binutils-gdb.git
10 years agoFix ChangeLog.
Mark Kettenis [Sat, 1 Mar 2014 11:03:30 +0000 (12:03 +0100)]
Fix ChangeLog.

10 years agoEnable rthreads support on OpenBSD/mips64.
Mark Kettenis [Sat, 1 Mar 2014 11:00:19 +0000 (12:00 +0100)]
Enable rthreads support on OpenBSD/mips64.

gdb/ChangeLog:

        * mips64obsd-nat.c: Include "obsd-nath".
        (_initialize_mips64obsd_nat): Call obsd_add_target instead of
        add_target
        * config/mips/obsd64.mh (NATDEPFILES): Add obsd-nat.o.

10 years agoEnable rthreads support on OpenBSD/amd64.
Mark Kettenis [Sat, 1 Mar 2014 10:49:58 +0000 (11:49 +0100)]
Enable rthreads support on OpenBSD/amd64.

gdb/ChangeLog:

        * amd64obsd-nat.c (_initialize_amd64obsd_nat): Call
        obsd_add_target instead of add_target.
        * config/i386/obsd64.mh (NATDEPFILES): Add obsd-nat.o.

10 years ago * config/tc-avr.c: Remove atxmega16x1.
Denis Chertykov [Sat, 1 Mar 2014 09:12:49 +0000 (13:12 +0400)]
* config/tc-avr.c: Remove atxmega16x1.

10 years agoRemove the unnecesary argument METHOD to valops.c:find_oload_champ.
Siva Chandra [Thu, 27 Feb 2014 13:51:46 +0000 (05:51 -0800)]
Remove the unnecesary argument METHOD to valops.c:find_oload_champ.

* valops.c (find_oload_champ): Remove unneccesary argument METHOD.
(find_overload_match): Update call to find_oload_champ.
(find_oload_champ_namespace_loop): Likewise

10 years agodaily update
Alan Modra [Fri, 28 Feb 2014 23:01:08 +0000 (09:31 +1030)]
daily update

10 years agoAdd sparc64obsd-nat.c to ALLDEPFILES.
Mark Kettenis [Fri, 28 Feb 2014 22:20:57 +0000 (23:20 +0100)]
Add sparc64obsd-nat.c to ALLDEPFILES.

gdb/ChangeLog:

        * Makefile.in (ALLDEPFILES): Add sparc64obsd-nat.c.

10 years agoEnable rthreads supports on OpenBSD/sparc64.
Mark Kettenis [Fri, 28 Feb 2014 22:06:14 +0000 (23:06 +0100)]
Enable rthreads supports on OpenBSD/sparc64.

gdb/ChangeLog:

        * configure.host (sparc64-*-openbsd*): Set gdb_host to obsd64.
        * config/sparc/obsd64.mh: New file.
        * sparc64obsd-nat.c: New file.

10 years agoStupid git never fucking does what I want!
Mark Kettenis [Fri, 28 Feb 2014 21:58:57 +0000 (22:58 +0100)]
Stupid git never fucking does what I want!

10 years agoSupport rthreads on OpenBSD 5.2 and later.
Mark Kettenis [Thu, 27 Feb 2014 22:23:46 +0000 (23:23 +0100)]
Support rthreads on OpenBSD 5.2 and later.

OpenBSD 5.2 and later have a proper threads implementation based on
kernel threads.  Debugging support is provided through additional
ptrace(2) requests, so this diff extends the generic code in
inf-ptrace.c with OpenBSD-specific code to discover additional threads.

gdb/ChangeLog:

        * obsd-nat.h: New file.
        * obsd-nat.c: New file.
        * Makefile.in (HFILES_NO_SRCDIR): Add obsd-nat.h.
        (ALLDEPFILES): Add obsd-nat.c.

10 years agoconstify ui_out_impl
Tom Tromey [Tue, 24 Dec 2013 03:53:23 +0000 (20:53 -0700)]
constify ui_out_impl

This patch constifies ui_out_impl in struct ui_out, and various
instances of ui_out_impl.

This removes a couple of FIXME comments (near cli_ui_out_impl and
mi_ui_out_impl) that did not make sense to me.

Tested by rebuilding.

2014-02-28  Tom Tromey  <tromey@redhat.com>

* cli-out.c (cli_ui_out_impl): Now const.  Remove comment.
* cli-out.h (cli_ui_out_impl): Now const.
* mi/mi-out.c (mi_ui_out_impl): Now const.  Remove comment.
* ui-out.c (struct ui_out) <impl>: Now const.
(default_ui_out_impl): Now const.
(ui_out_new): Make 'impl' parameter const.
* ui-out.h (ui_out_new): Update.

10 years agoIncorrect .debug_line prologue header length for 64-bit DWARF
Alan Modra [Fri, 28 Feb 2014 04:39:32 +0000 (15:09 +1030)]
Incorrect .debug_line prologue header length for 64-bit DWARF

Don't be clever, calculate the length directly as the difference of
two symbols.

* dwarf2dbg.c (out_debug_line): Correct .debug_line header_length
field for 64-bit dwarf.

10 years agoFix check_relocs/gc_sweep_hook mismatch
Alan Modra [Fri, 28 Feb 2014 04:08:27 +0000 (14:38 +1030)]
Fix check_relocs/gc_sweep_hook mismatch

PR ld/16643
* elflink.c (elf_gc_sweep): Call gc_sweep_hook for exactly
the same conditions we called check_relocs.

10 years agoRegenerate opcodes/msp430-decode.c
Alan Modra [Fri, 28 Feb 2014 04:02:26 +0000 (14:32 +1030)]
Regenerate opcodes/msp430-decode.c

Cosmetic changes.

* msp430-decode.c: Regenerate.

10 years agodaily update
Alan Modra [Thu, 27 Feb 2014 23:01:17 +0000 (09:31 +1030)]
daily update

10 years agoPrevent compiler warning.
Mark Kettenis [Thu, 27 Feb 2014 20:51:08 +0000 (21:51 +0100)]
Prevent compiler warning.

GCC 4.2.1 complains about first_l_name may be used uninitialized, and my brain
agrees.

gdb/ChangeLog:

        * solib-svr4.c (svr4_read_so_list): Initialize first_l_name to 0.

10 years agoStackGhost cookie is per-process.
Mark Kettenis [Thu, 27 Feb 2014 20:22:29 +0000 (21:22 +0100)]
StackGhost cookie is per-process.

gdb/ChangeLog:

        * sparc-nat.c (sparc_xfer_wcookie): Always use process ID.

10 years agoAdditional PR 8882 fix.
Jan Kratochvil [Thu, 27 Feb 2014 20:16:04 +0000 (21:16 +0100)]
Additional PR 8882 fix.

runtest gdb.base/corefile.exp

==23174== ERROR: AddressSanitizer: heap-use-after-free on address 0x604400008c88 at pc 0x68f0be bp 0x7fffae9d7490 sp
0x7fffae9d7480
READ of size 8 at 0x604400008c88 thread T0
    #0 0x68f0bd in svr4_read_so_list (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x68f0bd)
    #1 0x68f64e in svr4_current_sos_direct (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x68f64e)
    #2 0x68f757 in svr4_current_sos (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x68f757)
    #3 0xcebbff in update_solib_list (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xcebbff)
0x604400008c88 is located 8 bytes inside of 1104-byte region [0x604400008c80,0x6044000090d0)
freed by thread T0 here:
    #0 0x7f52677500f9 (/lib64/libasan.so.0+0x160f9)
    #1 0xd2c68a in xfree (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xd2c68a)
    #2 0xceb364 in free_so (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xceb364)
    #3 0xca59f8 in do_free_so (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xca59f8)
    #4 0x93432a in do_my_cleanups (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x93432a)
    #5 0x934406 in do_cleanups (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x934406)
    #6 0x68efa9 in svr4_read_so_list (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x68efa9)

I did not notice it during my review in:
Re: [PATCH v2] Skip vDSO when reading SO list (PR 8882)
https://sourceware.org/ml/gdb-patches/2013-09/msg00888.html

gdb/
2014-02-27  Jan Kratochvil  <jan.kratochvil@redhat.com>

Additional PR 8882 fix.
* solib-svr4.c (svr4_read_so_list): Change first to first_l_name.

Message-ID: <20140226220918.GA10431@host2.jankratochvil.net>

10 years agoopcodes/
Yufeng Zhang [Thu, 27 Feb 2014 15:06:56 +0000 (15:06 +0000)]
opcodes/

* aarch64-tbl.h (aarch64_opcode_table): Replace IMM0 with
FPIMM0 for fcmeq, fcmgt, fcmge, fcmlt and fcmle.

gas/testsuite/

* gas/aarch64/fp-const0-parsing.s: New test.
* gas/aarch64/fp-const0-parsing.d: Likewise.

10 years agoopcodes/
Yufeng Zhang [Thu, 27 Feb 2014 14:55:46 +0000 (14:55 +0000)]
opcodes/

* aarch64-opc.c (print_register_offset_address): Call
get_int_reg_name to prepare the register name.

gas/testsuite/

* gas/aarch64/ldst-reg-reg-offset.s: Add tests.
* gas/aarch64/ldst-reg-reg-offset.d: Update.

10 years agoLinux waitpid/__WALL emulation wrapper: If WNOHANG is set, don't touch sigprocmask.
Pedro Alves [Thu, 27 Feb 2014 14:44:16 +0000 (14:44 +0000)]
Linux waitpid/__WALL emulation wrapper: If WNOHANG is set, don't touch sigprocmask.

Just a small optimization.  No need to block/unblock signals if we're
not going to call sigsuspend.

gdb/
2014-02-27  Pedro Alves  <palves@redhat.com>

* nat/linux-waitpid.c (my_waitpid): Only block signals if WNOHANG
isn't set.

10 years agoThis patch adds support for ARM PLT entries that support a full 32-bit offset range.
Yuri Gribov [Thu, 27 Feb 2014 14:35:37 +0000 (14:35 +0000)]
This patch adds support for ARM PLT entries that support a full 32-bit offset range.
Enabled via the use of a new linker command line option: --long-plt.

* bfd-in.h: Add export of bfd_elf32_arm_use_long_plt.
* bfd-in2.h: Regenerate.
* elf32-arm.c (elf32_arm_plt_entry_long): New array.
(elf32_arm_link_hash_table_create): Set plt_entry_size to 16 if
using long PLT entries.
(bfd_elf32_arm_use_long_plt): New function.
(elf32_arm_populate_plt_entry): Add support for long PLT entries.

* emultempl/armelf.em (OPTION_LONG_PLT): Define.
(PARSE_AND_LIST_LONGOPTS): Add long-plt.
(PARSE_AND_LIST_OPTIONS): Likewise.
(PARSE_AND_LIST_ARGS_CASES): Handle long-plt.
* ld.texinfo: Document --long-plt.

* ld-arm/long-plt-format.s: New test case.
* ld-arm/long-plt-format.d: Expected disassembly.
* ld-arm/arm-elf.exp: Run the new test.

10 years agoTeach GDBserver's Linux backend about no unwaited-for children (TARGET_WAITKIND_NO_RE...
Pedro Alves [Thu, 27 Feb 2014 14:30:08 +0000 (14:30 +0000)]
Teach GDBserver's Linux backend about no unwaited-for children (TARGET_WAITKIND_NO_RESUMED).

GDBserver currently hangs forever in waitpid if the leader thread
exits before other threads, or if all resumed threads exit - e.g.,
next over a thread exit with sched-locking on.  This is exposed by
leader-exit.exp.  leader-exit.exp is part of a series of tests for a
set of related problems.  See
<http://www.sourceware.org/ml/gdb-patches/2011-10/msg00704.html>:

 "
 To recap, on the Linux kernel, ptrace/waitpid don't allow reaping the
 leader thread until all other threads in the group are reaped.  When
 the leader exits, it goes zombie, but waitpid will not return an exit
 status until the other threads are gone.  This is presently exercised
 by the gdb.threads/leader-exit.exp test.  The fix for that test, in
 linux-nat.c:wait_lwp, handles the case where we see the leader gone
 when we're stopping all threads to report an event to some other
 thread to the core.

 (...)

 The latter bit about not blocking if there no resumed threads in the
 process also applies to some other thread exiting, not just the main
 thread.  E.g., this test starts a thread, and runs to a breakpoint in
 that thread:

 ...
 (gdb) c
 Continuing.
 [New Thread 0x7ffff75a4700 (LWP 23397)]
 [Switching to Thread 0x7ffff75a4700 (LWP 23397)]

 Breakpoint 2, thread_a (arg=0x0) at ../../../src/gdb/testsuite/gdb.threads/no-unwaited-for-left.c:28
 28        return 0; /* break-here */
 (gdb) info threads
 * 2 Thread 0x7ffff75a4700 (LWP 23397)  thread_a (arg=0x0) at ../../../src/gdb/testsuite/gdb.threads/no-unwaited-for-left.c:28
   1 Thread 0x7ffff7fcb720 (LWP 23391)  0x00007ffff7bc606d in pthread_join (threadid=140737343276800, thread_return=0x0) at pthread_join.c:89

 The thread will exit as soon as we resume it.  But if we only resume
 that thread, leaving the rest of the threads stopped:

 (gdb) set scheduler-locking on
 (gdb) c
 Continuing.
 ^C^C^C^C^C^C^C^C
 "

This patch fixes the issues by implementing TARGET_WAITKIND_NO_RESUMED
on GDBserver, similarly to what the patch above did for native
Linux GDB.

gdb.threads/leader-exit.exp now passes.

gdb.threads/no-unwaited-for-left.exp now at least errors out instead
of hanging:

 continue
 Continuing.
 warning: Remote failure reply: E.No unwaited-for children left.

 [Thread 15454] #1 stopped.
 0x00000034cf408e60 in pthread_join (threadid=140737353922368, thread_return=0x0) at pthread_join.c:93
 93          lll_wait_tid (pd->tid);
 (gdb) FAIL: gdb.threads/no-unwaited-for-left.exp: continue stops when the main thread exits

The gdb.threads/non-ldr-exc-*.exp tests are skipped because GDBserver
unfortunately doesn't support fork/exec yet, but I'm confident this
fixes the related issues.

I'm leaving modeling TARGET_WAITKIND_NO_RESUMED in the RSP for a
separate pass.

(BTW, in case of error in response to a vCont, it would be better for
GDB to query the target for the current thread, or re-select one,
instead of assuming current inferior_ptid is still the selected
thread.)

This implementation is a little different from GDB's, because I'm
avoiding bringing in more of this broken use of waitpid(PID) into
GDBserver.  Specifically, this avoids waitpid(PID) when stopping all
threads.  There's really no need for wait_for_sigstop to wait for each
LWP in turn.  Instead, with some refactoring, we make it reuse
linux_wait_for_event.

gdb/gdbserver/
2014-02-27  Pedro Alves  <palves@redhat.com>

PR 12702
* inferiors.h (A_I_NEXT, ALL_INFERIORS_TYPE, ALL_PROCESSES): New
macros.
* linux-low.c (delete_lwp, handle_extended_wait): Add debug
output.
(last_thread_of_process_p): Take a PID argument instead of a
thread pointer.
(linux_wait_for_lwp): Delete.
(num_lwps, check_zombie_leaders, not_stopped_callback): New
functions.
(linux_low_filter_event): New function, party factored out from
linux_wait_for_event.
(linux_wait_for_event): Rename to ...
(linux_wait_for_event_filtered): ... this.  Add new filter ptid
argument.  Partly rewrite.  Always use waitpid(-1, WNOHANG) and
sigsuspend.  Check for zombie leaders.
(linux_wait_for_event): Reimplement as wrapper around
linux_wait_for_event_filtered.
(linux_wait_1): Handle TARGET_WAITKIND_NO_RESUMED.  Assume that if
a normal or signal exit is seen, it's the whole process exiting.
(wait_for_sigstop): No longer a for_each_inferior callback.
Rewrite on top of linux_wait_for_event_filtered.
(stop_all_lwps): Call wait_for_sigstop directly.
* server.c (resume, handle_target_event): Handle
TARGET_WAITKIND_NO_RESUMED.

10 years agoMove linux-nat.c:status_to_str to nat/linux-waitpid.c.
Pedro Alves [Thu, 27 Feb 2014 14:30:08 +0000 (14:30 +0000)]
Move linux-nat.c:status_to_str to nat/linux-waitpid.c.

So that gdbserver's Linux backend can use it too.

gdb/
2014-02-27  Pedro Alves  <palves@redhat.com>

PR 12702
* linux-nat.c (status_to_str): Moved to nat/linux-waitpid.c.
* nat/linux-waitpid.c: Include string.h.
(status_to_str): Moved here and made extern.
* nat/linux-waitpid.h (status_to_str): New declaration.

10 years agoMove ptid_match to common/ptid.c.
Hui Zhu [Thu, 27 Feb 2014 14:30:07 +0000 (14:30 +0000)]
Move ptid_match to common/ptid.c.

So that gdbserver can use it too.

gdb/
2014-02-27  Hui Zhu  <hui@codesourcery.com>

PR 12702
* infrun.c (ptid_match): Move ...
* common/ptid.c (ptid_match): ... here.
* inferior.h (ptid_match): Move ...
* common/ptid.h (ptid_match): ... here.

10 years agoThis patch adds a default manifest in to the final links performed by the Cygwin...
Nick Clifton [Thu, 27 Feb 2014 14:09:29 +0000 (14:09 +0000)]
This patch adds a default manifest in to the final links performed by the Cygwin and MinGW targets.
The manifest is necessary in order for the linked binaries to be executed in a Windows 8 environment.

The manifest is added using a linker script so that this feature will be compiler-neutral.  The resource
merging code in the linker means that if an application provides its own manifest then the default
manifest will be ignored.

* configure.in (all_emul_extra_binaries): New variable.  Populated
by invoking configure.tgt.
(EMUL_EXTRA_BINARIES): New substitution.
* configure: Regenerate.
* configure.tgt (target_extra_binaries): New variable.  Set to
default-manifest.o for Cygwin and MinGW targets.
* Makefile.am (EMUL_EXTRA_BINARIES): New variable.  Initialised
by the configure script.
(ALL_EMUL_EXTRA_BINARIES): New variable.
(default-manifest.o): New rule to build the default manifest.
(ld_new_DEPENDENCIES): Add EMUL_EXTRA_BINARIES.
(install-data-local): Add EMUL_EXTRA_BINARIES.
* Makefile.in: Regenerate.
* ld.texinfo: Document default manifest support.
* emulparams/i386pe.sh (DEFAULT_MANIFEST): Define.
* emulparams/i386pep.sh (DEFAULT_MANIFEST): Define.
* emultempl/default-manifest.rc: New file.
* scripttempl/pe.sc (R_RSRC): Include DEFAULT_MANIFEST, if defined.
* scripttempl/pep.sc (R_RSRC): Likewise.

* ld-pe/longsecn-1.d: Allow for extra sections.
* ld-pe/longsecn-2.d: Likewise.
* ld-pe/longsecn.d: Likewise.
* ld-pe/secrel.d: Likewise.

10 years agoFix ELF ppc32 targets that don't use ppc32elf.em
Alan Modra [Thu, 27 Feb 2014 13:01:35 +0000 (23:31 +1030)]
Fix ELF ppc32 targets that don't use ppc32elf.em

5446cbdf82892a800ed7eef563a795e75223ec52 broke powerpc-lynxos,
powerpc-netware, powerpc-windiss and powerpc-vxworks.

bfd/
* elf32-ppc.c (ppc_elf_link_hash_table_create): Provide default
params for targets that don't use ppc32elf.em.
ld/
* emulparams/elf32ppcvxworks.sh: Source plt_unwind.sh and
use ppc32elf.em.
* emultempl/ppc32elf.em (ppc_after_open): Don't compile for
vxworks.
(LDEMUL_AFTER_OPEN): Don't set for vxworks.
(PARSE_AND_LIST_LONGOPTS, PARSE_AND_LIST_OPTIONS): Exclude
-secure-plt, -bss-plt and -sdata-got when vxworks.

10 years agoCall common OpenBSD ABI init code on OpenBSD/mips64.
Mark Kettenis [Thu, 27 Feb 2014 13:07:10 +0000 (14:07 +0100)]
Call common OpenBSD ABI init code on OpenBSD/mips64.

        * mips64obsd-tdep.c (mips64obsd_init_abi): Call obsd_init_abi.
        * configure.tgt (mips64*-*-openbsd*): Add obsd-tdep.c to
        gdb_target_obs.

10 years agoAdd gdbarch auxv parsing for OpenBSD.
Mark Kettenis [Thu, 27 Feb 2014 12:47:00 +0000 (13:47 +0100)]
Add gdbarch auxv parsing for OpenBSD.

gdb/Changelog:

        * obsd-tdep.c (obsd_auxv_parse): New function.
        (obsd_init_abi): Set auxv_parse.

10 years agoAdd auxv parsing to the architecture vector.
Mark Kettenis [Thu, 27 Feb 2014 12:40:15 +0000 (13:40 +0100)]
Add auxv parsing to the architecture vector.

Necessary to fix parsing auxv entries from core files on systems that use
the layout specified by ELF instead of the incompatible variant used by Linux.

gdb/Changelog:

        * gdbarch.sh (auxv_parse): New.
        * gdbarch.h: Regenerated.
        * gdbarch.c: Regenerated.
        * auxv.c (target_auxv_parse): Call gdbarch_parse_auxv if provided.

10 years agodaily update
Alan Modra [Wed, 26 Feb 2014 23:01:09 +0000 (09:31 +1030)]
daily update

10 years agoguile: Add 'history-append!' procedure.
Ludovic Courtès [Tue, 18 Feb 2014 22:35:46 +0000 (23:35 +0100)]
guile: Add 'history-append!' procedure.

gdb/
2014-02-26  Ludovic Courtès  <ludo@gnu.org>

* guile/scm-value.c (gdbscm_history_append_x): New function.
(value_functions): Add it.

gdb/testsuite/
2014-02-26  Ludovic Courtès  <ludo@gnu.org>

* gdb.guile/scm-value.exp (test_value_in_inferior): Add
test for 'history-append!'.

gdb/doc/
2014-02-26  Ludovic Courtès  <ludo@gnu.org>

* gdb/doc/guile.texi (Basic Guile): Document 'history-append!'.

10 years agogdbserver/Windows: Rely purely on event info when handling DLL load event
Joel Brobecker [Mon, 24 Feb 2014 23:21:14 +0000 (15:21 -0800)]
gdbserver/Windows: Rely purely on event info when handling DLL load event

This is the GDBserver counterpart of a change we recently made in
GDB to only rely on get_image_name to determine its name.

This simplification, in turn, allows us to remove a fair amount of
functions and globals which now become unused.

gdb/gdbserver/ChangeLog:

* win32-low.c (psapi_get_dll_name,
* win32_CreateToolhelp32Snapshot): Delete.
(win32_CreateToolhelp32Snapshot, win32_Module32First)
(win32_Module32Next, load_toolhelp, toolhelp_get_dll_name):
Delete.
(handle_load_dll): Add function description.
Remove code using psapi_get_dll_name and toolhelp_get_dll_name.

10 years agowindows: Factorize handling of DLL load address offset
Joel Brobecker [Mon, 24 Feb 2014 23:21:13 +0000 (15:21 -0800)]
windows: Factorize handling of DLL load address offset

This patch is a small cleanup that moves the magic 0x1000 offset
to apply to a DLL's base address inside the win32_add_one_solib
function, rather than delegate that reponsibility to its callers.

gdb/gdbserver/ChangeLog:

* win32-low.c (win32_add_one_solib): Add 0x1000 to load_addr.
Add comment.
(win32_add_all_dlls): Remove 0x1000 offset applied to DLL
base address when calling win32_add_one_solib.
(handle_load_dll): Delete local variable load_addr.
Remove 0x1000 offset applied to DLL base address when calling
win32_add_one_solib.
(handle_unload_dll): Add comment.

10 years agogdbserver/windows: Ignore DLL load/unload events during child initialization.
Joel Brobecker [Mon, 24 Feb 2014 23:21:12 +0000 (15:21 -0800)]
gdbserver/windows: Ignore DLL load/unload events during child initialization.

This GDBserver patch mirrors a change made in GDB wich aims at
simplifying DLL handling during the inferior initialization
(process creation during the "run", or during an "attach").
Instead of processing each DLL load event, which is sometimes
incomplete, we ignore these events until the inferior has completed
its startup phase, and then just iterate over all DLLs via
EnumProcessModules.

As a side-effect, it fixes a small bug where win32_ensure_ntdll_loaded
was missing a 0x1000 offset in the DLL base address. This problem
should only be visible on the 64bit version of Windows 8.1, since
this is the only platform where win32_ensure_ntdll_loaded is actually
needed.

gdb/gdbserver/ChangeLog:

* win32-low.c (win32_add_all_dlls): Renames
win32_ensure_ntdll_loaded.  Rewrite function documentation.
Adjust implementation to always load all DLLs.
Add 0x1000 offset to DLL base address when calling
win32_add_one_solib.
(child_initialization_done): New static global.
(do_initial_child_stuff): Set child_initialization_done to
zero during child initialization, and 1 after.  Replace call
to win32_ensure_ntdll_loaded by call to win32_add_all_dlls.
Add comment.
(match_dll_by_basename, dll_is_loaded_by_basename): Delete.
(handle_unload_dll): Add function documentation.
(get_child_debug_event): Ignore load and unload DLL events
during child initialization.

10 years agoDWARF: Read constant-class addresses correctly
Joel Brobecker [Sat, 15 Feb 2014 15:09:58 +0000 (19:09 +0400)]
DWARF: Read constant-class addresses correctly

Starting with DWARF version 4, the description of the DW_AT_high_pc
attribute was amended to say:

   if it is of class constant, the value is an unsigned integer offset
   which when added to the low PC gives the address of the first
   location past the last instruction associated with the entity.

A change was made in Apr 27th, 2012 to reflect that change:

  | commit 91da14142c0171e58a91ad58a32fd010b700e761
  | Author: Mark Wielaard <mjw@redhat.com>
  | Date:   Fri Apr 27 18:55:19 2012 +0000
  |
  |     * dwarf2read.c (dwarf2_get_pc_bounds): Check DW_AT_high_pc form to
  |     see whether it is an address or a constant offset from DW_AT_low_pc.
  |     (dwarf2_record_block_ranges): Likewise.
  |     (read_partial_die): Likewise.

Unfortunately, this new interpretation is now used regardless of
the CU's DWARF version. It turns out that one of WindRiver's compilers
(FTR: Diabdata 4.4) is generating DWARF version 2 info with
DW_AT_high_pc attributes improperly using the data4 form. Because of
that, we miscompute all high PCs incorrectly. This leads to a lot of
symtabs having overlapping ranges, which in turn causes havoc in
pc-to-symtab-and-line translations.

One visible effect is when inserting a breakpoint on a given function:

    (gdb) b world
    Breakpoint 1 at 0x4005c4

The source location of the breakpoint is missing. The output should be:

    (gdb) b world
    Breakpoint 1 at 0x4005c8: file dw2-rel-hi-pc-world.c, line 24.

What happens in this case is that the pc-to-SAL translation first
starts be trying to find the symtab associated to our PC using
each symtab's ranges. Because of the high_pc miscomputation,
many symtabs end up matching, and the heuristic trying to select
the most probable one unfortunately returns one that is unrelated
(it really had no change in this case to do any better). Once we
have the wrong symtab, the start searching the associated linetable,
where the addresses are correct, thus finding no match, and therefore
no SAL.

This patch is an attempt at handling the situation as gracefully
as we can, without guarantees.  It introduces a new function
"attr_value_as_address" which uses the correct accessor for getting
the value of a given attribute.  It then adjust the code throughout
this unit to use this function instead of assuming that addresses always
have the DW_FORM_addr format.

It also fixes the original issue of miscomputing the high_pc
by limiting the new interpretation of constant form DW_AT_high_pc
attributes to units using DWARF version 4 or later.

gdb/ChangeLog:

        * dwarf2read.c (attr_value_as_address): New function.
        (dwarf2_find_base_address, read_call_site_scope): Use
        attr_value_as_address in place of DW_ADDR.
        (dwarf2_get_pc_bounds): Use attr_value_as_address to get
        the low and high addresses.  Slight rework of the handling
        of the high pc being a constant form, and limit it to
        DWARF verson 4 or higher.
        (dwarf2_record_block_ranges): Likewise.
        (read_partial_die): Likewise.
        (new_symbol_full): Use attr_value_as_address in place of DW_ADDR.

gdb/testsuite/ChangeLog:

        * gdb.dwarf2/dw2-abs-hi-pc-hello-dbg.S: New file.
        * gdb.dwarf2/dw2-abs-hi-pc-hello.c: New file.
        * gdb.dwarf2/dw2-abs-hi-pc-world-dbg.S: New file.
        * gdb.dwarf2/dw2-abs-hi-pc-world.c: New file.
        * gdb.dwarf2/dw2-abs-hi-pc.c: New file.
        * gdb.dwarf2/dw2-abs-hi-pc.exp: New file.

Tested on x86_64-linux.

10 years agomake "file" use the BFD cache better
Tom Tromey [Thu, 17 Oct 2013 18:03:06 +0000 (12:03 -0600)]
make "file" use the BFD cache better

Right now the "file" command will discard the exec_bfd and then
possibly open a new one.

If this ends up reopening the same file, it can cause needless work by
gdb -- destroying all the per-BFD data just to re-read it again.

This patch changes the code to hold a reference to the old exec_bfd
while opening the new one.

The possible downside of this is a higher peak memory use.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* exec.c (exec_file_attach): Hold a reference to exec_bfd.

10 years agoadd short-circuit logic to elfread.c
Tom Tromey [Tue, 15 Oct 2013 17:50:58 +0000 (11:50 -0600)]
add short-circuit logic to elfread.c

If minimal symbols have already been read into a per-BFD object, then
a symbol reader can skip re-reading them.  This changes the ELF reader
to do so.

We only skip the work if the file is ELF+DWARF.  If it has stabs or
mdebug sections, then I think extra information is computed during the
minsym creation pass; and so we must still repeat it.  Eventually even
this will go away, once all symbol types have switched to being
progspace-independent.  In the meantime this has no negative effect --
it is just a missing optimization for a small set of users.

This change also required a somewhat non-obvious change to the OBJSTAT
accounting code.  If a symbol reader skips re-reading minimal symbols,
then the corresponding OBJSTAT will not be updated.  This leads to a
test failure in gdb.base/maint.exp.

To fix this, I've moved the needed stat field out of objfile and into
the per-BFD object.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* elfread.c (elf_read_minimal_symbols): Return early if
minimal symbols have already been read.  Add "ei" parameter.
(elf_symfile_read): Call elf_read_minimal_symbols earlier.
* minsyms.c (prim_record_minimal_symbol_full): Update.
* objfiles.h (struct objstats) <n_minsyms>: Move...
(struct objfile_per_bfd_storage) <n_minsyms>: ... here.
* symmisc.c (print_objfile_statistics): Update.

10 years agosplit out elf_read_minimal_symbols
Tom Tromey [Tue, 15 Oct 2013 17:43:46 +0000 (11:43 -0600)]
split out elf_read_minimal_symbols

This is just a simple refactoring in elfread.c to split out the
minsym-reading code into its own function.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* elfread.c (elf_read_minimal_symbols): New function, from
elf_symfile_read.
(elf_symfile_read): Call it.

10 years agomove minimal symbols to per-bfd
Tom Tromey [Tue, 15 Oct 2013 17:37:19 +0000 (11:37 -0600)]
move minimal symbols to per-bfd

Now that minimal symbols are independent of the program space, we can
move them to the per-BFD object.  This lets us save memory in the
multi-inferior case; and, once the symbol readers are updated, time.

The other prerequisite for this move is that all the objects referred
to by the minimal symbols have a lifetime at least as long as the
per-BFD object.  I think this is satisfied partially by this patch
(moving the copied names there) and partially by earlier patches
moving the demangled name hash.

This patch contains a bit of logic to avoid creating new minimal
symbols if they have already been read for a given BFD.  This allows
us to avoid trying to update all the symbol readers for this
condition.  At first glance this may seem like a hack, but some symbol
readers mix psym and minsym reading, and would require logic just like
this regardless -- and it is simpler and less error-prone to just do
the work in a central spot.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* minsyms.c (lookup_minimal_symbol, iterate_over_minimal_symbols)
(lookup_minimal_symbol_text, lookup_minimal_symbol_by_pc_name)
(lookup_minimal_symbol_solib_trampoline)
(lookup_minimal_symbol_by_pc_section_1)
(lookup_minimal_symbol_and_objfile): Update.
(prim_record_minimal_symbol_full): Use the per-BFD obstack.
Don't allocate a minimal symbol if minsyms have already been read.
(build_minimal_symbol_hash_tables): Update.
(install_minimal_symbols): Do nothing if minsyms already read.
Use the per-BFD obstack.
(terminate_minimal_symbol_table): Use the per-BFD obstack.
* objfiles.c (allocate_objfile): Call
terminate_minimal_symbol_table later.
(have_minimal_symbols): Update.
* objfiles.h (struct objfile_per_bfd_storage) <msymbols,
minimal_symbol_count, msymbol_hash, msymbol_demangled_hash>:
Move from struct objfile.
<minsyms_read>: New field.
(struct objfile) <msymbols, minimal_symbol_count,
msymbol_hash, msymbol_demangled_hash>: Move.
(ALL_OBJFILE_MSYMBOLS): Update.
* symfile.c (read_symbols): Set minsyms_read.
(reread_symbols): Update.
* symmisc.c (dump_objfile, dump_msymbols): Update.

10 years agochange minsyms not to be relocated at read-time
Tom Tromey [Tue, 15 Oct 2013 19:28:57 +0000 (13:28 -0600)]
change minsyms not to be relocated at read-time

This removes the runtime offsets from minsyms.  Instead, these offsets
will now be applied whenever the minsym's address is computed.

This patch redefines MSYMBOL_VALUE_ADDRESS to actually use the offsets
from the given objfile.  Then, it updates all the symbol readers,
changing them so that they do not add in the section offset when
creating the symbol.

This change also lets us remove relocation of minsyms from
objfile_relocate1 and also msymbols_sort.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* minsyms.c (msymbols_sort): Remove.
* minsyms.h (msymbols_sort): Remove.
* objfiles.c (objfile_relocate1): Don't relocate minsyms.
* symtab.h (MSYMBOL_VALUE_ADDRESS): Use objfile offsets.
* elfread.c (elf_symtab_read): Don't add section offsets.
* xcoffread.c (record_minimal_symbol): Don't add section offset
to minimal symbol address.
* somread.c (text_offset, data_offset): Remove.
(som_symtab_read): Don't add section offsets to minimal symbol
addresses.
* coff-pe-read.c (add_pe_forwarded_sym, read_pe_exported_syms):
Don't add section offsets to minimal symbols.
* coffread.c (coff_symtab_read): Don't add section offsets
to minimal symbol addresses.
* machoread.c (macho_symtab_add_minsym): Don't add section offset
to minimal symbol addresses.
* mipsread.c (read_alphacoff_dynamic_symtab): Don't add
section offset to minimal symbol addresses.
* mdebugread.c (parse_partial_symbols): Don't add section
offset to minimal symbol addresses.
* dbxread.c (read_dbx_dynamic_symtab): Don't add section
offset to minimal symbol addresses.

10 years agostart change to progspace independence
Tom Tromey [Thu, 15 Aug 2013 14:46:35 +0000 (08:46 -0600)]
start change to progspace independence

This patch starts changing minimal symbols to be independent of the
program space.

Specifically, it adds a new objfile parameter to MSYMBOL_VALUE_ADDRESS
and changes all the code to use it.  This is needed so we can change
gdb to apply the section offset when a minsym's address is computed,
as opposed to baking the offsets into the symbol itself.

A few spots still need the unrelocated address.  For these, we
introduce MSYMBOL_VALUE_RAW_ADDRESS.

As a convenience, we also add the new macro BMSYMBOL_VALUE_ADDRESS,
which computes the address of a bound minimal symbol.  This just does
the obvious thing with the fields.

Note that this change does not actually enable program space
independence.  That requires more changes to gdb.  However, to ensure
that these changes compile properly, this patch does add the needed
section lookup code to MSYMBOL_VALUE_ADDRESS -- it just ensures it has
no effect at runtime by multiplying the offset by 0.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* ada-lang.c (ada_main_name): Update.
(ada_add_standard_exceptions): Update.
* ada-tasks.c (ada_tasks_inferior_data_sniffer): Update.
* aix-thread.c (pdc_symbol_addrs, pd_enable): Update.
* arm-tdep.c (skip_prologue_function, arm_skip_stub): Update.
* auxv.c (ld_so_xfer_auxv): Update.
* avr-tdep.c (avr_scan_prologue): Update.
* ax-gdb.c (gen_var_ref): Update.
* blockframe.c (get_pc_function_start)
(find_pc_partial_function_gnu_ifunc): Update.
* breakpoint.c (create_overlay_event_breakpoint)
(create_longjmp_master_breakpoint)
(create_std_terminate_master_breakpoint)
(create_exception_master_breakpoint): Update.
* bsd-uthread.c (bsd_uthread_lookup_address): Update.
* c-valprint.c (c_val_print): Update.
* coff-pe-read.c (add_pe_forwarded_sym): Update.
* common/agent.c (agent_look_up_symbols): Update.
* dbxread.c (find_stab_function_addr, end_psymtab): Update.
* dwarf2loc.c (call_site_to_target_addr): Update.
* dwarf2read.c (dw2_find_pc_sect_symtab): Update.
* elfread.c (elf_gnu_ifunc_record_cache)
(elf_gnu_ifunc_resolve_by_got): Update.
* findvar.c (default_read_var_value): Update.
* frame.c (inside_main_func): Update.
* frv-tdep.c (frv_frame_this_id): Update.
* glibc-tdep.c (glibc_skip_solib_resolver): Update.
* gnu-v3-abi.c (gnuv3_get_typeid, gnuv3_skip_trampoline):
Update.
* hppa-hpux-tdep.c (hppa64_hpux_search_dummy_call_sequence)
(hppa_hpux_find_dummy_bpaddr): Update.
* hppa-tdep.c (hppa_symbol_address): Update.
* infcmd.c (until_next_command): Update.
* jit.c (jit_read_descriptor, jit_breakpoint_re_set_internal):
Update.
* linespec.c (minsym_found, add_minsym): Update.
* linux-nat.c (get_signo): Update.
* linux-thread-db.c (inferior_has_bug): Update.
* m32c-tdep.c (m32c_return_value)
(m32c_m16c_address_to_pointer): Update.
* m32r-tdep.c (m32r_frame_this_id): Update.
* m68hc11-tdep.c (m68hc11_get_register_info): Update.
* machoread.c (macho_resolve_oso_sym_with_minsym): Update.
* maint.c (maintenance_translate_address): Update.
* minsyms.c (lookup_minimal_symbol_by_pc_name): Update.
(frob_address): New function.
(lookup_minimal_symbol_by_pc_section_1): Use raw addresses,
frob_address.  Rename parameter to "pc_in".
(compare_minimal_symbols, compact_minimal_symbols): Use raw
addresses.
(find_solib_trampoline_target, minimal_symbol_upper_bound):
Update.
* mips-linux-tdep.c (mips_linux_skip_resolver): Update.
* mips-tdep.c (mips_skip_pic_trampoline_code): Update.
* objc-lang.c (find_objc_msgsend): Update.
* objfiles.c (objfile_relocate1): Update.
* obsd-tdep.c (obsd_skip_solib_resolver): Update.
* p-valprint.c (pascal_val_print): Update.
* parse.c (write_exp_msymbol): Update.
* ppc-linux-tdep.c (ppc_linux_spe_context_lookup)
(ppc_elfv2_skip_entrypoint): Update.
* ppc-sysv-tdep.c (convert_code_addr_to_desc_addr): Update.
* printcmd.c (build_address_symbolic, msym_info)
(address_info): Update.
* proc-service.c (ps_pglobal_lookup): Update.
* psymtab.c (find_pc_sect_psymtab_closer)
(find_pc_sect_psymtab, find_pc_sect_symtab_from_partial):
Change msymbol parameter to bound_minimal_symbol.
* ravenscar-thread.c (get_running_thread_id): Update.
* remote.c (remote_check_symbols): Update.
* sh64-tdep.c (sh64_elf_make_msymbol_special): Use raw
address.
* sol2-tdep.c (sol2_skip_solib_resolver): Update.
* solib-dsbt.c (lm_base): Update.
* solib-frv.c (lm_base, main_got): Update.
* solib-irix.c (locate_base): Update.
* solib-som.c (som_solib_create_inferior_hook)
(link_map_start): Update.
* solib-spu.c (spu_enable_break, ocl_enable_break): Update.
* solib-svr4.c (elf_locate_base, enable_break): Update.
* spu-tdep.c (spu_get_overlay_table, spu_catch_start)
(flush_ea_cache): Update.
* stabsread.c (define_symbol, scan_file_globals): Update.
* stack.c (find_frame_funname): Update.
* symfile-debug.c (debug_qf_expand_symtabs_matching)
(debug_qf_find_pc_sect_symtab): Update.
* symfile.c (simple_read_overlay_table)
(simple_overlay_update): Update.
* symfile.h (struct quick_symbol_functions)
<find_pc_sect_symtab>: Change type of msymbol to
bound_minimal_symbol.
* symmisc.c (dump_msymbols): Update.
* symtab.c (find_pc_sect_symtab_via_partial)
(find_pc_sect_psymtab, find_pc_sect_line, skip_prologue_sal)
(search_symbols, print_msymbol_info): Update.
* symtab.h (MSYMBOL_VALUE_RAW_ADDRESS): New macro.
(MSYMBOL_VALUE_ADDRESS): Redefine.
(BMSYMBOL_VALUE_ADDRESS): New macro.
* tracepoint.c (scope_info): Update.
* tui/tui-disasm.c (tui_find_disassembly_address)
(tui_get_begin_asm_address): Update.
* valops.c (find_function_in_inferior): Update.
* value.c (value_static_field, value_fn_field): Update.

10 years agouse bound_minsym as result for lookup_minimal_symbol et al
Tom Tromey [Tue, 15 Oct 2013 01:53:29 +0000 (19:53 -0600)]
use bound_minsym as result for lookup_minimal_symbol et al

This patch changes a few minimal symbol lookup functions to return a
bound_minimal_symbol rather than a pointer to the minsym.  This change
helps prepare gdb for computing a minimal symbol's address at the
point of use.

Note that this changes even those functions that ostensibly search a
single objfile.  That was necessary because, in fact, those functions
can search an objfile and its separate debug objfiles; and it is
important for the caller to know in which objfile the minimal symbol
was actually found.

The bulk of this patch is mechanical.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* ada-lang.c (ada_update_initial_language): Update.
(ada_main_name, ada_has_this_exception_support): Update.
* ada-tasks.c (ada_tasks_inferior_data_sniffer): Update.
* aix-thread.c (pdc_symbol_addrs, pd_enable): Update.
* arm-tdep.c (arm_skip_stub): Update.
* auxv.c (ld_so_xfer_auxv): Update.
* avr-tdep.c (avr_scan_prologue): Update.
* ax-gdb.c (gen_var_ref): Update.
* breakpoint.c (struct breakpoint_objfile_data)
<overlay_msym, longjmp_msym, terminate_msym, exception_msym>: Change
type to bound_minimal_symbol.
(create_overlay_event_breakpoint)
(create_longjmp_master_breakpoint)
(create_std_terminate_master_breakpoint)
(create_exception_master_breakpoint): Update.
* bsd-uthread.c (bsd_uthread_lookup_address): Update.
* c-exp.y (classify_name): Update.
* coffread.c (coff_symfile_read): Update.
* common/agent.c (agent_look_up_symbols): Update.
* d-lang.c (d_main_name): Update.
* dbxread.c (find_stab_function_addr, end_psymtab): Update.
* dec-thread.c (enable_dec_thread): Update.
* dwarf2loc.c (call_site_to_target_addr): Update.
* elfread.c (elf_gnu_ifunc_resolve_by_got): Update.
* eval.c (evaluate_subexp_standard): Update.
* findvar.c (struct minsym_lookup_data) <result>: Change type
to bound_minimal_symbol.
<objfile>: Remove.
(minsym_lookup_iterator_cb, default_read_var_value): Update.
* frame.c (inside_main_func): Update.
* frv-tdep.c (frv_frame_this_id): Update.
* gcore.c (call_target_sbrk): Update.
* glibc-tdep.c (glibc_skip_solib_resolver): Update.
* gnu-v3-abi.c (gnuv3_get_typeid, gnuv3_skip_trampoline):
Update.
* go-lang.c (go_main_name): Update.
* hppa-hpux-tdep.c (hppa_hpux_skip_trampoline_code)
(hppa_hpux_find_import_stub_for_addr): Update.
* hppa-tdep.c (hppa_extract_17, hppa_lookup_stub_minimal_symbol):
Update.  Change return type.
* hppa-tdep.h (hppa_lookup_stub_minimal_symbol): Change return
type.
* jit.c (jit_breakpoint_re_set_internal): Update.
* linux-fork.c (inferior_call_waitpid, checkpoint_command):
Update.
* linux-nat.c (get_signo): Update.
* linux-thread-db.c (inferior_has_bug): Update
* m32c-tdep.c (m32c_return_value)
(m32c_m16c_address_to_pointer): Update.
* m32r-tdep.c (m32r_frame_this_id): Update.
* m68hc11-tdep.c (m68hc11_get_register_info): Update.
* machoread.c (macho_resolve_oso_sym_with_minsym): Update.
* minsyms.c (lookup_minimal_symbol_internal): Rename to
lookup_minimal_symbol.  Change return type.
(lookup_minimal_symbol): Remove.
(lookup_bound_minimal_symbol): Update.
(lookup_minimal_symbol_text): Change return type.
(lookup_minimal_symbol_solib_trampoline): Change return type.
* minsyms.h (lookup_minimal_symbol, lookup_minimal_symbol_text)
(lookup_minimal_symbol_solib_trampoline): Change return type.
* mips-linux-tdep.c (mips_linux_skip_resolver): Update.
* objc-lang.c (lookup_objc_class, lookup_child_selector)
(value_nsstring, find_imps): Update.
* obsd-tdep.c (obsd_skip_solib_resolver): Update.
* p-lang.c (pascal_main_name): Update.
* ppc-linux-tdep.c (ppc_linux_spe_context_lookup): Update.
* ppc-sysv-tdep.c (convert_code_addr_to_desc_addr): Update.
* proc-service.c (ps_pglobal_lookup): Update.
* ravenscar-thread.c (get_running_thread_msymbol): Change
return type.
(has_ravenscar_runtime, get_running_thread_id): Update.
* remote.c (remote_check_symbols): Update.
* sol-thread.c (ps_pglobal_lookup): Update.
* sol2-tdep.c (sol2_skip_solib_resolver): Update.
* solib-dsbt.c (lm_base): Update.
* solib-frv.c (lm_base, frv_relocate_section_addresses):
Update.
* solib-irix.c (locate_base): Update.
* solib-som.c (som_solib_create_inferior_hook)
(som_solib_desire_dynamic_linker_symbols, link_map_start):
Update.
* solib-spu.c (spu_enable_break): Update.
* solib-svr4.c (elf_locate_base, enable_break): Update.
* spu-tdep.c (spu_get_overlay_table, spu_catch_start)
(flush_ea_cache): Update.
* stabsread.c (define_symbol): Update.
* symfile.c (simple_read_overlay_table): Update.
* symtab.c (find_pc_sect_line): Update.
* tracepoint.c (scope_info): Update.
* tui-disasm.c (tui_get_begin_asm_address): Update.
* value.c (value_static_field): Update.

10 years agomake MSYMBOL_VALUE_ADDRESS an rvalue
Tom Tromey [Thu, 15 Aug 2013 14:44:43 +0000 (08:44 -0600)]
make MSYMBOL_VALUE_ADDRESS an rvalue

This changes MSYMBOL_VALUE_ADDRESS to be an rvalue.  In a later patch
we change this macro to compute its value; this patch introduces a
setter to make the break a bit cleaner.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* minsyms.c (prim_record_minimal_symbol_full): Use
SET_MSYMBOL_VALUE_ADDRESS.
* objfiles.c (objfile_relocate1): Use SET_MSYMBOL_VALUE_ADDRESS.
* sh64-tdep.c (sh64_elf_make_msymbol_special): Use
SET_MSYMBOL_VALUE_ADDRESS.
* symtab.h (MSYMBOL_VALUE_ADDRESS): Expand to an rvalue.
(SET_MSYMBOL_VALUE_ADDRESS): New macro.

10 years agochange minsym representation
Tom Tromey [Thu, 15 Aug 2013 14:43:43 +0000 (08:43 -0600)]
change minsym representation

In a later patch we're going to change the minimal symbol address
calculation to apply section offsets at the point of use.  To make it
simpler to catch potential problem spots, this patch changes the
representation of minimal symbols and introduces new
minimal-symbol-specific variants of the various accessors.  This is
necessary because it would be excessively ambitious to try to convert
all the symbol types at once.

The core of this change is just renaming a field in minimal_symbol;
the rest is just a fairly mechanical rewording.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* symtab.h (struct minimal_symbol) <mginfo>: Rename from ginfo.
(MSYMBOL_VALUE, MSYMBOL_VALUE_ADDRESS, MSYMBOL_VALUE_BYTES)
(MSYMBOL_BLOCK_VALUE, MSYMBOL_VALUE_CHAIN, MSYMBOL_LANGUAGE)
(MSYMBOL_SECTION, MSYMBOL_OBJ_SECTION, MSYMBOL_NATURAL_NAME)
(MSYMBOL_LINKAGE_NAME, MSYMBOL_PRINT_NAME, MSYMBOL_DEMANGLED_NAME)
(MSYMBOL_SET_LANGUAGE, MSYMBOL_SEARCH_NAME)
(MSYMBOL_MATCHES_SEARCH_NAME, MSYMBOL_SET_NAMES): New macros.
* ada-lang.c (ada_main_name): Update.
(ada_lookup_simple_minsym): Update.
(ada_make_symbol_completion_list): Update.
(ada_add_standard_exceptions): Update.
* ada-tasks.c (read_atcb, ada_tasks_inferior_data_sniffer): Update.
* aix-thread.c (pdc_symbol_addrs, pd_enable): Update.
* amd64-windows-tdep.c (amd64_skip_main_prologue): Update.
* arm-tdep.c (skip_prologue_function): Update.
(arm_skip_stack_protector, arm_skip_stub): Update.
* arm-wince-tdep.c (arm_pe_skip_trampoline_code): Update.
(arm_wince_skip_main_prologue): Update.
* auxv.c (ld_so_xfer_auxv): Update.
* avr-tdep.c (avr_scan_prologue): Update.
* ax-gdb.c (gen_var_ref): Update.
* block.c (call_site_for_pc): Update.
* blockframe.c (get_pc_function_start): Update.
(find_pc_partial_function_gnu_ifunc): Update.
* breakpoint.c (create_overlay_event_breakpoint): Update.
(create_longjmp_master_breakpoint): Update.
(create_std_terminate_master_breakpoint): Update.
(create_exception_master_breakpoint): Update.
(resolve_sal_pc): Update.
* bsd-uthread.c (bsd_uthread_lookup_address): Update.
* btrace.c (ftrace_print_function_name, ftrace_function_switched):
Update.
* c-valprint.c (c_val_print): Update.
* coff-pe-read.c (add_pe_forwarded_sym): Update.
* coffread.c (coff_symfile_read): Update.
* common/agent.c (agent_look_up_symbols): Update.
* dbxread.c (find_stab_function_addr): Update.
(end_psymtab): Update.
* dwarf2loc.c (call_site_to_target_addr): Update.
(func_verify_no_selftailcall): Update.
(tailcall_dump): Update.
(call_site_find_chain_1): Update.
(dwarf_expr_reg_to_entry_parameter): Update.
* elfread.c (elf_gnu_ifunc_record_cache): Update.
(elf_gnu_ifunc_resolve_by_got): Update.
* f-valprint.c (info_common_command): Update.
* findvar.c (read_var_value): Update.
* frame.c (get_prev_frame_1): Update.
(inside_main_func): Update.
* frv-tdep.c (frv_skip_main_prologue): Update.
(frv_frame_this_id): Update.
* glibc-tdep.c (glibc_skip_solib_resolver): Update.
* gnu-v2-abi.c (gnuv2_value_rtti_type): Update.
* gnu-v3-abi.c (gnuv3_rtti_type): Update.
(gnuv3_skip_trampoline): Update.
* hppa-hpux-tdep.c (hppa32_hpux_in_solib_call_trampoline): Update.
(hppa64_hpux_in_solib_call_trampoline): Update.
(hppa_hpux_skip_trampoline_code): Update.
(hppa64_hpux_search_dummy_call_sequence): Update.
(hppa_hpux_find_import_stub_for_addr): Update.
(hppa_hpux_find_dummy_bpaddr): Update.
* hppa-tdep.c (hppa_symbol_address)
(hppa_lookup_stub_minimal_symbol): Update.
* i386-tdep.c (i386_skip_main_prologue): Update.
(i386_pe_skip_trampoline_code): Update.
* ia64-tdep.c (ia64_convert_from_func_ptr_addr): Update.
* infcall.c (get_function_name): Update.
* infcmd.c (until_next_command): Update.
* jit.c (jit_breakpoint_re_set_internal): Update.
(jit_inferior_init): Update.
* linespec.c (minsym_found): Update.
(add_minsym): Update.
* linux-fork.c (info_checkpoints_command): Update.
* linux-nat.c (get_signo): Update.
* linux-thread-db.c (inferior_has_bug): Update.
* m32c-tdep.c (m32c_return_value): Update.
(m32c_m16c_address_to_pointer): Update.
(m32c_m16c_pointer_to_address): Update.
* m32r-tdep.c (m32r_frame_this_id): Update.
* m68hc11-tdep.c (m68hc11_get_register_info): Update.
* machoread.c (macho_resolve_oso_sym_with_minsym): Update.
* maint.c (maintenance_translate_address): Update.
* minsyms.c (add_minsym_to_hash_table): Update.
(add_minsym_to_demangled_hash_table): Update.
(msymbol_objfile): Update.
(lookup_minimal_symbol): Update.
(iterate_over_minimal_symbols): Update.
(lookup_minimal_symbol_text): Update.
(lookup_minimal_symbol_by_pc_name): Update.
(lookup_minimal_symbol_solib_trampoline): Update.
(lookup_minimal_symbol_by_pc_section_1): Update.
(lookup_minimal_symbol_and_objfile): Update.
(prim_record_minimal_symbol_full): Update.
(compare_minimal_symbols): Update.
(compact_minimal_symbols): Update.
(build_minimal_symbol_hash_tables): Update.
(install_minimal_symbols): Update.
(terminate_minimal_symbol_table): Update.
(find_solib_trampoline_target): Update.
(minimal_symbol_upper_bound): Update.
* mips-linux-tdep.c (mips_linux_skip_resolver): Update.
* mips-tdep.c (mips_stub_frame_sniffer): Update.
(mips_skip_pic_trampoline_code): Update.
* msp430-tdep.c (msp430_skip_trampoline_code): Update.
* objc-lang.c (selectors_info): Update.
(classes_info): Update.
(find_methods): Update.
(find_imps): Update.
(find_objc_msgsend): Update.
* objfiles.c (objfile_relocate1): Update.
* objfiles.h (ALL_OBJFILE_MSYMBOLS): Update.
* obsd-tdep.c (obsd_skip_solib_resolver): Update.
* p-valprint.c (pascal_val_print): Update.
* parse.c (write_exp_msymbol): Update.
* ppc-linux-tdep.c (powerpc_linux_in_dynsym_resolve_code)
(ppc_linux_spe_context_lookup, ppc_elfv2_skip_entrypoint): Update.
* ppc-sysv-tdep.c (convert_code_addr_to_desc_addr): Update.
* printcmd.c (build_address_symbolic): Update.
(sym_info): Update.
(address_info): Update.
* proc-service.c (ps_pglobal_lookup): Update.
* psymtab.c (find_pc_sect_psymtab_closer): Update.
(find_pc_sect_psymtab): Update.
* python/py-framefilter.c (py_print_frame): Update.
* ravenscar-thread.c (get_running_thread_id): Update.
* record-btrace.c (btrace_call_history, btrace_get_bfun_name):
Update.
* remote.c (remote_check_symbols): Update.
* rs6000-tdep.c (rs6000_skip_main_prologue): Update.
(rs6000_skip_trampoline_code): Update.
* sh64-tdep.c (sh64_elf_make_msymbol_special): Update.
* sol2-tdep.c (sol2_skip_solib_resolver): Update.
* solib-dsbt.c (lm_base): Update.
* solib-frv.c (lm_base): Update.
(main_got): Update.
* solib-irix.c (locate_base): Update.
* solib-som.c (som_solib_create_inferior_hook): Update.
(som_solib_desire_dynamic_linker_symbols): Update.
(link_map_start): Update.
* solib-spu.c (spu_enable_break): Update.
(ocl_enable_break): Update.
* solib-svr4.c (elf_locate_base): Update.
(enable_break): Update.
* spu-tdep.c (spu_get_overlay_table): Update.
(spu_catch_start): Update.
(flush_ea_cache): Update.
* stabsread.c (define_symbol): Update.
(scan_file_globals): Update.
* stack.c (find_frame_funname): Update.
(frame_info): Update.
* symfile.c (simple_read_overlay_table): Update.
(simple_overlay_update): Update.
* symmisc.c (dump_msymbols): Update.
* symtab.c (fixup_section): Update.
(find_pc_sect_line): Update.
(skip_prologue_sal): Update.
(search_symbols): Update.
(print_msymbol_info): Update.
(rbreak_command): Update.
(MCOMPLETION_LIST_ADD_SYMBOL): New macro.
(completion_list_objc_symbol): Update.
(default_make_symbol_completion_list_break_on): Update.
* tracepoint.c (scope_info): Update.
* tui/tui-disasm.c (tui_find_disassembly_address): Update.
(tui_get_begin_asm_address): Update.
* valops.c (find_function_in_inferior): Update.
* value.c (value_static_field): Update.
(value_fn_field): Update.

10 years agointroduce minimal_symbol_upper_bound
Tom Tromey [Mon, 19 Aug 2013 13:58:44 +0000 (07:58 -0600)]
introduce minimal_symbol_upper_bound

This introduces minimal_symbol_upper_bound and changes various bits of
code to use it.  Since this function is intimately tied to the
implementation of minimal symbol tables, I believe it belongs in
minsyms.c.

The new function is extracted from find_pc_partial_function_gnu_ifunc.
This isn't a "clean" move because the old function interleaved the
caching and the computation; but this doesn't make sense for the new
code.

2014-02-26  Tom Tromey  <tromey@redhat.com>

* blockframe.c (find_pc_partial_function_gnu_ifunc): Use
bound minimal symbols.  Move code that knows about minsym
table layout...
* minsyms.c (minimal_symbol_upper_bound): ... here.  New
function.
* minsyms.h (minimal_symbol_upper_bound): Declare.
* objc-lang.c (find_objc_msgsend): Use bound minimal symbols,
minimal_symbol_upper_bound.

10 years ago[Python] Make regexp collection printers work with typedefs as well.
Joel Brobecker [Thu, 30 Jan 2014 03:37:08 +0000 (07:37 +0400)]
[Python] Make regexp collection printers work with typedefs as well.

Consider the following type for which we would like to provide
a pretty-printer and manage it via RegexpCollectionPrettyPrinter:

        typedef long time_t;

Currently, this does not work because this framework only considers
the type's tag name:

        typename = gdb.types.get_basic_type(val.type).tag
        if not typename:
            return None

This patch extends it to use the type's name if the basic type
does not have a tag name, thus allowing the framework to also
work with typedefs like the above.

gdb/ChangeLog:

        * python/lib/gdb/printing.py (RegexpCollectionPrettyPrinter):
        Use the type's name if its basic type does not have a tag.

gdb/testsuite/ChangeLog:

        * testsuite/gdb.python/py-pp-re-notag.c: New file.
        * testsuite/gdb.python/py-pp-re-notag.ex: New file.
        * testsuite/gdb.python/py-pp-re-notag.p: New file.

10 years agoAdd comment in dwarf2read.c::read_subrange_type
Joel Brobecker [Wed, 26 Feb 2014 18:53:05 +0000 (10:53 -0800)]
Add comment in dwarf2read.c::read_subrange_type

This comment explains why we sometimes sign-extend the range type
bounds when we normally shouldn't have to.

gdb/ChangeLog:

        * dwarf2read.c (read_subrange_type): Add comment.

10 years agoDWARF: Set enum type "flag_enum" and "unsigned" flags at type creation.
Joel Brobecker [Wed, 22 Jan 2014 14:40:20 +0000 (18:40 +0400)]
DWARF: Set enum type "flag_enum" and "unsigned" flags at type creation.

Consider the following Ada code:

   --  An array whose index is an enumeration type with 128 enumerators.
   type Enum_T is (Enum_000, Enum_001, [...], Enum_128);
   type Table is array (Enum_T) of Boolean;

When the compiler is configured to generate pure DWARF debugging info,
trying to print type Table's description yields:

    ptype pck.table
    type = array (enum_000 .. -128) of boolean

The expected output was:

    ptype pck.table
    type = array (enum_000 .. enum_128) of boolean

The DWARF debugging info for our array looks like this:

    <1><44>: Abbrev Number: 5 (DW_TAG_array_type)
       <45>   DW_AT_name        : pck__table
       <50>   DW_AT_type        : <0x28>
    <2><54>: Abbrev Number: 6 (DW_TAG_subrange_type)
       <55>   DW_AT_type        : <0x5c>
       <59>   DW_AT_lower_bound : 0
       <5a>   DW_AT_upper_bound : 128

The array index type is, by construction with the DWARF standard,
a subrange of our enumeration type, defined as follow:

    <2><5b>: Abbrev Number: 0
    <1><5c>: Abbrev Number: 7 (DW_TAG_enumeration_type)
       <5d>   DW_AT_name        : pck__enum_t
       <69>   DW_AT_byte_size   : 1
    <2><6b>: Abbrev Number: 8 (DW_TAG_enumerator)
       <6c>   DW_AT_name        : pck__enum_000
       <7a>   DW_AT_const_value : 0
    [etc]

Therefore, while processing these DIEs, the array index type ends
up being a TYPE_CODE_RANGE whose target type is our enumeration type.
But the problem is that we read the upper bound as a negative value
(-128), which is then used as is by the type printer to print the
array upper bound. This negative value explains the "-128" in the
output.

To understand why the range type's upper bound is read as a negative
value, one needs to look at how it is determined, in read_subrange_type:

  orig_base_type = die_type (die, cu);
  base_type = check_typedef (orig_base_type);
  [... high is first correctly read as 128, but then ...]
  if (!TYPE_UNSIGNED (base_type) && (high & negative_mask))
    high |= negative_mask;

The negative_mask is applied, here, because BASE_TYPE->FLAG_UNSIGNED
is not set. And the reason for that is because the base_type was only
partially constructed during the call to die_type. While the enum
is constructed on the fly by read_enumeration_type, its flag_unsigned
flag is only set later on, while creating the symbols corresponding to
the enum type's enumerators (see process_enumeration_scope), after
we've already finished creating our range type - and therefore too
late.

My first naive attempt at fixing this problem consisted in extracting
the part in process_enumeration_scope which processes all enumerators,
to generate the associated symbols, but more importantly set the type's
various flags when necessary. However, this does not always work well,
because we're still in the subrange_type's scope, and it might be
different from the scope where the enumeration type is defined.

So, instead, what this patch does to fix the issue is to extract
from process_enumeration_scope the part that determines whether
the enumeration type should have the flag_unsigned and/or the
flag_flag_enum flags set. It turns out that, aside from the code
implementing the loop, this part is fairly independent of the symbol
creation. With that part extracted, we can then use it at the end
of our enumeration type creation, to produce a type which should now
no longer need any adjustment.

Once the enumeration type produced is correctly marked as unsigned,
the subrange type's upper bound is then correctly read as an unsigned
value, therefore giving us an upper bound of 128 instead of -128.

gdb/ChangeLog:

        * dwarf2read.c (update_enumeration_type_from_children): New
        function, mostly extracted from process_structure_scope.
        (read_enumeration_type): Call update_enumeration_type_from_children.
        (process_enumeration_scope): Do not set THIS_TYPE's flag_unsigned
        and flag_flag_enum fields.

gdb/testsuite/ChangeLog:

        * gdb.dwarf2/arr-subrange.c, gdb.dwarf2/arr-subrange.exp: New files.

10 years agoMention PR breakpoints/16292 in corresponding ChangeLog entry.
Pedro Alves [Wed, 26 Feb 2014 16:33:13 +0000 (16:33 +0000)]
Mention PR breakpoints/16292 in corresponding ChangeLog entry.

10 years agobsd-uthread.c: Don't install a to_xfer_partial method
Pedro Alves [Wed, 26 Feb 2014 14:36:05 +0000 (14:36 +0000)]
bsd-uthread.c: Don't install a to_xfer_partial method

Whatever the comment about deprecated_xfer_memory referred to,
deprecated_xfer_memory is gone now.  There's no need to install a
target method that just delegates, as that's what the default
delegator does already.

Tested by building an --enable-targets=all gdb on x86_64 Fedora 17.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* bsd-uthread.c (bsd_uthread_xfer_partial): Delete function.
(bsd_uthread_target): Don't install bsd_uthread_xfer_partial as
to_xfer_partial method.

10 years agoeliminate target_ops->deprecated_xfer_memory
Pedro Alves [Wed, 26 Feb 2014 14:36:04 +0000 (14:36 +0000)]
eliminate target_ops->deprecated_xfer_memory

As no target uses it anymore, it can finally go away.

After removing the deprecated_xfer_memory handling from
default_xfer_partial, we can delete the latter, because the only thing
it does is delegate to the target beneath unconditionally, which is
what the delegator installed by target-delegates.c will do for us if
no to_xfer_partial method is installed.

This was the last user of de_fault, so that goes away too.

Tested on x86_64 Fedora 17.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* target.c (complete_target_initialization): Don't install
default_xfer_partial as to_xfer_partial hook.
(nomemory): Delete.
(update_current_target): Don't INHERIT nor de_fault
deprecated_xfer_memory.  Delete de_fault macro.
(default_xfer_partial, deprecated_debug_xfer_memory): Delete.
(setup_target_debug): Don't install a deprecated_xfer_memory hook.
* target.h (struct target_ops) <deprecated_xfer_memory>: Delete
field.

10 years agogo32-nat.c: Don't install a deprecated_xfer_memory method
Pedro Alves [Wed, 26 Feb 2014 14:36:04 +0000 (14:36 +0000)]
go32-nat.c: Don't install a deprecated_xfer_memory method

This removes yet another instance of a deprecated_xfer_memory user.

Unfortunately djgpp's write_child function takes a non-const buffer
pointer, while GDB's xfer_partial api passes a const pointer.  To be
const-correct, we need to copy that buffer to a non-const buffer, and
pass the copy to write_child.  This is actually what
target.c:default_xfer_partial itself does, when calling into the
ops->deprecated_xfer_memory hook.

Tested by cross-building djgpp gdb, on x86-64 Fedora 17.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* go32-nat.c (my_write_child): New function.
(go32_xfer_memory): Rewrite as to_xfer_partial helper.
(go32_xfer_partial): New function.
(init_go32_ops): Don't install a deprecated_xfer_memory hook.
Instead install a to_xfer_partial hook.

10 years agonto-procfs.c: Don't install a deprecated_xfer_memory method
Pedro Alves [Wed, 26 Feb 2014 14:36:04 +0000 (14:36 +0000)]
nto-procfs.c: Don't install a deprecated_xfer_memory method

This removes yet another instance of a deprecated_xfer_memory user.

Completely untested.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* nto-procfs.c (procfs_xfer_memory): Adjust interface as a
to_xfer_partial helper.  Rewrite.
(procfs_xfer_partial): New function.
(init_procfs_ops): Don't install a deprecated_xfer_memory hook.
Install a to_xfer_partial hook.

10 years agoremote-m32r-sdi.c: Don't install a deprecated_xfer_memory method
Pedro Alves [Wed, 26 Feb 2014 14:36:03 +0000 (14:36 +0000)]
remote-m32r-sdi.c: Don't install a deprecated_xfer_memory method

This removes yet another instance of a deprecated_xfer_memory user.

Tested by building a --enable-targets=all gdb, on x86-64 Fedora 17.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* remote-m32r-sdi.c (send_data): Constify 'buf' parameter.
(m32r_xfer_memory): Adjust as a to_xfer_partial helper.
(m32r_xfer_partial): New function.
(init_m32r_ops): Don't install a deprecated_xfer_memory hook.
Install a to_xfer_partial hook.

10 years agoremote-mips.c: Don't install a deprecated_xfer_memory method
Pedro Alves [Wed, 26 Feb 2014 14:36:03 +0000 (14:36 +0000)]
remote-mips.c: Don't install a deprecated_xfer_memory method

This removes another yet instance of a deprecated_xfer_memory user.

Tested by building a --enable-targets=all gdb, on x86-64 Fedora 17.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* remote-mips.c (mips_xfer_memory): Adjust as to_xfer_partial
helper.
(mips_xfer_partial): New function.
(_initialize_remote_mips): Don't install a deprecated_xfer_memory
hook.  Install a to_xfer_partial hook.

10 years agoDWARF: Add array DW_AT_bit_stride and DW_AT_byte_stride support
Joel Brobecker [Wed, 29 Jan 2014 13:39:56 +0000 (17:39 +0400)]
DWARF: Add array DW_AT_bit_stride and DW_AT_byte_stride support

Consider the following declarations in Ada...

   type Item is range -32 .. 31;
   for Item'Size use 6;

   type Table is array (Natural range 0 .. 4) of Item;
   pragma Pack (Table);

... which declare a packed array whose elements are 6 bits long.
The debugger currently does not notice that the array is packed,
and thus prints values of this type incorrectly. This can be seen
in the "ptype" output:

    (gdb) ptype table
    type = array (0 .. 4) of foo.item

Normally, the debugger should print:

    (gdb) ptype table
    type = array (0 .. 4) of foo.item <packed: 6-bit elements>

The debugging information for this array looks like this:

        .uleb128 0xf    # (DIE (0x15c) DW_TAG_array_type)
        .long   .LASF9  # DW_AT_name: "pck__table"
        .byte   0x6     # DW_AT_bit_stride
        .long   0x1a9   # DW_AT_type
        .uleb128 0x10   # (DIE (0x16a) DW_TAG_subrange_type)
        .long   0x3b    # DW_AT_type
        .byte   0       # DW_AT_lower_bound
        .byte   0x4     # DW_AT_upper_bound
        .byte   0       # end of children of DIE 0x15c

The interesting part is the DW_AT_bit_stride attribute, which tells
the size of the array elements is 6 bits, rather than the normal
element type's size.

This patch adds support for this attribute by first creating
gdbtypes.c::create_array_type_with_stride, which is an enhanced
version of create_array_type taking an extra parameter as the stride.
The old create_array_type can then be re-implemented very simply
by calling the new create_array_type_with_stride.

We can then use this new function from dwarf2read, to create
arrays with or without stride.

gdb/ChangeLog:

        * gdbtypes.h (create_array_type_with_stride): Add declaration.
        * gdbtypes.c (create_array_type_with_stride): New function,
        renaming create_array_type, but with an added parameter
        called "bit_stride".
        (create_array_type): Re-implement using
        create_array_type_with_stride.
        * dwarf2read.c (read_array_type): Add support for DW_AT_byte_stride
        and DW_AT_bit_stride attributes.

gdb/testsuite/ChangeLog:

        * gdb.dwarf2/arr-stride.c: New file.
        * gdb.dwarf2/arr-stride.exp: New file.

The test, relying purely on generating an assembly file, only
verifies the type description of our array. But I was also
able to verify manually that the debugger print values of these
types correctly as well (which was not the case prior to this
patch).

10 years agoMultiple Ada task-specific breakpoints at the same address.
Pedro Alves [Wed, 26 Feb 2014 14:22:33 +0000 (14:22 +0000)]
Multiple Ada task-specific breakpoints at the same address.

With the test changed as in the patch, against current mainline, we get:

 (gdb) PASS: gdb.ada/tasks.exp: info tasks before inserting breakpoint
 break break_me task 1
 Breakpoint 2 at 0x4030b0: file /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.ada/tasks/foo.adb, line 27.
 (gdb) PASS: gdb.ada/tasks.exp: break break_me task 1
 break break_me task 3
 Note: breakpoint 2 also set at pc 0x4030b0.
 Breakpoint 3 at 0x4030b0: file /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.ada/tasks/foo.adb, line 27.
 (gdb) PASS: gdb.ada/tasks.exp: break break_me task 3
 continue
 Continuing.
 [Switching to Thread 0x7ffff7dc7700 (LWP 27133)]

 Breakpoint 2, foo.break_me () at /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.ada/tasks/foo.adb:27
 27       null;
 (gdb) FAIL: gdb.ada/tasks.exp: continue to breakpoint
 info tasks
    ID       TID P-ID Pri State                  Name
     1    63b010       48 Waiting on RV with 3   main_task
     2    63bd80    1  48 Accept or Select Term  task_list(1)
 *   3    63f510    1  48 Accepting RV with 1    task_list(2)
     4    642ca0    1  48 Accept or Select Term  task_list(3)
 (gdb) PASS: gdb.ada/tasks.exp: info tasks after hitting breakpoint

The breakpoint that caused a stop is breakpoint 3, but GDB end up
reporting (and running breakpoint commands of) "Breakpoint 2" instead.

The issue is that the bpstat_check_breakpoint_conditions logic of
"wrong thread" is missing the "wrong task" check.  This is usually
harmless, because the thread hop code in infrun.c code that handles
wrong-task-hitting-breakpoint does check for task-specific breakpoints
(within breakpoint_thread_match):

      /* Check if a regular breakpoint has been hit before checking
         for a potential single step breakpoint.  Otherwise, GDB will
         not see this breakpoint hit when stepping onto breakpoints.  */
      if (regular_breakpoint_inserted_here_p (aspace, stop_pc))
{
  if (!breakpoint_thread_match (aspace, stop_pc, ecs->ptid))
    thread_hop_needed = 1;
}

IOW, usually, when one only has a task specific breakpoint at a given
address, things work correctly.  Put another task-specific or
non-task-specific breakpoint there, and things break.

A patch that eliminates the special thread hop code in infrun.c is
what exposed this, as after that GDB solely relies on
bpstat_check_breakpoint_conditions to know whether the right or wrong
task hit a breakpoint.  IOW, given the latent bug, Ada task-specific
breakpoints become non-task-specific, and that is caught by the
testsuite, as:

 break break_me task 3
 Breakpoint 2 at 0x4030b0: file /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.ada/tasks/foo.adb, line 27.
 (gdb) PASS: gdb.ada/tasks.exp: break break_me task 3
 continue
 Continuing.
 [Switching to Thread 0x7ffff7fcb700 (LWP 17122)]

 Breakpoint 2, foo.break_me () at /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.ada/tasks/foo.adb:27
 27       null;
 (gdb) PASS: gdb.ada/tasks.exp: continue to breakpoint
 info tasks
    ID       TID P-ID Pri State                  Name
     1    63b010       48 Waiting on RV with 2   main_task
 *   2    63bd80    1  48 Accepting RV with 1    task_list(1)
     3    63f510    1  48 Accept or Select Term  task_list(2)
     4    642ca0    1  48 Accept or Select Term  task_list(3)
 (gdb) FAIL: gdb.ada/tasks.exp: info tasks after hitting breakpoint

It was after seeing this that I thought of how to expose the bug with
current mainline.

Tested on x86_64 Fedora 17.

gdb/
2014-02-26  Pedro Alves  <palves@redhat.com>

* breakpoint.c (bpstat_check_breakpoint_conditions): Handle
task-specific breakpoints.

gdb/testsuite/
2014-02-26  Pedro Alves  <palves@redhat.com>

* gdb.ada/tasks.exp: Set a task-specific breakpoint at break_me
that won't ever trigger.  Make sure that GDB reports the correct
breakpoint that caused the stop.

10 years agoResolve PR ld/16569 by emitting (and comparing) unmangled names, unless
Dan Mick [Wed, 26 Feb 2014 12:05:02 +0000 (12:05 +0000)]
Resolve PR ld/16569 by emitting (and comparing) unmangled names, unless
demangling has specifically been requested.

10 years agoRe-implement ia64-linux-nat.c::ia64_linux_xfer_partial
Pedro Alves [Wed, 26 Feb 2014 01:45:50 +0000 (20:45 -0500)]
Re-implement ia64-linux-nat.c::ia64_linux_xfer_partial

[description of this patch and ChangeLog entry by Joel Brobecker]
The recent implementation was questionable, and if it worked, it was
only by chance because the requested length is large enough that only
one read was sufficient.  Note that the implementation before that
also made that assumption, in the form of only handling
TARGET_OBJECT_UNWIND_TABLE xfer requests when offset was zero.

gdb/ChangeLog:

        * ia64-linux-nat.c (ia64_linux_xfer_partial): Reimplement
        handling of object == TARGET_OBJECT_UNWIND_TABLE.

10 years agoAnnotate comments for Doxygen.
Stan Shebs [Tue, 25 Feb 2014 23:47:58 +0000 (15:47 -0800)]
Annotate comments for Doxygen.

10 years agodaily update
Alan Modra [Tue, 25 Feb 2014 23:00:38 +0000 (09:30 +1030)]
daily update

10 years agoPR gdb/16626
Jan Kratochvil [Tue, 25 Feb 2014 19:47:09 +0000 (20:47 +0100)]
PR gdb/16626

gdb/testsuite/
2014-02-25  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR gdb/16626
* gdb.base/auto-load.exp: Fix out-of-srctree run.

Message-ID: <87k3cjt3jl.fsf@fleche.redhat.com>

10 years agoremove target_ignore
Tom Tromey [Mon, 24 Feb 2014 21:57:14 +0000 (14:57 -0700)]
remove target_ignore

This removes target_ignore, which isn't used any more.

2014-02-25  Tom Tromey  <tromey@redhat.com>

* target.h (target_ignore): Don't declare.
* target.c (target_ignore): Remove.

10 years agoPR gdb/16626
Jan Kratochvil [Tue, 25 Feb 2014 17:32:32 +0000 (18:32 +0100)]
PR gdb/16626

Fix auto-load 7.7 regression,
the regression affects any loading from /usr/share/gdb/auto-load .

5b2bf9471f1499bee578fcd60c05afe85794e280 is the first bad commit
commit 5b2bf9471f1499bee578fcd60c05afe85794e280
Author: Doug Evans <xdje42@gmail.com>
Date:   Fri Nov 29 21:29:26 2013 -0800
    Move .debug_gdb_script processing to auto-load.c.
    Simplify handling of auto-loaded objfile scripts.

Fedora 20 x86_64
$ gdb -q /usr/lib64/libgobject-2.0.so
Reading symbols from /usr/lib64/libglib-2.0.so.0.3800.2...Reading symbols from
/usr/lib/debug/usr/lib64/libglib-2.0.so.0.3800.2.debug...done.
done.
(gdb) _

Fedora Rawhide x86_64
$ gdb -q /usr/lib64/libgobject-2.0.so
Reading symbols from /usr/lib64/libglib-2.0.so...Reading symbols from
/usr/lib/debug/usr/lib64/libglib-2.0.so.0.3990.0.debug...done.
done.
warning: File "/usr/lib64/libglib-2.0.so.0.3990.0-gdb.py" auto-loading has been declined by your `auto-load safe-path'
set to "$debugdir:$datadir/auto-load:/usr/bin/mono-gdb.py".
To enable execution of this file add
        add-auto-load-safe-path /usr/lib64/libglib-2.0.so.0.3990.0-gdb.py
line to your configuration file "/home/jkratoch/.gdbinit".
To completely disable this security protection add
        set auto-load safe-path /
line to your configuration file "/home/jkratoch/.gdbinit".
For more information about this security protection see the
"Auto-loading safe path" section in the GDB manual.  E.g., run from the shell:
        info "(gdb)Auto-loading safe path"
(gdb) _

That is it tries to load "forbidden"
/usr/lib64/libglib-2.0.so.0.3990.0-gdb.py
but it should load instead
/usr/share/gdb/auto-load/usr/lib64/libglib-2.0.so.0.3990.0-gdb.py*
Although that is also not exactly this way, there does not exist any
/usr/lib64/libglib-2.0.so.0.3990.0-gdb.py
despite regressed GDB says so.

gdb/
2014-02-24  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR gdb/16626
* auto-load.c (auto_load_objfile_script_1): Change filename to
debugfile.

gdb/testsuite/
2014-02-24  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR gdb/16626
* gdb.base/auto-load-script: New file.
* gdb.base/auto-load.c: New file.
* gdb.base/auto-load.exp: New file.

Message-ID: <20140223212400.GA8831@host2.jankratochvil.net>

10 years agoFix dw2-icycle.exp -fsanitize=address GDB crash.
Jan Kratochvil [Tue, 25 Feb 2014 17:28:38 +0000 (18:28 +0100)]
Fix dw2-icycle.exp -fsanitize=address GDB crash.

binutils readelf -wi:
 <4><a2>: Abbrev Number: 26 (DW_TAG_inlined_subroutine)
    <a3>   DW_AT_abstract_origin: <0x5a>
    <a7>   DW_AT_low_pc      : 0x400590
    <ab>   DW_AT_high_pc     : 0x4
    <af>   DW_AT_call_file   : 1
    <b0>   DW_AT_call_line   : 20
    <b1>   DW_AT_sibling     : <0xb8>
 <2><b8>: Abbrev Number: 35 (DW_TAG_inlined_subroutine)
    <b9>   DW_AT_abstract_origin: <0x5a>
    <bd>   DW_AT_low_pc      : 0x400590
    <c1>   DW_AT_high_pc     : 0x4
    <c5>   DW_AT_call_file   : 1
    <c6>   DW_AT_call_line   : 29

<b1> DW_AT_sibling points to the next DIE - but that DIE is 2 levels
upwards - definitely not a sibling.  This confuses GDB up to a crash:

==32143== ERROR: AddressSanitizer: heap-buffer-overflow on address 0x6024000198ac at pc 0xb4d104 bp 0x7fff63e96e70 sp
0x7fff63e96e60
READ of size 1 at 0x6024000198ac thread T0
    #0 0xb4d103 in read_unsigned_leb128 (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb4d103)
    #1 0xb15f3c in peek_die_abbrev (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb15f3c)
    #2 0xb46185 in load_partial_dies (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb46185)
    #3 0xb103fb in process_psymtab_comp_unit_reader (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb103fb)
    #4 0xb0d2a9 in init_cutu_and_read_dies (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb0d2a9)
    #5 0xb1115f in process_psymtab_comp_unit (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb1115f)
    #6 0xb1235f in dwarf2_build_psymtabs_hard (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb1235f)
    #7 0xb05536 in dwarf2_build_psymtabs (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0xb05536)
    #8 0x86d5a5 in read_psyms (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x86d5a5)
    #9 0x9b1c37 in require_partial_symbols (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x9b1c37)
    #10 0x9bf2d0 in read_symbols (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x9bf2d0)
    #11 0x9c014c in syms_from_objfile_1 (/home/jkratoch/redhat/gdb-clean/gdb/gdb+0x9c014c)

gdb/testsuite/
2014-02-25  Jan Kratochvil  <jan.kratochvil@redhat.com>

Fix dw2-icycle.exp -fsanitize=address GDB crash.
* gdb.dwarf2/dw2-icycle.S: Remove all DW_AT_sibling.

Message-ID: <20140224201011.GA28926@host2.jankratochvil.net>

10 years agoRemove bogus vcvtps2ph variant.
Ilya Tocar [Thu, 20 Feb 2014 15:08:13 +0000 (19:08 +0400)]
Remove bogus vcvtps2ph variant.

We currently support version of vcvtps2ph with sae and only 1 register operand.
This version is encoded as if missing operand was equal to ymm0.
I didn't found any references to this variant in
http://download-software.intel.com/sites/default/files/managed/50/1a/319433-018.pdf
This patch removes it.

opcodes/

        * i386-opc.tbl: Remove wrong variant of vcvtps2ph
        * i386-tbl.h: Regenerate.

10 years agoAdjust ia64_linux_xfer_partial following to_xfer_partial API change.
Joel Brobecker [Tue, 25 Feb 2014 14:58:25 +0000 (09:58 -0500)]
Adjust ia64_linux_xfer_partial following to_xfer_partial API change.

ia64-linux-nat.c no longer compiles because ia64_linux_xfer_partial
no longer matches the to_xfer_partial prototype.  This patch fixes
the problem by adjusting it accordingly.

gdb/ChangeLog:

        * ia64-linux-nat.c (ia64_linux_xfer_partial): Add function
        documentation.  Adjust prototype to match the target_ops
        to_xfer_partial method.  Adjust implementation accordingly.

10 years agoFix a format issue of ChangeLog.
Hui Zhu [Tue, 25 Feb 2014 15:59:47 +0000 (23:59 +0800)]
Fix a format issue of ChangeLog.

10 years ago2014-02-25 Hui Zhu <hui@codesourcery.com>
Hui Zhu [Tue, 25 Feb 2014 15:55:42 +0000 (23:55 +0800)]
2014-02-25  Hui Zhu  <hui@codesourcery.com>

* target.h (target_ops): Fix TARGET_DEFAULT_RETURN of
to_traceframe_info.

10 years agoUse 16-bit integer type for rl78 register pairs.
Kevin Buettner [Tue, 25 Feb 2014 07:32:45 +0000 (00:32 -0700)]
Use 16-bit integer type for rl78 register pairs.

This patch changes rl78-tdep.c so that a 16-bit type is used for
register pairs instead of a pointer type as was previously the case.
This will cause these register pairs to be displayed as integers
instead of as a data address with a 0xf0000 ORed in.

E.g. registers ax, bc, de, and hl might now be displayed like this:

    (gdb) info registers ax bc de hl
    ax             0x6      6
    bc             0x0      0
    de             0x10c3   4291
    hl             0x108d   423

Whereas, before, they were displayed as follows:

    (gdb) info registers ax bc de hl
    ax             0xf0006  0xf0006
    bc             0xf0000  0xf0000
    de             0xf10c3  0xf10c3
    hl             0xf108d  0xf108d

These pairs are 16 bit quantities and should be displayed as such.
This change also affects the way that the banked register pairs are
displayed.  Within GDB, the banked register pairs are named bank0_rp0,
bank0_rp1, .., bank3_rp2, bank3_rp3.

However, these register pairs need to be used as addresses in DWARF
expressions.  I have added 16 pseudo registers corresponding to banked
register pairs.  These new pseudo registers are all hidden from the
user and have a pointer type.  Values from these registers are
intended to be used in DWARF expressions.  Therefore,
rl78_dwarf_reg_to_regnum() has been adjusted to return these new
pseudo register numbers.

I had a much simpler patch which only changed the types, but it showed
a number of regressions due to integer values from the banked register
pairs being used as part of an address expression.  This current patch
shows no regressions and now displays values of register pairs
correctly.

* rl78-tdep.c ( RL78_BANK0_RP0_PTR_REGNUM, RL78_BANK0_RP1_PTR_REGNUM)
(RL78_BANK0_RP2_PTR_REGNUM, RL78_BANK0_RP3_PTR_REGNUM)
(RL78_BANK1_RP0_PTR_REGNUM, RL78_BANK1_RP1_PTR_REGNUM)
(RL78_BANK1_RP2_PTR_REGNUM, RL78_BANK1_RP3_PTR_REGNUM)
(RL78_BANK2_RP0_PTR_REGNUM, RL78_BANK2_RP1_PTR_REGNUM)
(RL78_BANK2_RP2_PTR_REGNUM, RL78_BANK2_RP3_PTR_REGNUM)
(RL78_BANK3_RP0_PTR_REGNUM, RL78_BANK3_RP1_PTR_REGNUM)
(RL78_BANK3_RP2_PTR_REGNUM, RL78_BANK3_RP3_PTR_REGNUM):
New constants.
(rl78_register_type): Use a data pointer type for SP and
new pseudo registers mentioned above.  Use a 16 bit integer
type for all other register pairs.
(rl78_register_name, rl78_g10_register_name): Update for
new pseudo registers.
(rl78_pseudo_register_read): Likewise.
(rl78_pseudo_register_write): Likewise.
(rl78_dwarf_reg_to_regnum): Return register numbers representing
to the newly added pseudo registers.

10 years agodaily update
Alan Modra [Mon, 24 Feb 2014 23:00:39 +0000 (09:30 +1030)]
daily update

10 years ago * value.c (record_latest_value): Fix comment.
Doug Evans [Mon, 24 Feb 2014 22:01:45 +0000 (14:01 -0800)]
* value.c (record_latest_value): Fix comment.
* printcmd.c (print_command_1): Remove code to handle -1 return from
record_latest_value.

10 years agolib/gdb.exp (run_on_host): Log error output if program fails.
Doug Evans [Mon, 24 Feb 2014 21:39:14 +0000 (13:39 -0800)]
lib/gdb.exp (run_on_host): Log error output if program fails.

10 years agoprocfs.c: Don't install a deprecated_xfer_memory method
Pedro Alves [Wed, 19 Feb 2014 20:29:31 +0000 (20:29 +0000)]
procfs.c: Don't install a deprecated_xfer_memory method

This removes yet another instance of a deprecated_xfer_memory user,
and fixes a nasty regression as a side-effect:

    (gdb) start
    Temporary breakpoint 1 at 0x19070: file simple_main.adb, line 4.
    Starting program: /[...]/simple_main
    Warning:
    Cannot insert breakpoint 1.
    Cannot access memory at address 0x19070
    Cannot insert breakpoint -3.
    Temporarily disabling shared library breakpoints:
    breakpoint #-3

The regression was introduced by the to_xfer_partial transition
to return a status enum.  procfs_xfer_partial was updated but
not the case where object is TARGET_OBJECT_MEMORY.  As result,
procfs_xfer_partial was returning the length xfered rather than
the status, and the xfered buffer was left uninitialized.

gdb/
2014-02-19  Pedro Alves  <palves@redhat.com>

* procfs.c (procfs_target): Don't install procfs_xfer_memory as
deprecated_xfer_memory hook.
(procfs_xfer_partial): Call procfs_xfer_memory instead
of the deprecated_xfer_memory target hook.
(procfs_xfer_memory): Adjust interface as a to_xfer_partial
helper.

10 years agoFix a GDB assert failure on windows
Yuanhui Zhang [Mon, 24 Feb 2014 07:22:10 +0000 (15:22 +0800)]
Fix a GDB assert failure on windows

A GDB internal error is found on native mingw32 target.

(gdb) run
../../binutils-gdb/gdb/target.c:1483: internal-error:
target_xfer_partial: Assertion `*xfered_len > 0' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)

This error was introduced by the following snippet in commit
9b409511d07fe375284701af34909fb539029caf

> @@ -2536,27 +2538,30 @@ windows_xfer_shared_libraries (struct target_ops *ops,
>      }
>
>    obstack_free (&obstack, NULL);
> -  return len;
> +  *xfered_len = (ULONGEST) len;
> +  return TARGET_XFER_OK;
>  }

In the original code, len is returned, which could be 0, but after that
commit, only TARGET_XFER_OK is returned, which is wrong.  If len is 0,
TARGET_XFER_EOF should be returned.  (it is 0 in enum
target_xfer_status declaration).

gdb:

2014-02-24  Yuanhui Zhang  <asmwarrior@gmail.com>

* windows-nat.c (windows_xfer_shared_libraries): Return
TARGET_XFER_EOF if LEN is zero to fix an assert failure when
requested object is TARGET_OBJECT_LIBRARIES.

10 years agoRename TARGET_XFER_E_UNAVAILABLE to TARGET_XFER_UNAVAILABLE
Yao Qi [Sun, 23 Feb 2014 06:34:29 +0000 (14:34 +0800)]
Rename TARGET_XFER_E_UNAVAILABLE to TARGET_XFER_UNAVAILABLE

Nowadays, TARGET_XFER_E_UNAVAILABLE isn't regarded as an error in
to_xfer_partial interface, so _E_ looks odd.  This patch is to
replace TARGET_XFER_E_UNAVAILABLE with TARGET_XFER_UNAVAILABLE,
and change its value from -2 to 2.  Since there is no comparison
on the value of 'enum target_xfer_status', so it should be safe.

gdb:

2014-02-24  Yao Qi  <yao@codesourcery.com>

* target.h (enum target_xfer_status)
<TARGET_XFER_E_UNAVAILABLE>: Rename it to ...
<TARGET_XFER_UNAVAILABLE>: ... it with setting value 2
explicitly.  New.
* corefile.c (memory_error_message): User updated.
* exec.c (section_table_read_available_memory): Likewise.
* record-btrace.c (record_btrace_xfer_partial): Likewise.
* target.c (target_xfer_status_to_string): Likewise.
(raw_memory_xfer_partial): Likewise.
(memory_xfer_partial_1, target_xfer_partial): Likewise.
* valops.c (read_value_memory): Likewise.
* exec.h: Update comments.

10 years agoTweak target_xfer_status_to_string
Yao Qi [Sun, 23 Feb 2014 06:27:17 +0000 (14:27 +0800)]
Tweak target_xfer_status_to_string

This patch tweaks target_xfer_status_to_string on comments and argument
name.

gdb:

2014-02-24  Yao Qi  <yao@codesourcery.com>

* target.c (target_xfer_status_to_string): Rename argument err
to status.
* target.h (target_xfer_status_to_string): Update declaration.
Replace target_xfer_error_to_string with
target_xfer_status_to_string in comment.

10 years agoUnbreak mips native build
Yao Qi [Mon, 24 Feb 2014 04:03:05 +0000 (12:03 +0800)]
Unbreak mips native build

When I build mips native gdb today, I get the follow error,

../../../git/gdb/mips-linux-nat.c: In function '_initialize_mips_linux_nat':
../../../git/gdb/mips-linux-nat.c:792:15: error: assignment from incompatible pointer type [-Werror]
cc1: all warnings being treated as errors

It looks an oversight of recent target_ops delegation patches.  This
patch is to fix the build error.

gdb:

2014-02-24  Yao Qi  <yao@codesourcery.com>

* mips-linux-nat.c (super_close): Update its type.
(mips_linux_close): Pass 'self' to super_close.

10 years agoRemove TARGET_XFER_STATUS_ERROR_P
Yao Qi [Sun, 23 Feb 2014 04:24:08 +0000 (12:24 +0800)]
Remove TARGET_XFER_STATUS_ERROR_P

This patch removes macro TARGET_XFER_STATUS_ERROR_P, as Pedro pointed
out during patches review that TARGET_XFER_STATUS_ERROR_P tends to
be unnecessary.

gdb:

2014-02-24  Yao Qi  <yao@codesourcery.com>

* target.h (TARGET_XFER_STATUS_ERROR_P): Remove.
* corefile.c (read_memory): Adjusted.
* target.c (target_write_with_progress): Adjusted.

10 years agodaily update
Alan Modra [Sun, 23 Feb 2014 23:00:35 +0000 (09:30 +1030)]
daily update

10 years agoRevert previous tweaks
Yao Qi [Mon, 16 Dec 2013 13:13:36 +0000 (21:13 +0800)]
Revert previous tweaks

As we migrate to the new to_xfer_partial interface, some of previous
tweaks become unnecessary, we don't have to check traceframe is
selected in each target implementation, so this patch below is
reverted.

  [PATCH] Send qXfer:traceframe-info:read when traceframe is selected.
  https://sourceware.org/ml/gdb-patches/2013-10/msg00752.html

Third, to_traceframe_info is only called when traceframe is selected,
that means it is only called when target is remote, tfile or ctf, so
this patch can be partially reverted,

  https://sourceware.org/ml/gdb-patches/2013-04/msg00000.html

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

Revert two patches:

2013-10-25  Yao Qi  <yao@codesourcery.com>

* remote.c (remote_traceframe_info): Return early if
traceframe is not selected.

2013-07-19  Yao Qi  <yao@codesourcery.com>

* target.c (update_current_target): Change the default action
of 'to_traceframe_info' from tcomplain to return_zero.
* target.h (struct target_ops) <to_traceframe_info>: Add more
comments.

10 years agoAdjust read_value_memory to use to_xfer_partial
Yao Qi [Tue, 5 Nov 2013 01:50:21 +0000 (09:50 +0800)]
Adjust read_value_memory to use to_xfer_partial

As the new to_xfer_partial implementations are done in ctf and tfile
targets, read_value_memory can be simplified a lot.  Call
target_xfer_partial in a loop, check return value, and set bytes
unavailable when necessary.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* valops.c (read_value_memory): Rewrite it.  Call
target_xfer_partial in a loop.
* exec.h (section_table_available_memory): Remove declaration.
Move comments to ...
* exec.c (section_table_available_memory): ... here.  Make it static.

10 years agoUse new to_xfer_partial interface in ctf and tfile target
Yao Qi [Tue, 11 Feb 2014 04:20:05 +0000 (12:20 +0800)]
Use new to_xfer_partial interface in ctf and tfile target

This patch adjust both ctf and tfile target implementation of to_xfer_partial,
to return TARGET_XFER_E_UNAVAILABLE and set *XFERED_LEN if data is
unavailable.  Note that some code on xfer in exec.c can be shared, but
we can do it in a separate pass later.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* exec.c (section_table_read_available_memory): New function.
* exec.h (section_table_read_available_memory): Declare.
* ctf.c (ctf_xfer_partial): Call
section_table_read_available_memory.
* tracefile-tfile.c (tfile_xfer_partial): Likewise.

10 years agoShare code on to_xfer_partial for tfile and ctf target
Yao Qi [Mon, 10 Feb 2014 09:17:32 +0000 (17:17 +0800)]
Share code on to_xfer_partial for tfile and ctf target

In the to_xfer_partial implementations of ctf and tfile, the code on
reading from read-only sections is duplicated.  This patch moves it to
a separate function exec_read_partial_read_only.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* ctf.c (ctf_xfer_partial): Move code to ...
* exec.c (exec_read_partial_read_only): ... it.  New function.
* tracefile-tfile.c (tfile_xfer_partial): Likewise.
* tracefile.c: Include "exec.h".
* exec.h (exec_read_partial_read_only): Declare.

10 years agoLet tracefile has_memory and has_all_memory.
Yao Qi [Mon, 10 Feb 2014 07:28:33 +0000 (15:28 +0800)]
Let tracefile has_memory and has_all_memory.

At present, tfile target thinks it has memory but ctf doesn't.
This is an oversight when I added ctf target support.  This patch
moves the implementations of to_has_all_memory and to_has_memory to
upper layer.  After this change, both tfile and ctf target think
they have memory.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* tracefile-tfile.c (tfile_has_all_memory): Remove.
(tfile_has_memory): Remove.
(init_tfile_ops): Don't set fields to_has_all_memory and
to_has_memory of tfile_ops.
* tracefile.c (tracefile_has_all_memory): New function.
(tracefile_has_memory): New function.
(init_tracefile_ops): Initialize fields to_has_all_memory and
to_has_memory of 'ops'.

10 years agoShare some code between ctf and tfile target.
Yao Qi [Mon, 10 Feb 2014 07:22:55 +0000 (15:22 +0800)]
Share some code between ctf and tfile target.

This patch move the duplicated code between tfile and ctf
targets into file tracefile.c.  The common part of target_ops
fields are set in init_tracefile_ops.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* ctf.c (ctf_has_stack, ctf_has_registers): Remove.
(ctf_thread_alive, ctf_get_trace_status): Remove.
(init_ctf_ops): Don't set some fields of ctf_ops.  Call
init_tracefile_ops.
* tracefile-tfile.c (tfile_get_trace_status): Remove.
(tfile_has_stack, tfile_has_registers): Remove.
(tfile_thread_alive): Remove.
(init_tfile_ops): Don't set some fields of tfile_ops.  Call
init_tracefile_ops.
* tracefile.c (tracefile_has_stack): New function.
(tracefile_has_registers): New function.
(tracefile_thread_alive): New function.
(tracefile_get_trace_status): New function.
(init_tracefile_ops): New function.
* tracefile.h (init_tracefile_ops): Declare.

10 years agoMove tfile target to tracefile-tfile.c
Yao Qi [Mon, 10 Feb 2014 06:38:26 +0000 (14:38 +0800)]
Move tfile target to tracefile-tfile.c

This patch moves tfile target related code from tracepoint.c to
tracefile-tfile.c.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* tracepoint.c (TFILE_PID): Move it to tracefile-tfile.c.
(O_LARGEFILE): Likewise.
(tfile_ops): Likewise.
(TRACE_HEADER_SIZE): Likewise.
(trace_fd, trace_frames_offset, cur_offset): Likewise.
(cur_data_size): Likewise.
(tfile_read, tfile_open, tfile_interp_line): Likewise.
(tfile_close, tfile_files_info): Likewise.
(tfile_get_trace_status): Likewise.
(tfile_get_tracepoint_status): Likewise.
(tfile_get_traceframe_address): Likewise.
(tfile_trace_find, match_blocktype): Likewise.
(traceframe_walk_blocks, traceframe_find_block_type): Likewise.
(tfile_fetch_registers, tfile_xfer_partial): Likewise.
(tfile_get_trace_state_variable_value): Likewise.
(tfile_has_all_memory, tfile_has_memory): Likewise.
(tfile_has_stack, tfile_has_registers): Likewise.
(tfile_thread_alive, build_traceframe_info): Likewise.
(tfile_traceframe_info, init_tfile_ops): Likewise.
(_initialize_tracepoint): Don't call init_tfile_ops
and add_target_with_completer.
* tracefile-tfile.c: Include regcache.h, inferior.h, gdbthread.h,
exec.h, completer.h and filenames.h.
(_initialize_tracefile_tfile): New function.

10 years agoMove trace file writer out of tracepoint.c
Yao Qi [Mon, 10 Feb 2014 05:31:12 +0000 (13:31 +0800)]
Move trace file writer out of tracepoint.c

This patch is a refactor which moves trace file writer related code
out of tracepoint.c, which has 6k LOC.  It moves general trace file
writer to a new file tracefile.c and moves tfile specific writer to
tracefile-tfile.c.

gdb:

2014-02-23  Yao Qi  <yao@codesourcery.com>

* Makefile.in (REMOTE_OBS): Append tracefile.o and
tracefile-tfile.o.
(HFILES_NO_SRCDIR): Add tracefile.h.
* ctf.c: Include "tracefile.h".
* tracefile.h: New file.
* tracefile.c: New file
* tracefile-tfile.c: New file.
* tracepoint.c: Include "tracefile.h".
(free_uploaded_tps, free_uploaded_tsvs): Remove declarations.
(stop_reason_names): Add const.
(trace_file_writer_xfree): Move it to tracefile.c.
(trace_save, trace_save_command, trace_save_tfile): Likewise.
(trace_save_ctf): Likewise.
(struct tfile_trace_file_writer): Move it to tracefile-tfile.c.
(tfile_target_save, tfile_dtor, tfile_start): Likewise.
(tfile_write_header, tfile_write_regblock_type): Likewise.
(tfile_write_status, tfile_write_uploaded_tsv): Likewise.
(tfile_write_uploaded_tp, tfile_write_definition_end): Likewise.
(tfile_write_raw_data, tfile_end): Likewise.
(tfile_trace_file_writer_new): Likewise.
(free_uploaded_tp): Make it extern.
(free_uploaded_tsv): Make it extern.
(_initialize_tracepoint): Move code to register command 'tsave'
to tracefile.c.
* tracepoint.h (stop_reason_names): Declare.
(struct trace_frame_write_ops): Move it to tracefile.h.
(struct trace_file_write_ops): Likewise.
(struct trace_file_writer): Likewise.
(free_uploaded_tsvs, free_uploaded_tps): Declare.

10 years agodaily update
Alan Modra [Sat, 22 Feb 2014 23:00:37 +0000 (09:30 +1030)]
daily update

10 years agodaily update
Alan Modra [Fri, 21 Feb 2014 23:00:38 +0000 (09:30 +1030)]
daily update

10 years agoFix crash on process name "(sd-pam)" (PR 16594).
Jan Kratochvil [Fri, 21 Feb 2014 17:39:40 +0000 (18:39 +0100)]
Fix crash on process name "(sd-pam)" (PR 16594).

info os processes -fsanitize=address error
https://sourceware.org/bugzilla/show_bug.cgi?id=16594

info os processes
=================================================================
==5795== ERROR: AddressSanitizer: heap-use-after-free on address
0x600600214974 at pc 0x757a92 bp 0x7fff95dd9f00 sp 0x7fff95dd9ef0
READ of size 4 at 0x600600214974 thread T0
    #0 0x757a91 in get_cores_used_by_process (.../gdb/gdb+0x757a91)

At least Fedora 20 has process(es):
 6678 ?        Ss     0:00 /usr/lib/systemd/systemd --user
 6680 ?        S      0:00  \_ (sd-pam)

and GDB "info os processes" crashes on it as /proc/6680/stat contains:

6680 ((sd-pam)) S 6678 6678 6678 0 -1 1077961024 33 0 0 0 0 0 0 0 20 0 1 0 18568 73768960 120 18446744073709551615 1 1
0 0 0 0 0 4096 0 18446744073709551615 0 0 17 6 0 0 0 0 0 0 0 0 0 0 0 0 0

and GDB fails to find the proper end of the process name "((sd-pam))".
Therefore it reads core number off-by-one (it reads 17 instead of 6) and
overruns the array.

(1) Make the process name parsing more foolproof.

(2) Do not trust the parsed number from /proc/PID/stat and verify it against
    the array size.

I noticed that 'ps' gets this right, so I've peeked at its
sources, and it just looks for the first ')' starting at
the end.

https://gitorious.org/procps/procps/source/dc072aced7250fed9b01fb05f0d672678752a63e:proc/readproc.c

Look for stat2proc.

Given ps does that, I believe the kernel won't ever be changed
in a way that would break it.  So it sounds like could do strrchr
from the end of stat just as well without worry, which is simpler.

gdb/
2014-02-21  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR gdb/16594
* common/linux-osdata.c (linux_common_core_of_thread): Find the end of
process name.
(get_cores_used_by_process): New parameter num_cores, use it.
(linux_xfer_osdata_processes): Pass num_cores to it.
* linux-tdep.c (linux_info_proc, linux_fill_prpsinfo): Find the end of
process name.

Message-ID: <20140217212826.GA15080@host2.jankratochvil.net>

10 years agoAdd support for CPUID PREFETCHWT1
Ilya Tocar [Thu, 20 Feb 2014 14:57:31 +0000 (18:57 +0400)]
Add support for CPUID PREFETCHWT1

Latest AVX512 spec
http://download-software.intel.com/sites/default/files/managed/50/1a/319433-018.pdf
Has CPUID PREFETCHWT1 for prefetchwt1 instruction, which we list as AVX512PF.
This patch introduces CPUID PREFETCHWT1.

gas/

        * config/tc-i386.c (cpu_arch): Add .prefetchwt1.
        * doc/c-i386.texi: Document .prefetchwt1/prefetchwt1.

opcodes/

        * i386-gen.c (cpu_flag_init): Add CPU_PREFETCHWT1_FLAGS/
        (cpu_flags): Add CpuPREFETCHWT1.
        * i386-init.h: Regenerate.
        * i386-opc.h (CpuPREFETCHWT1): New.
        (i386_cpu_flags): Add cpuprefetchwt1.
        * i386-opc.tbl: Cahnge CPU of prefetchwt1 from CpuAVX512PF to CpuPREFETCHWT1.
        * i386-tbl.h: Regenerate.

gas/testsuite

        * gas/i386/avx512pf-intel.d: Remove prefetchwt1.
        * gas/i386/avx512pf.s: Ditto.
        * gas/i386/avx512pf.d: Ditto.
        * gas/i386/x86-64-avx512pf-intel.d: Ditto.
        * gas/i386/x86-64-avx512pf.s: Ditto.
        * gas/i386/x86-64-avx512pf.d: Ditto.
        * gas/i386/prefetchwt1-intel.d: New file.
        * gas/i386/prefetchwt1.s: Ditto.
        * gas/i386/prefetchwt1.d: Ditto.
        * gas/i386/x86-64-prefetchwt1-intel.d: Ditto.
        * gas/i386/x86-64-prefetchwt1.s: Ditto.
        * gas/i386/x86-64-prefetchwt1.d: Ditto.

10 years agoFix length arg in call to breakpoint_xfer_memory.
Andreas Krebbel [Fri, 21 Feb 2014 14:55:37 +0000 (15:55 +0100)]
Fix length arg in call to breakpoint_xfer_memory.

The patch "return target_xfer_status in to_xfer_partial" caused a
regression in various s390(x) test cases, because memory_xfer_partial
filled only the first byte of the read buffer from a breakpoint shadow:

    https://sourceware.org/ml/gdb-patches/2014-01/msg01071.html

This patch fixes the regression.

10 years agoRename test.
Pedro Alves [Fri, 21 Feb 2014 13:30:12 +0000 (13:30 +0000)]
Rename test.

I realized that the name of this test only made sense when considering
the old (never committed) implementation of the fix that came along
with the test originally, that forced a schedlock while a step-resume
(to get over the signal handler) was inserted.  The final solution
that went into the tree does not force that locking.

So this renames it to something more descriptive.

gdb/testsuite/
2014-02-21  Pedro Alves  <palves@redhat.com>

* gdb.threads/step-after-sr-lock.c: Rename to ...
* gdb.threads/signal-while-stepping-over-bp-other-thread.c: ... this.
* gdb.threads/step-after-sr-lock.exp: Rename to ...
* gdb.threads/signal-while-stepping-over-bp-other-thread.exp:
... this.

10 years agoAdd file missed from last commit.
Alan Modra [Fri, 21 Feb 2014 12:48:50 +0000 (23:18 +1030)]
Add file missed from last commit.

10 years agoAdd ppc476 workaround bootstrap test
Alan Modra [Thu, 20 Feb 2014 23:48:01 +0000 (10:18 +1030)]
Add ppc476 workaround bootstrap test

I was running this by hand to test out --ppc476-workaround.  Another
bootstrap test doesn't take all that long, so let's add it to the
testsuite.

* ld-bootstrap/bootstrap.exp: Add ppc476 workaround test.
* ld-bootstrap/ppc476.t: New file.

10 years agodaily update
Alan Modra [Thu, 20 Feb 2014 23:00:38 +0000 (09:30 +1030)]
daily update

10 years agoFix for PR tdep/16397: SystemTap SDT probe support for x86 doesn't work with "triplet...
Sergio Durigan Junior [Thu, 20 Feb 2014 21:49:09 +0000 (18:49 -0300)]
Fix for PR tdep/16397: SystemTap SDT probe support for x86 doesn't work with "triplet operands"

This is the continuation of what Joel proposed on:

<https://sourceware.org/ml/gdb-patches/2013-12/msg00977.html>

Now that I have already submitted and pushed the patch to split
i386_stap_parse_special_token into two smaller functions, it is indeed
simpler to understand this patch.

It occurs because, on x86, triplet displacement operands are allowed
(like "-4+8-20(%rbp)"), and the current parser for this expression is
buggy.  It does not correctly extract the register name from the
expression, which leads to incorrect evaluation.  The parser was also
being very "generous" with the expression, so I included a few more
checks to ensure that we're indeed dealing with a triplet displacement
operand.

This patch also includes testcases for the two different kind of
expressions that can be encountered on x86: the triplet displacement
(explained above) and the three-argument displacement (as in
"(%rbx,%ebx,-8)").  The tests are obviously arch-dependent and are
placed under gdb.arch/.

Message-ID: <m3mwj1j12v.fsf@redhat.com>
URL: <https://sourceware.org/ml/gdb-patches/2014-01/msg00310.html>

gdb/
2014-02-20  Sergio Durigan Junior  <sergiodj@redhat.com>

PR tdep/16397
* i386-tdep.c (i386_stap_parse_special_token_triplet): Check if a
number comes after the + or - signs.  Adjust length of register
name to be extracted.

gdb/testsuite/
2014-02-20  Sergio Durigan Junior  <sergiodj@redhat.com>

PR tdep/16397
* gdb.arch/amd64-stap-special-operands.exp: New file.
* gdb.arch/amd64-stap-three-arg-disp.S: Likewise.
* gdb.arch/amd64-stap-three-arg-disp.c: Likewise.
* gdb.arch/amd64-stap-triplet.S: Likewise.
* gdb.arch/amd64-stap-triplet.c: Likewise.

10 years agoAdd single sentence description to top of entry of previous commit.
Doug Evans [Thu, 20 Feb 2014 21:33:44 +0000 (13:33 -0800)]
Add single sentence description to top of entry of previous commit.