Introduced a regression when compiling for mingw*:
/gdb/common/pathstuff.c: In function 'gdb::unique_xmalloc_ptr<char>
gdb_realpath(const char*)':
/gdb/common/pathstuff.c:56:14: error: 'MAX_PATH' was not declared in this scope
char buf[MAX_PATH];
^
/gdb/common/pathstuff.c:57:5: error: 'DWORD' was not declared in this scope
DWORD len = GetFullPathName (filename, MAX_PATH, buf, NULL);
^
/gdb/common/pathstuff.c:57:11: error: expected ';' before 'len'
DWORD len = GetFullPathName (filename, MAX_PATH, buf, NULL);
^
/gdb/common/pathstuff.c:63:9: error: 'len' was not declared in this scope
if (len > 0 && len < MAX_PATH)
^
/gdb/common/pathstuff.c:64:54: error: 'buf' was not declared in this scope
return gdb::unique_xmalloc_ptr<char> (xstrdup (buf));
^
make[2]: *** [pathstuff.o] Error 1
The proper fix is to conditionally include "<windows.h>". This commit
does that, without introducing any regressions as per tests made by
our BuildBot.
Change order of error message printed when gdbserver can't find CWD
I forgot to address Pedro's comment about my last patch and change the
order of the message printed when getcwd returns NULL on gdbserver.
This obvious commit does it.
Simon mentioned on IRC that, after the startup-with-shell feature has
been implemented on gdbserver, it is not possible to specify a
filename-only binary, like:
$ gdbserver :1234 a.out
/bin/bash: line 0: exec: a.out: not found
During startup program exited with code 127.
Exiting
This happens on systems where the current directory "." is not listed
in the PATH environment variable. Although including "." in the PATH
variable is a possible workaround, this can be considered a regression
because before startup-with-shell it was possible to use only the
filename (due to reason that gdbserver used "exec*" directly).
The idea of the patch is to verify if the program path provided by the
user (or by the remote protocol) contains a directory separator
character. If it doesn't, it means we're dealing with a filename-only
binary, so we call "gdb_abspath" to properly expand it and transform
it into a full path. Otherwise, we leave the program path untouched.
This mimicks the behaviour seen on GDB (look at "openp" and
"attach_inferior", for example).
I am also submitting a testcase which exercises the scenario described
above. This test requires gdbserver to be executed in a different CWD
than the original, so I also created a helper function, "with_cwd" (on
testsuite/lib/gdb.exp), which takes care of cd'ing into and out of the
specified dir.
Built and regtested on BuildBot, without regressions.
gdb/ChangeLog:
2018-02-28 Sergio Durigan Junior <sergiodj@redhat.com>
Simon Marchi <simon.marchi@polymtl.ca>
* common/common-utils.c: Include "sys/stat.h".
(is_regular_file): Move here from "source.c"; change return
type to "bool".
* common/common-utils.h (is_regular_file): New prototype.
* common/pathstuff.c (contains_dir_separator): New function.
* common/pathstuff.h (contains_dir_separator): New prototype.
* source.c: Don't include "sys/stat.h".
(is_regular_file): Move to "common/common-utils.c".
* server.c: Include "filenames.h" and "pathstuff.h".
(program_name): Delete variable.
(program_path): New anonymous class.
(get_exec_wrapper): Use "program_path" instead of
"program_name".
(handle_v_run): Likewise.
(captured_main): Likewise.
(process_serial_event): Likewise.
This commit moves the path manipulation routines found on utils.c to a
new common/pathstuff.c, and updates the Makefile.in's accordingly.
The routines moved are "gdb_realpath", "gdb_realpath_keepfile" and
"gdb_abspath".
This will be needed because gdbserver will have to call "gdb_abspath"
on my next patch, which implements a way to expand the path of the
inferior provided by the user in order to allow specifying just the
binary name when starting gdbserver, like:
$ gdbserver :1234 a.out
With the recent addition of the startup-with-shell feature on
gdbserver, this scenario doesn't work anymore if the user doesn't have
the current directory listed in the PATH variable.
I had to do a minor adjustment on "gdb_abspath" because we don't have
access to "tilde_expand" on gdbserver, so now the function is using
"gdb_tilde_expand" instead. Otherwise, the code is the same.
Regression tested on the BuildBot, without regressions.
Yao Qi [Thu, 15 Feb 2018 14:48:30 +0000 (14:48 +0000)]
Reset inferior::control on inferior exit
When we kill an inferior, the inferior is not deleted. What is more, it
is reused when the new process is created, so we need to reset inferior's
state when it exits.
Eli Zaretskii [Sat, 27 Jan 2018 16:34:46 +0000 (18:34 +0200)]
Avoid compilation errors in MinGW native builds of GDB
The error is triggered by including python-internal.h, and the
error message is:
In file included from d:\usr\lib\gcc\mingw32\6.3.0\include\c++\math.h:36:0,
from build-gnulib/import/math.h:27,
from d:/usr/Python26/include/pyport.h:235,
from d:/usr/Python26/include/Python.h:58,
from python/python-internal.h:94,
from python/py-arch.c:24:
d:\usr\lib\gcc\mingw32\6.3.0\include\c++\cmath:1157:11: error: '::hypot' has not been declared
using ::hypot;
^~~~~
This happens because Python headers define 'hypot' to expand to
'_hypot' in the Windows builds.
gdb/ChangeLog:
2018-01-27 Eli Zaretskii <eliz@gnu.org>
* python/python-internal.h (_hypot) [__MINGW32__]: Define back to
'hypoth'. This avoids a compilation error.
In some cases objects of type range_bounds are memset when allocated,
but then their dynamic_prop low/high fields are copied over from some
template dynamic_prop object that wasn't memset. E.g.,
create_static_range_type's low/high locals are left with garbage in
the padding, and then that padding is copied over to the range_bounds
object's low/high fields.
At first, I considered making sure to always memset range_bounds
objects, thinking that maybe type objects are being put in some bcache
instance somewhere. But then I hacked bcache/bcache_full to poison
non-pod types, and made dynamic_prop a non-pod, and GDB still
compiled.
So given that, it seems safest to not assume padding will always be
memset, and instead treat them as regular value types, implementing
(in)equality operators and using those instead of memcmp.
Pedro Alves [Mon, 22 Jan 2018 19:54:17 +0000 (19:54 +0000)]
Fix segfault with 'set print object on' + 'whatis <struct>' & co
Compiling GDB with a recent GCC exposes a problem:
../../gdb/typeprint.c: In function 'void whatis_exp(const char*, int)':
../../gdb/typeprint.c:515:12: warning: 'val' may be used uninitialized in this function [-Wmaybe-uninitialized]
real_type = value_rtti_type (val, &full, &top, &using_enc);
~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The warning is correct. There are indeed code paths that use
uninitialized 'val', leading to crashes. Inside the
value_rtti_indirect_type/value_rtti_type calls here in whatis_exp:
because "whatis" with a type argument enters this branch:
/* The behavior of "whatis" depends on whether the user
expression names a type directly, or a language expression
(including variable names). If the former, then "whatis"
strips one level of typedefs, only. If an expression,
"whatis" prints the type of the expression without stripping
any typedef level. "ptype" always strips all levels of
typedefs. */
if (show == -1 && expr->elts[0].opcode == OP_TYPE)
{
which does not initialize VAL. Trying the above triggers crashes like
this:
(gdb) set print object on
(gdb) whatis some_structure_type
Thread 1 "gdb" received signal SIGSEGV, Segmentation fault.
0x00000000005dda90 in check_typedef (type=0x6120736573756170) at src/gdb/gdbtypes.c:2388
2388 int instance_flags = TYPE_INSTANCE_FLAGS (type);
...
This is a regression caused by a recent-ish refactoring of the code on
'whatis_exp', introduced by:
Fix warning on gdb/compile/compile.c (C++-ify "triplet_rx")
This fixes a GCC warning that happens when compiling
gdb/compile/compile.c on some GCC versions (e.g., "gcc (GCC) 7.2.1 20180104 (Red Hat 7.2.1-6)"):
../../gdb/compile/compile.c: In function 'void eval_compile_command(command_line*, const char*, compile_i_scope_types, void*)':
../../gdb/compile/compile.c:548:19: warning: 'triplet_rx' may be used uninitialized in this function [-Wmaybe-uninitialized]
error_message = compiler->fe->ops->set_arguments_v0 (compiler->fe, triplet_rx,
~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
argc, argv);
~~~~~~~~~~~
../../gdb/compile/compile.c:466:9: note: 'triplet_rx' was declared here
char *triplet_rx;
^~~~~~~~~~
It's a simple patch that converts "triplet_rx" from "char *" to
"std::string", thus guaranteeing that it will be always initialized.
I've regtested this patch and did not find any regressions. OK to
apply on both master and 8.1 (after creating a bug for it)?
Eldar Abusalimov [Wed, 17 Jan 2018 13:13:15 +0000 (13:13 +0000)]
configure: Fix test for fs_base/gs_base in <sys/user.h>
Make <sys/types.h> be included prior to including <sys/user.h>.
glibc versions older than 2.14 use __uintNN_t types within certain
structures defined in <sys/user.h> probably assuming these types are
defined prior to including the header. This results in the following
`configure` feature test compilation error that makes it think that
`struct user_regs_struct` doesn't have `fs_base`/`gs_base` fields,
althouh it does.
configure:13617: checking for struct user_regs_struct.fs_base
configure:13617: gcc -c -g -O2 -I/linux/include conftest.c >&5
In file included from conftest.c:158:0:
/usr/include/sys/user.h:32:3: error: unknown type name '__uint16_t'
__uint16_t cwd;
^
/usr/include/sys/user.h:33:3: error: unknown type name '__uint16_t'
__uint16_t swd;
^
/usr/include/sys/user.h:34:3: error: unknown type name '__uint16_t'
__uint16_t ftw;
^
/usr/include/sys/user.h:35:3: error: unknown type name '__uint16_t'
__uint16_t fop;
^
/usr/include/sys/user.h:36:3: error: unknown type name '__uint64_t'
__uint64_t rip;
^
/usr/include/sys/user.h:37:3: error: unknown type name '__uint64_t'
__uint64_t rdp;
^
/usr/include/sys/user.h:38:3: error: unknown type name '__uint32_t'
__uint32_t mxcsr;
^
/usr/include/sys/user.h:39:3: error: unknown type name '__uint32_t'
__uint32_t mxcr_mask;
^
/usr/include/sys/user.h:40:3: error: unknown type name '__uint32_t'
__uint32_t st_space[32]; /* 8*16 bytes for each FP-reg = 128 bytes */
^
/usr/include/sys/user.h:41:3: error: unknown type name '__uint32_t'
__uint32_t xmm_space[64]; /* 16*16 bytes for each XMM-reg = 256 bytes */
^
/usr/include/sys/user.h:42:3: error: unknown type name '__uint32_t'
__uint32_t padding[24];
^
configure:13617: $? = 1
configure: failed program was:
| /* confdefs.h */
...
| /* end confdefs.h. */
| #include <sys/user.h>
|
| int
| main ()
| {
| static struct user_regs_struct ac_aggr;
| if (ac_aggr.fs_base)
| return 0;
| ;
| return 0;
| }
Recent glibc versions don't use typedef'ed int types in <sys/user.h>,
thus allowing it to be included as is
(glibc commit d79a9c949c84e7f0ba33e87447c47af833e9f11a).
However there're still some distros alive that use older glibc,
for instance, RHEL/CentOS 6 package glibc 2.12.
As noted by Andrew Paprocki, who submitted the PR
(https://sourceware.org/bugzilla/show_bug.cgi?id=21559#c3):
> It should be noted that modifying `configure` to force on
> `HAVE_STRUCT_USER_REGS_STRUCT_FS_BASE` and
> `HAVE_STRUCT_USER_REGS_STRUCT_GS_BASE` fixes this issue. For some
> reason the `configure` tests for `fs_base` and `gs_base` fail
> even though `sys/user.h` on RHEL5 has the fields defined in
> `user_regs_struct`.
Note that this patch does NOT fix the root cause of PR gdb/21559,
although now that `configure` properly detects the presence of the
fields and sets HAVE_XXX accordingly, the execution takes another
path, which doesn't lead to the assertion failure in question.
gdb/ChangeLog:
2018-01-17 Eldar Abusalimov <eldar.abusalimov@jetbrains.com>
PR gdb/21559
* configure.ac: Include <sys/types.h> prior to <sys/user.h> when
checking for fs_base/gs_base fields in struct user_regs_struct.
* configure: Regenerate.
gdb/gdbserver/ChangeLog:
2018-01-17 Eldar Abusalimov <eldar.abusalimov@jetbrains.com>
PR gdb/21559
* configure.ac: Include <sys/types.h> prior to <sys/user.h> when
checking for fs_base/gs_base fields in struct user_regs_struct.
* configure: Regenerate.
Yao Qi [Wed, 17 Jan 2018 11:19:03 +0000 (11:19 +0000)]
Relax gdb.compile/compile.exp to match the address printed for frame
One test in gdb.compile/compile.exp passes on one fedora builder,
bt
#0 0x00007ffff7ff43f6 in _gdb_expr (__regs=0x7ffff7ff2000) at gdb
command line:1^M
#1 <function called from gdb>^M
#2 main () at /home/gdb-buildbot/fedora-x86-64-1/fedora-x86-64/build/gdb/testsuite/../../../binutils-gdb/gdb/testsuite/gdb.compile/compile.c:106^M
(gdb) PASS: gdb.compile/compile.exp: bt
but fails on my machine with gcc trunk,
bt^M
#0 _gdb_expr (__regs=0x7ffff7ff3000) at gdb command line:1^M
#1 <function called from gdb>^M
#2 main () at gdb/testsuite/gdb.compile/compile.c:106^M
(gdb) FAIL: gdb.compile/compile.exp: bt
The test should be tweaked to match both cases (pc in the start of line
vs pc in the middle of line). Note that I am not clear that why libcc1
emits debug info this way so that the address is in the middle of line.
gdb/testsuite:
2018-01-17 Yao Qi <yao.qi@linaro.org>
* gdb.compile/compile.exp: Match the address printed for
frame in the output of command "bt".
Tom Tromey [Wed, 3 Jan 2018 18:12:34 +0000 (11:12 -0700)]
Fix scm-ports.exp regression
In https://sourceware.org/ml/gdb-patches/2017-12/msg00215.html, Jan
pointed out that the scalar printing patches caused a regression in
scm-ports.exp on x86.
What happens is that on x86, this:
set sp_reg [get_integer_valueof "\$sp" 0]
... ends up setting sp_reg to a negative value, because
get_integer_valueof uses "print/d":
Pedro Alves [Fri, 12 Jan 2018 18:59:40 +0000 (18:59 +0000)]
Add testcase for GDB hang fixed by previous commit
This adds a testcase for the previous commit. The regression was
related to in-line step overs. The reason we didn't see it on native
x86-64/s390 GNU/Linux testing is that native debugging uses displaced
stepping by default (because native debugging defaults to "maint set
target-non-stop on"), unlike remote debugging.
So in order to trigger the bug with native debugging as well, the
testcase disables displaced stepping explicitly.
Also, instead of using watchpoints to trigger the regression, the
testcase uses a breakpoint at address 0, which should be more
portable.
gdb/testsuite/ChangeLog:
2018-01-12 Pedro Alves <palves@redhat.com>
Andreas Arnez [Fri, 12 Jan 2018 18:59:40 +0000 (18:59 +0000)]
Fix GDB hang with remote after error from resume
Since this commit --
Fix PR18360 - internal error when using "interrupt -a"
(https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=c65d6b55)
-- the testsuite shows long delays on s390 with native-gdbserver when
executing certain tests, such as watchpoints.exp. These hangs have been
discussed before in the context of buildbot problems, see here:
The problem can easily be triggered by stopping on a breakpoint, then
setting impossible watchpoints, and finally doing "continue". Then, after
having set the step-over state (in keep_going_pass_signal in infrun.c),
GDB tries to insert breakpoints and watchpoints into the inferior. This
fails, and the "continue" command is aborted. But the step-over state is
not cleared in this case, which causes future step-over attempts to be
skipped since GDB thinks that "we already have an in-line step-over
operation ongoing" (see start_step_over in infrun.c). Thus the next
"continue" just goes on to wait for events from the remote, which will
never occur.
The problem can also be reproduced on amd64 with native-gdbserver, using
the following change to watchpoints.exp:
To fix the hang, this patch clears the step-over info when
insert_breakpoints has failed. Of course, with native-gdbserver the
watchpoints.exp test case still causes many FAILs on s390, because
gdbserver does not support watchpoints for that target. This is a
separate issue.
gdb/ChangeLog:
2018-01-12 Andreas Arnez <arnez@linux.vnet.ibm.com>
* infrun.c (keep_going_pass_signal): Clear step-over info when
insert_breakpoints fails.
caused regressions on software single-set targets, specifically:
FAIL: gdb.base/breakpoint-in-ro-region.exp: always-inserted off: auto-hw off: single-step breakpoint is not left behind
FAIL: gdb.base/breakpoint-in-ro-region.exp: always-inserted off: auto-hw on: single-step breakpoint is not left behind
FAIL: gdb.base/breakpoint-in-ro-region.exp: always-inserted on: auto-hw off: step in ro region (cannot insert hw break)
FAIL: gdb.base/breakpoint-in-ro-region.exp: always-inserted on: auto-hw off: single-step breakpoint is not left behind
FAIL: gdb.base/breakpoint-in-ro-region.exp: always-inserted on: auto-hw on: single-step breakpoint is not left behind
and indeed detailed logs indicate a breakpoint is left lingering, e.g.:
(gdb) PASS: gdb.base/breakpoint-in-ro-region.exp: always-inserted off: auto-hw off: step in ro region (cannot insert sw break)
maint info breakpoints 0
Num Type Disp Enb Address What
0 sw single-step keep y 0x00400774 in main at [...]/gdb/testsuite/gdb.base/breakpoint-in-ro-region.c:24 inf 1 thread 1
stop only in thread 1
(gdb) FAIL: gdb.base/breakpoint-in-ro-region.exp: always-inserted off: auto-hw off: single-step breakpoint is not left behind
vs:
(gdb) PASS: gdb.base/breakpoint-in-ro-region.exp: always-inserted off: auto-hw off: step in ro region (cannot insert sw break)
maint info breakpoints 0
No breakpoint or watchpoint matching '0'.
(gdb) PASS: gdb.base/breakpoint-in-ro-region.exp: always-inserted off: auto-hw off: single-step breakpoint is not left behind
Before commit d930703d68ae, we had a cleanup installed in 'resume'
that would delete single-step breakpoints on error:
/* Resuming. */
/* Things to clean up if we QUIT out of resume (). */
static void
resume_cleanups (void *ignore)
{
if (!ptid_equal (inferior_ptid, null_ptid))
delete_single_step_breakpoints (inferior_thread ());
normal_stop ();
}
That whole function was removed by d930703d68ae mainly to eliminate
the normal_stop call:
~~~~
Note that the exception called from within resume ends up calling
normal_stop via resume_cleanups. That's very borked though, because
normal_stop is going to re-handle whatever was the last reported
event, possibly even re-running a hook stop...
~~~~
But as the regression shows, removing resume_cleanups completely went
a bit too far, as the delete_single_step_breakpoints call is still
necessary.
So fix the regression by reinstating the
delete_single_step_breakpoints call on error. However, since we're
trying to eliminate cleanups, restore it in a different form (using
TRY/CATCH).
Tested on x86-64 GNU/Linux both top of master and on top of a series
that implements software single-step on x86.
gdb/ChangeLog:
2018-01-11 Pedro Alves <palves@redhat.com>
PR gdb/22583
* infrun.c (resume): Rename to ...
(resume_1): ... this.
(resume): Reimplement as wrapper around resume_1.
Pedro Alves [Thu, 11 Jan 2018 00:24:59 +0000 (00:24 +0000)]
Fix backwards compatibility with old GDBservers (PR remote/22597)
At <https://sourceware.org/ml/gdb-patches/2017-12/msg00285.html>,
Maciej reported that commit:
commit 5cd63fda035d4ba949e6478406162c4673b3c9ef
Date: Wed Oct 4 18:21:10 2017 +0100
Subject: Fix "Remote 'g' packet reply is too long" problems with multiple inferiors
made GDB stop working with older stubs. Any attempt to continue
execution after the initial connection fails with:
[...]
Process .../gdb/testsuite/outputs/gdb.base/advance/advance created; pid = 2670
Listening on port 2346
target remote [...]:2346
Remote debugging using [...]:2346
Reading symbols from .../lib64/ld.so.1...done.
[Switching to Thread <main>]
(gdb) continue
Cannot execute this command without a live selected thread.
(gdb)
The problem is:
(gdb) c
Cannot execute this command without a live selected thread.
(gdb) info threads
Id Target Id Frame
1 Thread 14917 0x00007f341cd98ed0 in _start () from /lib64/ld-linux-x86-64.so.2
The current thread <Thread ID 2> has terminated. See `help thread'.
^^^^^^^^^^^
(gdb)
Note, thread _2_. There's really only one thread in the inferior
(it's still at the entry point), but still GDB added a bogus second
thread.
The reason GDB started adding a second thread after 5cd63fda035d is
this hunk:
Note the else branch that falls back to magic_null_ptid. We reach
that when we process the initial stop reply sent back in response to
the the "?" (status) packet early in the connection setup:
meaning, when we process that stop reply, we treat the event as coming
from a thread with ptid == magic_null_ptid, which is not yet in the
thread list, so we add it then:
(top-gdb) p ptid
$1 = {m_pid = 42000, m_lwp = -1, m_tid = 1}
(top-gdb) bt
#0 0x0000000000840a8c in add_thread_silent(ptid_t) (ptid=...) at src/gdb/thread.c:269
#1 0x00000000007ad61d in remote_add_thread(ptid_t, int, int) (ptid=..., running=0, executing=0)
at src/gdb/remote.c:1838
#2 0x00000000007ad8de in remote_notice_new_inferior(ptid_t, int) (currthread=..., executing=0)
at src/gdb/remote.c:1921
#3 0x00000000007b758b in process_stop_reply(stop_reply*, target_waitstatus*) (stop_reply=0x1158860, status=0x7fffffffcc00)
at src/gdb/remote.c:7217
#4 0x00000000007b7a38 in remote_wait_as(ptid_t, target_waitstatus*, int) (ptid=..., status=0x7fffffffcc00, options=0)
at src/gdb/remote.c:7380
#5 0x00000000007b7cd1 in remote_wait(target_ops*, ptid_t, target_waitstatus*, int) (ops=0x102fac0 <remote_ops>, ptid=..., status=0x7fffffffcc00, options=0) at src/gdb/remote.c:7446
#6 0x000000000081587b in delegate_wait(target_ops*, ptid_t, target_waitstatus*, int) (self=0x102fac0 <remote_ops>, arg1=..., arg2=0x7fffffffcc00, arg3=0) at src/gdb/target-delegates.c:138
#7 0x0000000000827d77 in target_wait(ptid_t, target_waitstatus*, int) (ptid=..., status=0x7fffffffcc00, options=0)
at src/gdb/target.c:2179
#8 0x0000000000715fda in do_target_wait(ptid_t, target_waitstatus*, int) (ptid=..., status=0x7fffffffcc00, options=0)
at src/gdb/infrun.c:3589
#9 0x0000000000716351 in wait_for_inferior() () at src/gdb/infrun.c:3707
#10 0x0000000000715435 in start_remote(int) (from_tty=1) at src/gdb/infrun.c:3212
things go downhill from this.
We don't see the problem with current master gdbserver, because that
version always sends the ";thread:" part in the initial stop reply:
Years ago I had added a "--disable-packet=" command line option to
gdbserver which comes in handy for testing this, since the existing
"--disable-packet=Tthread" precisely makes gdbserver not send that
";thread:" part in stop replies. The testcase added by this commit
emulates old gdbserver making use of that.
I've compared a testrun at 5cd63fda035d^ (before regression) with
'current master+patch', against old gdbserver at f8b73d13b7ca^. I
hacked out --once, and "monitor exit" to be able to test. The results
are a bit too unstable to tell accurately, but it looked like there
were no regressions. Maciej confirmed this worked for him as well.
No regressions on master (against master gdbserver).
gdb/ChangeLog:
2018-01-11 Pedro Alves <palves@redhat.com>
PR remote/22597
* remote.c (remote_parse_stop_reply): Default to the last-set
general thread instead of to 'magic_null_ptid'.
gdb/testsuite/ChangeLog:
2018-01-11 Pedro Alves <palves@redhat.com>
PR remote/22597
* gdb.server/stop-reply-no-thread.c: New file.
* gdb.server/stop-reply-no-thread.exp: New file.
Pedro Alves [Wed, 10 Jan 2018 23:14:07 +0000 (23:14 +0000)]
Ada: make verbatim matcher override other language matchers (PR gdb/22670)
A previous patch fixed verbatim matching in the lookup at the minimal
symbol level, but we should also be finding that same symbol through
the partial/full symtab search.
For example, this is what happens if we use "print" instead of
"break":
(gdb) p <MixedCaseFunc>
$1 = {<text variable, no debug info>} 0x4024dc <MixedCaseFunc>
Before the C++ wildmatching series, GDB knows that MixedCaseFunc is a
function without parameters, and the expression above means calling
it. If you try it before having started the inferior, you'd get the
following (expected) error:
(gdb) print <MixedCaseFunc>
You can't do that without a process to debug.
The main idea behind making the name matcher be determined by the
symbol's language is so that C++ (etc.) wildmatching in linespecs
works even if the current language is not C++, as e.g., when you step
through C or assembly code.
Ada's verbatim matching syntax however ("<...>") isn't quite the same.
It is more a property of the current language than of a particular
symbol's language. We want to support this syntax when debugging an
Ada program, but it's reason of existence is to find non-Ada symbols.
This suggests going back to enabling it depending on current language
instead of language of the symbol being matched.
I'm not entirely happy with the "current_language" reference (though I
think that it's harmless). I think we could try storing the current
language in the lookup_name_info object, and then convert a bunch of
functions more to pass around lookup_name_info objects instead of
"const char *" names. I.e., build the lookup_name_info higher up.
I'm not sure about that, I'll have to think more about it. Maybe
something different will be better. Meanwhile, this gets us going.
I've extended the testcase to also exercise a no-debug-info function,
for extra coverage of the minsyms-only paths.
gdb/ChangeLog:
2018-01-10 Pedro Alves <palves@redhat.com>
PR gdb/22670
* dwarf2read.c
(gdb_index_symbol_name_matcher::gdb_index_symbol_name_matcher):
Adjust to use language_get_symbol_name_matcher instead of
language_defn::la_get_symbol_name_matcher.
* language.c (language_get_symbol_name_matcher): If in Ada mode
and the lookup name is a verbatim match, return Ada's matcher.
* language.h (language_get_symbol_name_matcher): Adjust comment.
(ada_lookup_name_info::verbatim_p):: New method.
gdb/testsuite/ChangeLog:
2018-01-10 Pedro Alves <palves@redhat.com>
PR gdb/22670
* gdb.ada/bp_c_mixed_case.exp: Add intro comment. Test printing C
functions too. Test setting breakpoints and printing C functions
with no debug info too.
* gdb.ada/bp_c_mixed_case/qux.c: New file.
commit b5ec771e60c1a0863e51eb491c85c674097e9e13
Date: Wed Nov 8 14:22:32 2017 +0000
Subject: Introduce lookup_name_info and generalize Ada's FULL/WILD name matching
The gist of it is that linespec completion in Ada mode is generating
additional matches that should not appear in the match list
(internally generated symbols, or symbols that should be enclosed
between "<...>"). These extraneous entries have uppercase characters, such as:
These matches come from minimal symbols. The problem is that Ada
minsyms end up with no language set (language_auto), and thus we end
up using the generic symbol name matcher for those instead of Ada's.
We already had a special case for in compare_symbol_name to handle
this, but it was limited to expressions, while the case at hand is
completing a linespec. Fix this by applying the special case to
linespec completion as well. I.e., remove the EXPRESSION check from
compare_symbol_name. That alone turns out to not be sufficient still
-- GDB would still show a couple entries that shouldn't be there:
The reason is that these minimal symbols end up with their language
set to language_cplus / C++, because those encoded names manage to
demangle successfully as C++ symbols (using an old C++ mangling
scheme):
It's unfortunate that Ada's encoding scheme doesn't start with some
unique prefix like "_Z" in the C++ Itanium ABI mangling scheme. For
now, paper over that by treating C++ minsyms as Ada minsyms.
gdb/ChangeLog:
2018-01-10 Pedro Alves <palves@redhat.com>
PR gdb/22670
* ada-lang.c (ada_collect_symbol_completion_matches): If the
minsym's language is language_auto or language_cplus, pass down
language_ada instead.
* symtab.c (compare_symbol_name): Don't frob symbol language here.
gdb/testsuite/ChangeLog:
2018-01-10 Pedro Alves <palves@redhat.com>
PR gdb/22670
* gdb.ada/complete.exp ("complete break ada"): Replace kfail with
a fail.
Pedro Alves [Wed, 10 Jan 2018 23:14:06 +0000 (23:14 +0000)]
Fix gdb.ada/bp_c_mixed_case.exp (PR gdb/22670)
The problem here is that we are using the user-provided lookup name
literally for name comparisons. I.e., "<MixedCase>" with the "<>"s
included.
This commit fixes the minsym lookup case. psymbol/symbol lookup will
be fixed in a follow up.
In the minsym case, we're using using the user-provided lookup name
literally for linkage name comparisons. That obviously can't work
since the "<>" are not really part of the linkage name.
The original idea was that we'd use the symbol's language to select
the right symbol name matching algorithm, but that doesn't work for
Ada because it's not really possible to unambiguously tell from the
linkage name alone whether we're dealing with Ada symbols, so Ada
minsyms end up with no language set, or sometimes C++ set.
So fix this by treating Ada mode specially when determining the
linkage name to match against.
gdb/ChangeLog:
2018-01-10 Pedro Alves <palves@redhat.com>
PR gdb/22670
* minsyms.c (linkage_name_str): New function.
(iterate_over_minimal_symbols): Use it.
gdb/testsuite/ChangeLog:
2018-01-10 Pedro Alves <palves@redhat.com>
Samuel Thibault [Mon, 8 Jan 2018 15:30:16 +0000 (10:30 -0500)]
hurd: Add enough auxv support for AT_ENTRY for PIE binaries
Add PIE support for hurd, by faking an AT_ENTRY auxv entry. That value
is expected to be read by svr4_exec_displacement, which will propagate
the executable displacement.
gdb/ChangeLog:
* gdb/gnu-nat.c: Include <elf.h> and <link.h>.
(gnu_xfer_auxv): New function.
(gnu_xfer_partial): Call gnu_xfer_auxv when `object' is
TARGET_OBJECT_AUXV.
Yao Qi [Mon, 8 Jan 2018 10:09:33 +0000 (10:09 +0000)]
Fix GDBserver build failure when $development is false
When we set bfd/development.sh:$development to false, GDBserver failed to
build,
selftest.o: In function `selftests::run_tests(char const*)':
binutils-gdb/gdb/gdbserver/../common/selftest.c:97:undefined reference to `selftests::reset()'
collect2: error: ld returned 1 exit status
selftest.o shouldn't be compiled and linked when $development is false.
With this patch, in release mode, GDBserver doesn't nothing with option
--selftest,
$ ./gdbserver --selftest=foo
Selftests are not available in a non-development build.
$ ./gdbserver --selftest
Selftests are not available in a non-development build.
gdb/gdbserver:
2018-01-08 Yao Qi <yao.qi@linaro.org>
Simon Marchi <simon.marchi@ericsson.com>
* Makefile.in (OBS): Remove selftest.o.
* configure.ac: Set srv_selftest_objs if $development is true.
(GDBSERVER_DEPFILES): Append $srv_selftest_objs.
* configure: Re-generated.
* server.c (captured_main): Wrap variable selftest_filter with
GDB_SELF_TEST.
gdb/testsuite:
2018-01-08 Simon Marchi <simon.marchi@ericsson.com>
* gdb.server/unittest.exp: Match the output in non-development
mode.
Yao Qi [Mon, 8 Jan 2018 10:09:32 +0000 (10:09 +0000)]
Fix GDB build failure when $development is false
We don't build GDB selftests bits when $development is false. However, if
we turn bfd/development.sh:$development to false, common/selftest.c is
compiled which is not expected. It causes the build failure,
selftest.o: In function `selftests::run_tests(char const*)':
binutils-gdb/gdb/common/selftest.c:97: undefined reference to `selftests::reset()'
collect2: error: ld returned 1 exit status
I fix this issue by putting selftest.o selftest-arch.o into CONFIG_OBS
only when $development is true. After this is fixed, there are other
build failures in maint.c, this patch fixes them as well.
In the release mode, the output of these commands are:
(gdb) maintenance selftest
Selftests are not available in a non-development build.
(gdb) maintenance selftest foo
Selftests are not available in a non-development build.
(gdb) maintenance info selftests
Selftests are not available in a non-development build.
gdb:
2018-01-08 Yao Qi <yao.qi@linaro.org>
Simon Marchi <simon.marchi@ericsson.com>
* Makefile.in (COMMON_SFILES): Remove selftest-arch.c and
common/selftest.c.
(COMMON_OBS): Remove selftest.o.
* configure.ac: Append selftest-arch.c and common/selftest.c to
CONFIG_SRCS. Append selftest-arch.o and selftest.o to COMMON_OBS.
* configure: Re-generated.
* maint.c (maintenance_selftest): Wrap selftests::run_tests with
GDB_SELF_TEST.
(maintenance_info_selftests): Likewise.
gdb/testsuite:
2018-01-08 Simon Marchi <simon.marchi@ericsson.com>
* gdb.gdb/unittest.exp: Match output in non-development mode.