[binutils-gdb/users/hjl/linux/master] Don't propagate our current terminal state to the inferior

[H.J.Lu <hjl at sourceware dot org>]

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=6a06d66006d33293215eaf706ee416f6a99da273

commit 6a06d66006d33293215eaf706ee416f6a99da273
Author: Patrick Palka <patrick@parcs.ath.cx>
Date:   Sat Nov 22 14:12:49 2014 -0500

    Don't propagate our current terminal state to the inferior
    
    Currently when we start an inferior we have the inferior inherit our
    terminal state.  Under TUI, our terminal is highly modified by ncurses
    and readline.  So when starting an inferior under TUI, the inferior will
    have a highly modified terminal state which will interfere with standard
    I/O. For example,
    
    $ gdb gdb
    (gdb) break main
    (gdb) run
    (gdb) print puts ("a\nb")
    a
    b
    $1 = 4
    (gdb) [enter TUI mode]
    (gdb) run
    (gdb) [exit TUI mode]
    (gdb) print puts ("a\nb")
    a
     b
      $2 = 4
    (gdb) print puts ("a\r\nb\r")
    a
    b
    $3 = 6
    
    As you can see, when we start the inferior under the regular interface,
    puts() prints the text properly.  But when we start the inferior under
    TUI, puts() does not print the text properly.  This is because when we
    start the inferior under TUI it inherits our current terminal state
    which has been modified by ncurses to, among other things, require an
    explicit \r\n to print a new line. As a result the inferior performs
    standard I/O in an unexpected way.
    
    Because of this discrepancy, it doesn't seem like a good idea to have
    the inferior inherit our _current_ terminal state for it may have been
    modified by readline and/or ncurses.  Instead, we should have the
    inferior inherit a pristine snapshot of our terminal state taken before
    readline or ncurses have had a chance to alter it.  This enables the
    inferior to run in a more accurate way, more closely mimicking the
    program's behavior had it run standalone.  And it fixes the above
    mentioned issue.
    
    Tested on x86_64-unknown-linux-gnu.
    
    gdb/ChangeLog:
    
    	* terminal.h (set_initial_gdb_ttystate): Declare.
    	* inflow.c (initial_gdb_ttystate): New static variable.
    	(set_initial_gdb_ttystate): New setter.
    	(child_terminal_init_with_pgrp): Copy initial_gdb_ttystate
    	instead of our current terminal state.
    	* top.c (gdb_init): Call set_initial_gdb_ttystate.

Diff:
---
 gdb/ChangeLog  |  9 +++++++++
 gdb/inflow.c   | 13 ++++++++++++-
 gdb/terminal.h |  2 ++
 gdb/top.c      |  4 ++++
 4 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 0b63d34..6477dc1 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,12 @@
+2015-01-07  Patrick Palka  <patrick@parcs.ath.cx>
+
+	* terminal.h (set_initial_gdb_ttystate): Declare.
+	* inflow.c (initial_gdb_ttystate): New static variable.
+	(set_initial_gdb_ttystate): New setter.
+	(child_terminal_init_with_pgrp): Copy initial_gdb_ttystate
+	instead of our current terminal state.
+	* top.c (gdb_init): Call set_initial_gdb_ttystate.
+
 2015-01-07  Joel Brobecker  <brobecker@adacore.com>
 
 	* guile/scm-type.c (tyscm_array_1): Add comment.
diff --git a/gdb/inflow.c b/gdb/inflow.c
index 8947e63..3c121a3 100644
--- a/gdb/inflow.c
+++ b/gdb/inflow.c
@@ -79,6 +79,10 @@ struct terminal_info
    unimportant.  */
 static struct terminal_info our_terminal_info;
 
+/* The initial tty state given to each new inferior.  It is a snapshot of our
+   own tty state taken during initialization of GDB.  */
+static serial_ttystate initial_gdb_ttystate;
+
 static struct terminal_info *get_inflow_inferior_data (struct inferior *);
 
 #ifdef PROCESS_GROUP_TYPE
@@ -156,6 +160,13 @@ show_interactive_mode (struct ui_file *file, int from_tty,
     fprintf_filtered (file, "Debugger's interactive mode is %s.\n", value);
 }
 
+/* Set the initial tty state that is to be inherited by new inferiors.  */
+void
+set_initial_gdb_ttystate (void)
+{
+  initial_gdb_ttystate = serial_get_tty_state (stdin_serial);
+}
+
 /* Does GDB have a terminal (on stdin)?  */
 int
 gdb_has_a_terminal (void)
@@ -227,7 +238,7 @@ child_terminal_init_with_pgrp (int pgrp)
     {
       xfree (tinfo->ttystate);
       tinfo->ttystate = serial_copy_tty_state (stdin_serial,
-					       our_terminal_info.ttystate);
+					       initial_gdb_ttystate);
 
       /* Make sure that next time we call terminal_inferior (which will be
          before the program runs, as it needs to be), we install the new
diff --git a/gdb/terminal.h b/gdb/terminal.h
index ae25c98..17acde2 100644
--- a/gdb/terminal.h
+++ b/gdb/terminal.h
@@ -103,6 +103,8 @@ extern int gdb_has_a_terminal (void);
 
 extern void gdb_save_tty_state (void);
 
+extern void set_initial_gdb_ttystate (void);
+
 /* Set the process group of the caller to its own pid, or do nothing
    if we lack job control.  */
 extern int gdb_setpgid (void);
diff --git a/gdb/top.c b/gdb/top.c
index 524a92b..b85ea1a 100644
--- a/gdb/top.c
+++ b/gdb/top.c
@@ -1890,6 +1890,10 @@ gdb_init (char *argv0)
 
   initialize_stdin_serial ();
 
+  /* Take a snapshot of our tty state before readline/ncurses have had a chance
+     to alter it.  */
+  set_initial_gdb_ttystate ();
+
   async_init_signals ();
 
   /* We need a default language for parsing expressions, so simple