Fix "attach" command vs user input race

On async targets, a synchronous attach is done like this:

   #1 - target_attach is called (PTRACE_ATTACH is issued)
   #2 - a continuation is installed
   #3 - we go back to the event loop
   #4 - target reports stop (SIGSTOP), event loop wakes up, and
        attach continuation is called
   #5 - among other things, the continuation calls
        target_terminal_inferior, which removes stdin from the event
        loop

Note that in #3, GDB is still processing user input.  If the user is
fast enough, e.g., with something like:

  echo -e "attach PID\nset xxx=1" | gdb

... then the "set" command is processed before the attach completes.

We get worse behavior even, if input is a tty and therefore
readline/editing is enabled, with e.g.,:

 (gdb) attach PID\nset xxx=1

we then crash readline/gdb, with:

 Attaching to program: attach-wait-input, process 14537
 readline: readline_callback_read_char() called with no handler!
 Aborted
 $

Fix this by calling target_terminal_inferior before #3 above.

The test covers both scenarios by running with editing/readline forced
to both on and off.

gdb/
2014-07-09  Pedro Alves  <palves@redhat.com>

	* infcmd.c (attach_command_post_wait): Don't call
	target_terminal_inferior here.
	(attach_command): Call it here instead.

gdb/testsuite/
2014-07-09  Pedro Alves  <palves@redhat.com>

	* gdb.base/attach-wait-input.exp: New file.
	* gdb.base/attach-wait-input.c: New file.

gdb/ChangeLog

@@ -1,3 +1,9 @@
+2014-07-09  Pedro Alves  <palves@redhat.com>
+
+	* infcmd.c (attach_command_post_wait): Don't call
+	target_terminal_inferior here.
+	(attach_command): Call it here instead.
+
 2014-07-09  Andrew Burgess  <andrew.burgess@embecosm.com>
 
 	* ada-varobj.c (ada_varobj_ops): Fill in is_path_expr_parent

gdb/infcmd.c

@@ -2380,9 +2380,6 @@ attach_command_post_wait (char *args, int from_tty, int async_exec)
 
   post_create_inferior (&current_target, from_tty);
 
-  /* Install inferior's terminal modes.  */
-  target_terminal_inferior ();
-
   if (async_exec)
     {
       /* The user requested an `attach&', so be sure to leave threads
@@ -2498,6 +2495,22 @@ attach_command (char *args, int from_tty)
      based on what modes we are starting it with.  */
   target_terminal_init ();
 
+  /* Install inferior's terminal modes.  This may look like a no-op,
+     as we've just saved them above, however, this does more than
+     restore terminal settings:
+
+     - installs a SIGINT handler that forwards SIGINT to the inferior.
+       Otherwise a Ctrl-C pressed just while waiting for the initial
+       stop would end up as a spurious Quit.
+
+     - removes stdin from the event loop, which we need if attaching
+       in the foreground, otherwise on targets that report an initial
+       stop on attach (which are most) we'd process input/commands
+       while we're in the event loop waiting for that stop.  That is,
+       before the attach continuation runs and the command is really
+       finished.  */
+  target_terminal_inferior ();
+
   /* Set up execution context to know that we should return from
      wait_for_inferior as soon as the target reports a stop.  */
   init_wait_for_inferior ();

gdb/testsuite/ChangeLog

@@ -1,3 +1,8 @@
+2014-07-09  Pedro Alves  <palves@redhat.com>
+
+	* gdb.base/attach-wait-input.exp: New file.
+	* gdb.base/attach-wait-input.c: New file.
+
 2014-07-09  Andrew Burgess  <andrew.burgess@embecosm.com>
 
 	* gdb.mi/var-cmd.c (do_nested_struct_union_tests): New function

gdb/testsuite/gdb.base/attach-wait-input.c

@@ -0,0 +1,40 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+   Copyright 2014 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 <sys/types.h>
+#include <unistd.h>
+
+volatile int should_exit = 0;
+int mypid;
+
+static void
+setup_done (void)
+{
+}
+
+int
+main (void)
+{
+  unsigned int counter = 1;
+
+  mypid = getpid ();
+  setup_done ();
+
+  for (counter = 0; !should_exit && counter < 100; counter++)
+    sleep (1);
+  return 0;
+}

gdb/testsuite/gdb.base/attach-wait-input.exp

@@ -0,0 +1,119 @@
+# Copyright 2014 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/>.  */
+
+# Verify that GDB waits for the "attach" command to finish before
+# processing the following command.
+#
+# GDB used to have a race where on async targets, in the small window
+# between the attach request and the initial stop for the attach, GDB
+# was still processing user input.
+#
+# The issue was originally detected with:
+#
+#  echo -e "attach PID\nset xxx=1" | gdb
+#
+# In that scenario, stdin is not a tty, which disables readline.
+# Explicitly turning off editing exercises the same code path, and is
+# simpler to do, so we test with both editing on and off.
+
+# The test uses the "attach" command.
+if [target_info exists use_gdb_stub] {
+    return
+}
+
+standard_testfile
+
+if {[build_executable "failed to build" $testfile $srcfile debug]} {
+    return -1
+}
+
+# Start the program running, and return its PID, ready for attaching.
+
+proc start_program {binfile} {
+    global gdb_prompt
+    global decimal
+
+    clean_restart $binfile
+
+    if ![runto setup_done] then {
+	fail "Can't run to setup_done"
+	return 0
+    }
+
+    # Get the PID of the test process.
+    set testpid ""
+    set test "get inferior process ID"
+    gdb_test_multiple "p mypid" $test {
+	-re " = ($decimal)\r\n$gdb_prompt $" {
+	    set testpid $expect_out(1,string)
+	    pass $test
+	}
+    }
+
+    gdb_test "detach" "Detaching from program: .*"
+
+    if {$testpid == ""} {
+	return
+    }
+
+    return $testpid
+}
+
+# Do test proper.  EDITING indicates whether "set editing" is on or
+# off.
+
+proc test { editing } {
+    global gdb_prompt
+    global binfile
+    global decimal
+
+    with_test_prefix "editing $editing" {
+
+	set testpid [start_program $binfile]
+	if {$testpid == ""} {
+	    return
+	}
+
+	# Enable/disable readline.
+	gdb_test_no_output "set editing $editing"
+
+	# Send both commands at once.
+	send_gdb "attach $testpid\nprint should_exit = 1\n"
+
+	# Use gdb_expect directly instead of gdb_test_multiple to
+	# avoid races with the double prompt.
+	set test "attach and print"
+	gdb_expect {
+	    -re "Attaching to program.*process $testpid\r\n.*$gdb_prompt.*$decimal = 1\r\n$gdb_prompt $" {
+		pass "$test"
+	    }
+	    timeout {
+		fail "$test (timeout)"
+	    }
+	}
+
+	# As we've used attach, on quit, we'll detach from the
+	# program.  Explicitly kill it in case we failed above.
+	gdb_test "kill" \
+	    "" \
+	    "after attach, exit" \
+	    "Kill the program being debugged.*y or n. $" \
+	    "y"
+    }
+}
+
+foreach editing {"on" "off"} {
+    test $editing
+}