Unset attach_flag when running a new process
Commit Message
We currently set attach_flag when attaching to a process, so we should
make sure to unset it when forking a new process. Otherwise attach_flag
would remain set after forking, if the previous process associated with
the inferior was attached to.
[ I fixed the extended-gdbserver test failure by using
prepare_for_testing instead of using build_executable + gdb_start. The
former sets remote exec-file which is what makes the "run" command work as
expected under extended-gdbserver. ]
gdb/ChangeLog:
* gdb/infcmd.c (run_comand_1): Unset attach_flag.
gdb/testsuite/ChangeLog:
* gdb.base/run-after-attach.exp: New test file.
* gdb.base/run-after-attach.c: New test file.
---
gdb/infcmd.c | 4 ++
gdb/testsuite/gdb.base/run-after-attach.c | 25 +++++++++
gdb/testsuite/gdb.base/run-after-attach.exp | 80 +++++++++++++++++++++++++++++
3 files changed, 109 insertions(+)
create mode 100644 gdb/testsuite/gdb.base/run-after-attach.c
create mode 100644 gdb/testsuite/gdb.base/run-after-attach.exp
Comments
On 30/07/15 16:28, Patrick Palka wrote:
> + /* Unset attach_flag, it may be set if the previous process associated with
> + the inferior was attached to. */
> + current_inferior ()->attach_flag = 0;
It is better to do such reset in target.c:target_pre_inferior, which
is created for such purpose, as far as I know.
I don't read the test part of this patch.
On Thu, Jul 30, 2015 at 12:24 PM, Yao Qi <qiyaoltc@gmail.com> wrote:
> On 30/07/15 16:28, Patrick Palka wrote:
>>
>> + /* Unset attach_flag, it may be set if the previous process associated
>> with
>> + the inferior was attached to. */
>> + current_inferior ()->attach_flag = 0;
>
>
> It is better to do such reset in target.c:target_pre_inferior, which
> is created for such purpose, as far as I know.
That makes sense, although I am a bit hesitant about such a change,
since target_pre_inferior has a lot of callers. Is it OK, for
instance, to assume that current_inferior points to the right inferior
when target_pre_inferior gets called? Currently current_inferior is
not called directly or indirectly from target_pre_inferior.
I wonder how this attach_flag issues manifests itself elsewhere, if at all.
On Thu, Jul 30, 2015 at 11:28 AM, Patrick Palka <patrick@parcs.ath.cx> wrote:
> We currently set attach_flag when attaching to a process, so we should
> make sure to unset it when forking a new process. Otherwise attach_flag
> would remain set after forking, if the previous process associated with
> the inferior was attached to.
>
> [ I fixed the extended-gdbserver test failure by using
> prepare_for_testing instead of using build_executable + gdb_start. The
> former sets remote exec-file which is what makes the "run" command work as
> expected under extended-gdbserver. ]
>
> gdb/ChangeLog:
>
> * gdb/infcmd.c (run_comand_1): Unset attach_flag.
>
> gdb/testsuite/ChangeLog:
>
> * gdb.base/run-after-attach.exp: New test file.
> * gdb.base/run-after-attach.c: New test file.
Ping.
On 08/02/2015 09:20 PM, Patrick Palka wrote:
> On Thu, Jul 30, 2015 at 12:24 PM, Yao Qi <qiyaoltc@gmail.com> wrote:
>> On 30/07/15 16:28, Patrick Palka wrote:
>>>
>>> + /* Unset attach_flag, it may be set if the previous process associated
>>> with
>>> + the inferior was attached to. */
>>> + current_inferior ()->attach_flag = 0;
>>
>>
>> It is better to do such reset in target.c:target_pre_inferior, which
>> is created for such purpose, as far as I know.
>
> That makes sense, although I am a bit hesitant about such a change,
> since target_pre_inferior has a lot of callers.
You mean, the whole lot of 3 callers. :-)
infcmd.c:542: target_pre_inferior (from_tty);
infcmd.c:2603: target_pre_inferior (from_tty);
target.c:2191: target_pre_inferior (from_tty);
> Is it OK, for
> instance, to assume that current_inferior points to the right inferior
> when target_pre_inferior gets called? Currently current_inferior is
> not called directly or indirectly from target_pre_inferior.
Should be. It's called from the run command, and from the
attach command. Those assume that the current inferior is
the inferior that we're about to run/attach.
The call in target_preopen is more dubious. That should probably
be iterating over all inferiors and calling target_pre_inferior
on each. For the case where you have multiple inferiors, and then
connect to a remote target with "target remote".
> +set test "kill process"
> +send_gdb "kill\n"
> +gdb_expect {
> + -re ".y or n. $" {
> + send_gdb "y\n"
> + exp_continue
> + }
> + -re "$gdb_prompt $" {
> + pass $test
> + }
> + default {
> + fail $test
> + }
> +}
Avoid raw gdb_expect. You can write:
gdb_test "kill" \
"" \
"kill process" \
"Kill the program being debugged.*y or n. $" \
"y"
> +
> +set test "restart process"
> +gdb_test "start" "Starting program.*Temporary breakpoint .*" $test
> +
> +set test "attempt kill via quit"
> +send_gdb "quit\n"
> +set ok 0
> +# The quit prompt should warn about killing the process, not about detaching the
> +# process, since this process was not attached to.
> +gdb_expect {
Please use gdb_test_multiple. That catches internal errors and such.
> + -re "will be killed.*.y or n. $" {
> + set ok 1
> + send_gdb "n\n"
> + exp_continue
> + }
> + -re "will be detached.*.y or n. $" {
> + send_gdb "n\n"
> + exp_continue
> + }
> + -re "$gdb_prompt $" {
> + if $ok {
> + pass $test
> + } else {
> + fail $test
> + }
gdb_assert $ok $test
> + }
> + default {
> + fail $test
> + }
> +}
> +
> +remote_exec host "kill $testpid"
>
Thanks,
Pedro Alves
On Thu, Aug 13, 2015 at 12:30 PM, Pedro Alves <palves@redhat.com> wrote:
> On 08/02/2015 09:20 PM, Patrick Palka wrote:
>> On Thu, Jul 30, 2015 at 12:24 PM, Yao Qi <qiyaoltc@gmail.com> wrote:
>>> On 30/07/15 16:28, Patrick Palka wrote:
>>>>
>>>> + /* Unset attach_flag, it may be set if the previous process associated
>>>> with
>>>> + the inferior was attached to. */
>>>> + current_inferior ()->attach_flag = 0;
>>>
>>>
>>> It is better to do such reset in target.c:target_pre_inferior, which
>>> is created for such purpose, as far as I know.
>>
>> That makes sense, although I am a bit hesitant about such a change,
>> since target_pre_inferior has a lot of callers.
>
> You mean, the whole lot of 3 callers. :-)
>
> infcmd.c:542: target_pre_inferior (from_tty);
> infcmd.c:2603: target_pre_inferior (from_tty);
> target.c:2191: target_pre_inferior (from_tty);
The thing is that one of the callers, target_preopen, itself has lots
of callers all over the place.
>
>> Is it OK, for
>> instance, to assume that current_inferior points to the right inferior
>> when target_pre_inferior gets called? Currently current_inferior is
>> not called directly or indirectly from target_pre_inferior.
>
> Should be. It's called from the run command, and from the
> attach command. Those assume that the current inferior is
> the inferior that we're about to run/attach.
Yeah, those are obviously fine.
>
> The call in target_preopen is more dubious. That should probably
> be iterating over all inferiors and calling target_pre_inferior
> on each. For the case where you have multiple inferiors, and then
> connect to a remote target with "target remote".
I see. I will move the call to target_pre_inferior then.
>
>
>
>> +set test "kill process"
>> +send_gdb "kill\n"
>> +gdb_expect {
>> + -re ".y or n. $" {
>> + send_gdb "y\n"
>> + exp_continue
>> + }
>> + -re "$gdb_prompt $" {
>> + pass $test
>> + }
>> + default {
>> + fail $test
>> + }
>> +}
>
> Avoid raw gdb_expect. You can write:
>
> gdb_test "kill" \
> "" \
> "kill process" \
> "Kill the program being debugged.*y or n. $" \
> "y"
Cool, didn't know about that.
>
>> +
>> +set test "restart process"
>> +gdb_test "start" "Starting program.*Temporary breakpoint .*" $test
>> +
>> +set test "attempt kill via quit"
>> +send_gdb "quit\n"
>> +set ok 0
>> +# The quit prompt should warn about killing the process, not about detaching the
>> +# process, since this process was not attached to.
>> +gdb_expect {
>
> Please use gdb_test_multiple. That catches internal errors and such.
Hmm, I wonder why I didn't use gdb_test_multiple in the first place.
Probably for no good reason. I'll make this change then.
>
>> + -re "will be killed.*.y or n. $" {
>> + set ok 1
>> + send_gdb "n\n"
>> + exp_continue
>> + }
>> + -re "will be detached.*.y or n. $" {
>> + send_gdb "n\n"
>> + exp_continue
>> + }
>> + -re "$gdb_prompt $" {
>> + if $ok {
>> + pass $test
>> + } else {
>> + fail $test
>> + }
>
> gdb_assert $ok $test
Done.
>
>
>> + }
>> + default {
>> + fail $test
>> + }
>> +}
>> +
>> +remote_exec host "kill $testpid"
>>
>
> Thanks,
> Pedro Alves
>
@@ -601,6 +601,10 @@ run_command_1 (char *args, int from_tty, int tbreak_at_main)
/* Done with ARGS. */
do_cleanups (args_chain);
+ /* Unset attach_flag, it may be set if the previous process associated with
+ the inferior was attached to. */
+ current_inferior ()->attach_flag = 0;
+
/* We call get_inferior_args() because we might need to compute
the value now. */
run_target->to_create_inferior (run_target, exec_file, get_inferior_args (),
new file mode 100644
@@ -0,0 +1,25 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2015 Free Software Foundation, Inc.
+
+ This program is free software; you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <unistd.h>
+
+int
+main (void)
+{
+ sleep (600);
+ return 0;
+}
new file mode 100644
@@ -0,0 +1,80 @@
+# Copyright (C) 2015 Free Software Foundation, Inc.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+# Check that forking a process after a previous process was attached to unsets
+# attach_flag. This is done indirectly by inspecting GDB's quit prompt.
+
+if ![can_spawn_for_attach] {
+ return 0
+}
+
+standard_testfile
+set executable $testfile
+
+if [prepare_for_testing $testfile.exp $executable] {
+ return -1
+}
+
+set testpid [spawn_wait_for_attach $binfile]
+
+set test "attach to process"
+gdb_test "attach $testpid" "Attaching to program.*" $test
+
+set test "kill process"
+send_gdb "kill\n"
+gdb_expect {
+ -re ".y or n. $" {
+ send_gdb "y\n"
+ exp_continue
+ }
+ -re "$gdb_prompt $" {
+ pass $test
+ }
+ default {
+ fail $test
+ }
+}
+
+set test "restart process"
+gdb_test "start" "Starting program.*Temporary breakpoint .*" $test
+
+set test "attempt kill via quit"
+send_gdb "quit\n"
+set ok 0
+# The quit prompt should warn about killing the process, not about detaching the
+# process, since this process was not attached to.
+gdb_expect {
+ -re "will be killed.*.y or n. $" {
+ set ok 1
+ send_gdb "n\n"
+ exp_continue
+ }
+ -re "will be detached.*.y or n. $" {
+ send_gdb "n\n"
+ exp_continue
+ }
+ -re "$gdb_prompt $" {
+ if $ok {
+ pass $test
+ } else {
+ fail $test
+ }
+ }
+ default {
+ fail $test
+ }
+}
+
+remote_exec host "kill $testpid"