.o, monitor.o and
dsrec.o from gdb_target_obs.
(microblaze*-*-*): Remove microblaze-rom.o, monitor.o and dsrec.o
from gdb_target_obs.
(powerpc-*-lynx*178): Remove monitor.o and dsrec.o from
gdb_target_obs.
(powerpc*-*-*): Remove monitor.o, dsrec.o, ppcbug-rom.o and
dink32-rom.o from gdb_target_obs.
(sh*-*-linux*): Remove monitor.o and dsrec.o from gdb_target_obs.
(sh*): Remove monitor.o and dsrec.o from gdb_target_obs.
* dbug-rom.c, dink32-rom.c, dsrec.c, m32r-rom.c, microblaze-rom.c,
monitor.c, monitor.h, ppcbug-rom.c, srec.h: Delete files.
opcodes * i386-dis.c (print_insn): Fix decoding of three byte operands.
tests * gas/i386/intel.s: Add test of disassembly of a potential
three byte instuction at the end of a function.
* gas/i386/intel.d: Update expected disassembly.
This avoids two more types of FAILs with the gnu_vector test case.
First, for POWER targets newer GCCs emit an ABI note when invoked with
"-mcpu=native". Then the test case fell back to non-native compile,
producing code for a non-vector ABI. But that is not supported by GDB.
Thus the compiler note is now suppressed with "-Wno-psabi".
Second, on s390 the test case produced FAILs after falling back to a
non-vector ABI when using "finish" or "return" in a vector-valued
function. This was due to a long-standing known bug (Bug 8549). This
case is now detected, and KFAILs are emitted instead.
gdb/testsuite/ChangeLog:
* gdb.base/gnu_vector.exp: Try compilation with "-mcpu=native
-Wno-psabi" if "-mcpu=native" fails. For the tests with "finish"
and "return" use KFAIL when GDB can not read/write the vector
return value.
(Commit d23d1dff missed pushing the config/iconv.m4 change, and to
regenerate binutils' configure.)
Original description:
This is the second in a series of patches to make a build with an
in-tree GNU libiconv work as designed.
Currently GDB is the only toolchain component which actually uses an
in-tree libiconv. This patch modifies the common AM_ICONV to use an
in-tree libiconv when present and not already provided by libc.
(GDB's workaround uses an in-tree libiconv even when libc provides
iconv(3); I'm not sure when or why that would be desirable.)
config/ChangeLog:
2015-08-24 Yaakov Selkowitz <yselkowi@redhat.com>
* iconv.m4 (AM_ICONV_LINK): Use in-tree libiconv when present.
binutils/ChangeLog:
2015-08-24 Yaakov Selkowitz <yselkowi@redhat.com>
* configure: Regenerate.
Belt and braces fix. Either of the changes here is sufficient to
ensure vd_nodename is initialized properly.
* elf.c (_bfd_elf_slurp_version_tables): Always init vd_nodename.
Don't copy fields not set by _bfd_elf_swap_verdef_in.
In C++ mode:
src/gdb/gdbserver/ax.c: In function ‘eval_result_type gdb_eval_agent_expr(eval_agent_expr_context*, agent_expr*, ULONGEST*)’:
src/gdb/gdbserver/ax.c:1335:11: error: invalid conversion from ‘int’ to ‘eval_result_type’ [-fpermissive]
return 1;
^
"1" as an enum eval_result_type is expr_eval_empty_expression, but
clearly this wants to return expr_eval_unhandled_opcode.
gdb/gdbserver/ChangeLog:
2015-08-21 Pedro Alves <palves@redhat.com>
* ax.c (gdb_eval_agent_expr): Return expr_eval_unhandled_opcode
instead of literal 1.
Fixes:
../../src/gdb/dwarf2read.c:127:15: error: declaration of ‘asection* dwarf2_section_info::<anonymous union>::asection’ [-fpermissive]
asection *asection;
^
In file included from ../../src/gdb/common/common-types.h:35:0,
from ../../src/gdb/common/common-defs.h:44,
from ../../src/gdb/defs.h:28,
from ../../src/gdb/dwarf2read.c:31:
../bfd/bfd.h:1596:3: error: changes meaning of ‘asection’ from ‘typedef struct bfd_section asection’ [-fpermissive]
} asection;
^
gdb/ChangeLog:
2015-08-21 Tom Tromey <tromey@redhat.com>
* dwarf2read.c (struct dwarf2_section_info): Rename field
'asection' to 'section'.
(dwarf2_has_info, get_section_bfd_owner, get_section_bfd_section)
(dwarf2_locate_sections, dwarf2_locate_sections)
(locate_dwz_sections, locate_v1_virtual_dwo_sections)
(dwarf2_locate_dwo_sections, dwarf2_locate_dwo_sections)
(dwarf2_locate_v2_dwp_sections): Adjust.
This patch fixes the following bug in TUI:
(gdb) break foo
No symbol table is loaded. Use the "file" command.
Make breakpoint pending on future shared library load? (y or [n]) <ENTER>
By submitting an empty command line to a secondary prompt, the line
corresponding to the secondary prompt is undesirably cleared and
overwritten. Outside of a secondary prompt, clearing the prompt line
after submitting an empty command line is intended behavior which
complements GDB's repeat-command shorthand. But inside a secondary
prompt, this behavior is undesired since the shorthand is not applicable
in that case. We should retain the secondary-prompt line even when it's
given no input.
This patch makes sure that a prompt that was given an empty command line
is cleared and overwritten only if it's not a secondary prompt. To
acheive this, a new predicate is defined which informs us whether the
current input handler is a secondary prompt.
gdb/ChangeLog:
* top.h (gdb_in_secondary_prompt_p): Declare.
* top.c (gdb_secondary_prompt_depth): Define.
(gdb_in_secondary_prompt_p): Define.
(gdb_readline_wrapper_cleanup): Decrement
gdb_secondary_prompt_depth.
(gdb_readline_wrapper): Increment gdb_secondary_prompt_depth.
* tui/tui-io.c (tui_getc): Don't clear the prompt line if we
are in a secondary prompt.
This is necessary to make sure that start_line is updated after a
command has been entered. Usually, start_line gets updated anyway
because most commands output text, and outputting text is done through
the function tui_puts, which updates start_line. However if a command
does not output text, then tui_puts will not get called and start_line
will not get updated in time for the next prompt to be displayed.
One can observe this bug by executing the command "delete" within TUI.
After executing, the prompt line
(gdb) delete
gets overwritten by the next prompt. With this patch, the prompt line
gets preserved.
gdb/ChangeLog:
* tui/tui-io.c (tui_getc): Use tui_putc instead of waddch to
emit the newline.
Running that test in a loop, I found a gdbserver core dump with the
following back trace:
Core was generated by `../gdbserver/gdbserver --once --multi :2346'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0 0x0000000000406ab6 in inferior_regcache_data (inferior=0x0) at src/gdb/gdbserver/inferiors.c:236
236 return inferior->regcache_data;
(gdb) up
#1 0x0000000000406d7f in get_thread_regcache (thread=0x0, fetch=1) at src/gdb/gdbserver/regcache.c:31
31 regcache = (struct regcache *) inferior_regcache_data (thread);
(gdb) bt
#0 0x0000000000406ab6 in inferior_regcache_data (inferior=0x0) at src/gdb/gdbserver/inferiors.c:236
#1 0x0000000000406d7f in get_thread_regcache (thread=0x0, fetch=1) at src/gdb/gdbserver/regcache.c:31
#2 0x0000000000409271 in prepare_resume_reply (buf=0x20dd593 "", ptid=..., status=0x20edce0) at src/gdb/gdbserver/remote-utils.c:1147
#3 0x000000000040ab0a in vstop_notif_reply (event=0x20edcc0, own_buf=0x20dd590 "T05") at src/gdb/gdbserver/server.c:183
#4 0x0000000000426b38 in notif_write_event (notif=0x66e6c0 <notif_stop>, own_buf=0x20dd590 "T05") at src/gdb/gdbserver/notif.c:69
#5 0x0000000000426c55 in handle_notif_ack (own_buf=0x20dd590 "T05", packet_len=8) at src/gdb/gdbserver/notif.c:113
#6 0x000000000041118f in handle_v_requests (own_buf=0x20dd590 "T05", packet_len=8, new_packet_len=0x7fff742c77b8)
at src/gdb/gdbserver/server.c:2862
#7 0x0000000000413850 in process_serial_event () at src/gdb/gdbserver/server.c:4148
#8 0x0000000000413945 in handle_serial_event (err=0, client_data=0x0) at src/gdb/gdbserver/server.c:4196
#9 0x000000000041a1ef in handle_file_event (event_file_desc=5) at src/gdb/gdbserver/event-loop.c:429
#10 0x00000000004199b6 in process_event () at src/gdb/gdbserver/event-loop.c:184
#11 0x000000000041a735 in start_event_loop () at src/gdb/gdbserver/event-loop.c:547
#12 0x00000000004123d2 in captured_main (argc=4, argv=0x7fff742c7ac8) at src/gdb/gdbserver/server.c:3562
#13 0x000000000041252e in main (argc=4, argv=0x7fff742c7ac8) at src/gdb/gdbserver/server.c:3631
Clearly this means that a thread pushed a stop reply in the event
queue, and then before GDB confused the event, the whole process died,
along with its thread. But the pending thread event was left
dangling. When GDB fetched that event, gdbserver looked up the
corresponding thread, but found NULL; not expecting this, gdbserver
crashes when it tries to read this thread's registers.
gdb/gdbserver/
2015-08-21 Pedro Alves <palves@redhat.com>
PR gdb/18749
* inferiors.c (remove_thread): Discard any pending stop reply for
this thread.
* server.c (remove_all_on_match_pid): Rename to ...
(remove_all_on_match_ptid): ... this. Work with a filter ptid
instead of a pid.
(discard_queued_stop_replies): Change parameter to a ptid. Now
extern.
(handle_v_kill, kill_inferior_callback)
(process_serial_event): Adjust.
(captured_main): Call initialize_notif before starting the
program, thus before threads are created.
* server.h (discard_queued_stop_replies): Declare.
In all-stop mode, if the current thread disappears while stopping all
threads, gdbserver calls set_desired_thread(0) ['0' means "I want the
continue thread"] which just picks the first thread in the list.
This looks like a dangerous thing to do. GDBserver continues
processing whatever it was doing, but to the wrong thread. If
debugging more than one process, we may even pick the wrong process.
Instead, GDBserver should detect the situation and bail out of
whatever is was doing.
The backends used to pay attention to the set 'cont_thread' (the Hc
thread, used in the old way to resume threads, before vCont), but all
such 'cont_thread' checks have been eliminated meanwhile. The
remaining implicit dependencies that I found on there being a selected
thread in the backends are in the Ctrl-C handling, which some backends
use as thread to send a signal to. Even that seems to me to be better
handled by always using the first thread in the list or by using the
signal_pid PID.
In order to make this a systematic approach, I'm making
set_desired_thread never fallback to a random thread, and instead end
up with current_thread == NULL, like already done in non-stop mode.
Then I updated all callers to handle the situation.
I stumbled on this while fixing other bugs exposed by
gdb.threads/fork-plus-threads.exp test. The problems I saw were fixed
in a different way, but in any case, I think the potential for
problems is more or less obvious, and the resulting code looks a bit
less magical to me.
Tested on x86-64 Fedora 20, w/ native-extended-gdbserver board.
gdb/gdbserver/ChangeLog:
2015-08-21 Pedro Alves <palves@redhat.com>
* linux-low.c (wait_for_sigstop): Always switch to no thread
selected if the previously current thread dies.
* lynx-low.c (lynx_request_interrupt): Use the first thread's
process instead of the current thread's.
* remote-utils.c (input_interrupt): Don't check if there's no
current thread.
* server.c (gdb_read_memory, gdb_write_memory): If setting the
current thread to the general thread fails, error out.
(handle_qxfer_auxv, handle_qxfer_libraries)
(handle_qxfer_libraries_svr4, handle_qxfer_siginfo)
(handle_qxfer_spu, handle_qxfer_statictrace, handle_qxfer_fdpic)
(handle_query): Check if there's a thread selected instead of
checking whether there's any thread in the thread list.
(handle_qxfer_threads, handle_qxfer_btrace)
(handle_qxfer_btrace_conf): Don't error out early if there's no
thread in the thread list.
(handle_v_cont, myresume): Don't set the current thread to the
continue thread.
(process_serial_event) <Hg handling>: Also set thread_id if the
previous general thread is still alive.
(process_serial_event) <g/G handling>: If setting the current
thread to the general thread fails, error out.
* spu-low.c (spu_resume, spu_request_interrupt): Use the first
thread's lwp instead of the current thread's.
* target.c (set_desired_thread): If the desired thread was not
found, leave the current thread pointing to NULL. Return an int
(boolean) indicating success.
* target.h (set_desired_thread): Change return type to int.
GDB provides no indicator of progress during file operations, and can
appear to have locked up during slow remote transfers. This commit
updates GDB to print a warning each time a file is accessed over RSP.
An additional message detailing how to avoid remote transfers is
printed for the first transfer only.
gdb/ChangeLog:
* target.h (struct target_ops) <to_fileio_open>: New argument
warn_if_slow. Update comment. All implementations updated.
(target_fileio_open_warn_if_slow): New declaration.
* target.c (target_fileio_open): Renamed as...
(target_fileio_open_1): ...this. New argument warn_if_slow.
Pass warn_if_slow to implementation. Update debug printing.
(target_fileio_open): New function.
(target_fileio_open_warn_if_slow): Likewise.
* gdb_bfd.c (gdb_bfd_iovec_fileio_open): Use new function
target_fileio_open_warn_if_slow.
gdb/testsuite/ChangeLog:
* gdb.trace/pending.exp: Cope with remote transfer warnings.
This commit fixes a stale cleanup left by linux_mntns_access_fs.
gdb/ChangeLog:
* nat/linux-namespaces.c (linux_mntns_access_fs):
Do not overwrite old_chain.
These changes allow debugging multithreaded NPTL xtensa applications.
2015-08-20 Max Filippov <jcmvbkbc@gmail.com>
gdb/gdbserver/
* configure.srv (xtensa*-*-linux*): Add srv_linux_thread_db=yes.
* linux-xtensa-low.c (arch/xtensa.h gdb_proc_service.h): New
#includes.
(ps_get_thread_area): New function.
2015-08-20 Max Filippov <jcmvbkbc@gmail.com>
gdb/
* arch/xtensa.h: New file.
* xtensa-linux-nat.c (gdb_proc_service.h): New #include.
(ps_get_thread_area): New function.
* xtensa-linux-tdep.c (xtensa_linux_init_abi): Add call to
set_gdbarch_fetch_tls_load_module_address to enable TLS support.
* xtensa-tdep.c (osabi.h): New #include.
(xtensa_gdbarch_init): Call gdbarch_init_osabi to register
xtensa-specific hooks.
* xtensa-tdep.h (struct xtensa_elf_gregset_t): Add threadptr
member and move the structure to arch/xtensa.h.
This patch almost halves the time it takes to "target remote + run to
main" on a higher-latency connection.
E.g., I've got a ping time of ~85ms to an x86-64 machine on the gcc
compile farm (almost 2000km away from me), and I'm behind a ~16Mbit
ADSL. When I connect to a gdbserver debugging itself on that machine
and run to main, it takes almost 55 seconds:
[palves@gcc76] $ ./gdbserver :9999 ./gdbserver
[palves@home] $ ssh -L 9999:localhost:9999 gcc76.fsffrance.org
[palves@home] $ time ./gdb -data-directory=data-directory -ex "tar rem :9999" -ex "b main" -ex "c" -ex "set confirm off" -ex "quit"
Pristine gdb 7.10.50.20150820-cvs gets us:
...
Remote debugging using :9999
Reading symbols from target:/home/palves/gdb/build/gdb/gdbserver/gdbserver...done.
Reading symbols from target:/lib64/ld-linux-x86-64.so.2...(no debugging symbols found)...done.
0x00007ffff7ddd190 in ?? () from target:/lib64/ld-linux-x86-64.so.2
Breakpoint 1 at 0x41200c: file ../../../src/gdb/gdbserver/server.c, line 3635.
Continuing.
Breakpoint 1, main (argc=1, argv=0x7fffffffe3d8) at ../../../src/gdb/gdbserver/server.c:3635
3635 ../../../src/gdb/gdbserver/server.c: No such file or directory.
/home/palves/gdb/build/gdb/gdbserver/gdbserver: No such file or directory.
real 0m54.803s
user 0m0.329s
sys 0m0.064s
While with the readahead cache added by this patch, it drops to:
real 0m29.462s
user 0m0.454s
sys 0m0.054s
I added a few counters to show cache hit/miss, and got:
readahead cache miss 142
readahead cache hit 310
Tested on x86_64 Fedora 20.
gdb/ChangeLog:
2015-08-21 Pedro Alves <palves@redhat.com>
* remote.c (struct readahead_cache): New.
(struct remote_state) <readahead_cache>: New field.
(remote_open_1): Invalidate the cache.
(readahead_cache_invalidate, readahead_cache_invalidate_fd): New
functions.
(remote_hostio_pwrite): Invalidate the readahead cache.
(remote_hostio_pread): Rename to ...
(remote_hostio_pread_vFile): ... this.
(remote_hostio_pread_from_cache): New function.
(remote_hostio_pread): Reimplement.
(remote_hostio_close): Invalidate the readahead cache.
Fixes implicit function declaration
error in gdb/procfs.c:4927 about undeclared
make_cleanup_close().
gdb/ChangeLog:
PR build/18843
* procfs.c: Include "filestuff.h".
These fields are currently used to track the location of the cursor
inside the command window. But their usefulness is questionable because
ncurses already internally keeps track of the location of the cursor,
whose coordinates we can query using the functions getyx(), getcurx() or
getcury(). It is an unnecessary pain to keep these fields in sync with
ncurses, and their meaning is not well-defined anyway. For instance, it
is not clear whether the coordinates held in these fields are
authoritative, or whether the coordinates reported by ncurses are.
So to keep things simple, this patch removes these fields and replaces
existing reads of these fields with calls to the appropriate ncurses
querying functions, and replaces writes to these fields with calls to
wmove() (when necessary and applicable).
In the function tui_cont_sig(), I removed the call to wmove() entirely
because moving to (start_line, curch) makes no sense. The move should
have been to (cur_line, curch) -- which would now be a no-op.
Tested on x86_64 Fedora 22, no obvious regressions.
gdb/ChangeLog:
* tui/tui-data.h (tui_command_info): Remove fields cur_line and
curch.
* tui/tui-data.c (tui_clear_win_detail) [CMD_WIN]: Don't set
cur_line or curch, instead call wmove().
(init_win_info) [CMD_WIN]: Likewise.
* tui/tui-io.c (tui_puts): Likewise. Don't read cur_line,
instead call getcury().
(tui_redisplay_readline): Don't set cur_line or curch.
(tui_mld_erase_entire_line): Don't read cur_line, instead call
getcury().
(tui_cont_sig): Remove call to wmove.
(tui_getc): Don't read cur_line or curch, instead call getcury()
or getyx(). Don't set curch.
* tui/tui-win.c (make_visible_with_new_height) [CMD_WIN]: Don't
set cur_line or curch. Always move cursor to (0,0).
Commit 221e1a37 (remote non-stop: Process initially stopped threads
before other commands) caused a test regression when testing with the
native-extended-gdbserver board:
FAIL: gdb.server/solib-list.exp: non-stop 1: non-stop interior stop (timeout)
This "interior stop" now happens before "target remote" prints the
prompt, so we should no longer explicitly expect it.
gdb/testsuite/ChangeLog:
2015-08-20 Pedro Alves <palves@redhat.com>
* gdb.server/solib-list.exp: No longer expect an interior stop in
non-stop mode.
The main motivation for this is making non-stop / all-stop behave
similarly on initial connection, in order to move in the direction of
reimplementing all-stop mode with the remote target always running in
non-stop mode.
When we connect to a remote target in non-stop mode, we may find
threads either running or already stopped. The act of connecting
itself does not force threads to stop. To handle that, the remote
non-stop connection is currently roughly like this:
#1 - Fetch list of remote threads (qXfer:threads:read, qfThreadInfo,
etc). All threads are assumed to be running until the target
reports an asynchronous stop reply for them.
#2 - Fetch the initial set of threads that were already stopped, with
the '?' packet. (In non-stop, this is coupled with the vStopped
mechanism to be able to retrieve the status of more than one
thread.)
The stop replies fetched in #2 are placed in the pending stop reply
queue, and left for the regular event loop to process. That is,
"target remote" finishes and returns _before_ those stops are
processed.
That means that it's possible to have GDB process further commands
before the initial set of stopped threads is reported to the user.
E.g., before the patch, note how the prompt is printed before the
frame:
Remote debugging using :9999
(gdb)
[Thread 15296] #1 stopped.
0x0000003615a011f0 in ?? ()
Even though thread #1 was not running, for a moment, the user can see
it as such:
$ gdb a.out -ex "set non-stop 1" -ex "tar rem :9999" -ex "info threads" -ex "info registers"
Remote debugging using :9999
Id Target Id Frame
* 1 Thread 4772 (running)
Target is executing. <<<<<<< info registers
(gdb)
[Thread 4772] #1 stopped.
0x0000003615a011f0 in ?? ()
To fix that, this commit makes gdb process all threads found already
stopped at connection time, before giving the prompt to the user.
The fix takes a cue from fork-child.c:startup_inferior [1], and
processes the events locally in remote.c, avoiding the whole
wait_for_inferior/handle_inferior_event path. I decided to try this
approach after noticing that:
- several cases in handle_inferior_event miss checking stop_soon.
- we don't want to fetch the thread list in normal_stop.
and trying to fix them was resulting in sprinkling stop_soon checks in
many places, and uglifying normal_stop even more.
While with this patch, I'm avoiding changing GDB's output other than
when the prompt is printed, I think this approach is more flexible if
we do want to change it. And also, it's likely easier to get rid of
the MI *running event that is still sent for threads that are
initially found stopped, if we want to.
This happens to fix the testsuite too. All non-stop tests are racy
against "target remote" / gdbserver testing currently. That is,
sometimes the tests run, but other times they're just skipped without
any indication of PASS/FAIL. When that happens, the logs show:
target remote localhost:2346
Remote debugging using localhost:2346
(gdb)
[Thread 25418] #1 stopped.
0x0000003615a011f0 in ?? ()
^CQuit
(gdb) Remote debugging from host 127.0.0.1
Killing process(es): 25418
monitor exit
(gdb) Remote connection closed
(gdb) testcase /home/pedro/gdb/mygit/build/../src/gdb/testsuite/gdb.threads/multi-create-ns-info-thr.exp completed in 61 seconds
The trouble here is that there's output after the prompt, and the
regex in question doesn't expect that:
-re "Remote debugging using .*$serialport_re.*$gdb_prompt $" {
verbose "Set target to $targetname"
return 0
}
[1] - before startup_inferior was added, we'd go through
wait_for_inferior/handle_inferior_event while going through the shell,
and that turned out problematic.
Tested on x86_64 Fedora 20, gdbserver.
gdb/ChangeLog:
2015-08-20 Pedro Alves <palves@redhat.com>
* infrun.c (print_target_wait_results): Make extern.
* infrun.h (print_target_wait_results): Declare.
* remote.c (set_stop_requested_callback): Delete.
(process_initial_stop_replies): New function.
(remote_start_remote): Use it.
(stop_reply_queue_length): New function.
gdb/testsuite/ChangeLog:
2015-08-20 Pedro Alves <palves@redhat.com>
* gdb.server/connect-stopped-target.c: New file.
* gdb.server/connect-stopped-target.exp: New file.
Here, in dwarfread.c:process_full_comp_unit:
/* Set symtab language to language from DW_AT_language. If the
compilation is from a C file generated by language preprocessors, do
not set the language if it was already deduced by start_subfile. */
if (!(cu->language == language_c
&& COMPUNIT_FILETABS (cust)->language != language_c))
COMPUNIT_FILETABS (cust)->language = cu->language;
in case start_subfile doesn't manage to deduce a language
COMPUNIT_FILETABS(cust)->language ends up as language_unknown, not
language_c. So the condition above evals false and we never set the
language from the cu's language.
gdb/ChangeLog:
2015-08-20 Pedro Alves <palves@redhat.com>
* dwarf2read.c (process_full_comp_unit): To tell whether
start_subfile managed to deduce a language, test for
language_unknown instead of language_c.
gdb/testsuite/ChangeLog:
2015-08-20 Pedro Alves <palves@redhat.com>
* gdb.dwarf2/comp-unit-lang.exp: New file.
* gdb.dwarf2/comp-unit-lang.c: New file.
Before this change, trying to evaluate the following Ada expression
yielded a syntax error, even though it's completely legal:
(gdb) p s'first = 'a'
Error in expression, near `'.
The problem lies in the lexer (gdb/ada-lex.l): at the point we reach "'a'",
we're still in the BEFORE_QUAL_QUOTE start condition (the mechanism to
distinguish character literals from other "tick" usages: qualified
expressions and attributes), so we consider that this quote is actually a
separate "tick".
This changes resets the start condition to INITIAL in the
{TICK}[a-zA-Z][a-zA-Z]+ rule (for attributes): attributes activate this
BEFORE_QUAL_QUOTE condition and in this case the above rule is always
executed rather than the <BEFORE_QUAL_QUOTE>"'" one (in flex, it's
always the longest match that is chosen). We now have instead:
(gdb) p s'first = 'a'
$1 = true
gdb/ChangeLog:
* ada-lex.l: Reset the start condition to INITIAL in the rule
that matches attributes.
gdb/testsuite/ChangeLog:
* gdb.ada/attr_ref_and_charlit.exp: New testcase.
* gdb.ada/attr_ref_and_charlit/foo.adb: New file.
Tested on x86_64-linux, no regression.
A symbol value in an ELF final linked binary is absolute, in contrast
to a relocatable object file where the value is section relative. For
--emit-relocs it is therefore incorrect to use the value of a section
symbol as the addend when adjusting relocs against input section
symbols to output section symbols.
PR gold/18846
* target-reloc.h (relocate_relocs <RELOC_ADJUST_FOR_SECTION_RELA>):
Subtract os->address() from addend.
* powerpc.cc (relocate_relocs): Likewise.
This change introduces a new function, dwarf2_string_attr(), which is
a wrapper for dwarf2_attr(). dwarf2read.c has been updated to
call dwarf2_string_attr in most instances where a string-valued
attribute is decoded to produce a string value. In most cases, it
simplifies the code; in some instances, the complexity of the code
remains unchanged.
I performed this change by looking for instances where the
result of DW_STRING was used in an assignment. Many of these
had a pattern which (roughly) looks something like this:
struct attribute *attr = NULL;
attr = dwarf2_attr (die, name, cu);
if (attr != NULL && DW_STRING (attr))
{
const char *str;
...
str = DW_STRING (attr);
... /* Use str in some fashion. */
}
Code of this form is transformed to look like this instead:
const char *str;
str = dwarf2_string_attr (die, name, cu)
if (str != NULL)
{
...
/* Use str in some fashion. */
...
}
In addition to invoking dwarf2_attr() and DW_STRING(),
dwarf2_string_attr() checks to make sure that the attribute's
`form' field matches one of DW_FORM_strp, DW_FORM_string, or
DW_FORM_GNU_strp_alt. If it does not match one of these forms,
it will return a NULL value in addition to calling complaint().
An earlier version of this patch did this type checking for one
particular instance where a string attribute was being decoded.
The situation that I was attempting to handle in that earlier patch is
this:
The Texas Instruments compiler uses the encoding for
DW_AT_MIPS_linkage_name for other purposes. TI uses the encoding,
0x2007, for TI_AT_TI_end_line which, unlike DW_AT_MIPS_linkage_name,
does not have a string-typed value. In this instance, GDB was attempting
to use an integer value as a string pointer, with predictable results.
(GDB would die with a segmentation fault.)
I've added a test which reproduces the problem that I was orignally
wanting to fix. It uses DW_AT_MIPS_linkage name with an associate
value which is a string, and again, where the value is a small
integer.
My test case causes GDB to segfault in an unpatched GDB. There
will be two PASSes in a patched GDB.
Unpatched GDB:
(gdb) ptype f
ERROR: Process no longer exists
UNRESOLVED: gdb.dwarf2/dw2-bad-mips-linkage-name.exp: ptype f
ERROR: Couldn't send ptype g to GDB.
UNRESOLVED: gdb.dwarf2/dw2-bad-mips-linkage-name.exp: ptype g
Patched GDB:
(gdb) ptype f
type = bool ()
(gdb) PASS: gdb.dwarf2/dw2-bad-mips-linkage-name.exp: ptype f
ptype g
type = bool ()
(gdb) PASS: gdb.dwarf2/dw2-bad-mips-linkage-name.exp: ptype g
I see no regressions on an x86_64 native target.
gdb/ChangeLog:
* dwarf2read.c (dwarf2_string_attr): New function.
(lookup_dwo_unit, process_psymtab_comp_unit_reader)
(dwarf2_compute_name, dwarf2_physname, find_file_and_directory)
(read_call_site_scope, namespace_name, guess_full_die_structure_name)
(anonymous_struct_prefix, prepare_one_comp_unit): Use
dwarf2_string_attr in place of dwarf2_attr and DW_STRING.
gdb/testsuite/ChangeLog:
* gdb.dwarf2/dw2-bad-mips-linkage-name.c: New file.
* gdb.dwarf2/dw2-bad-mips-linkage-name.exp: New file.
