[PATCH] Ctrl-D

[PATCH] Ctrl-D

[Andrew STUBBS]

[-- Attachment #1: Type: text/plain, Size: 720 bytes --]

Hi,

In response to the recent discussion of Ctrl-D behaviour on the GDB 
list, here is a patch which attempts to address the issues identified.

This example shows two presses of Crtl-D:

Old behaviour:

(gdb) <C-d>The program is running.  Exit anyway? (y or n) <C-d>bash$

New behaviour:

(gdb) <C-d>quit
The program is running.  Exit anyway? (y or n) <C-d>Y
bash$


In addition to these cosmetic changes, in have adjusted the code such 
that hook-quit will work, as will trace-commands:

(gdb) define hook-quit
Type commands for definition of "hook-quit".
End with a line saying just "end".
 >set confirm 0
 >end
(gdb) set trace-commands on
(gdb) <C-d>quit
+quit
++set confirm 0
bash$


:ADDPATCH CLI:

Andrew Stubbs

[-- Attachment #2: ctrl-d.patch --]
[-- Type: text/plain, Size: 1290 bytes --]

2006-11-30  Andrew Stubbs  <andrew.stubbs@st.com>

	* event-top.c (command_handler): On EOF, print 'quit' and run quit
	command via execute_command such that hooks and trace work.
	* utils.c (defaulted_query): On EOF, print default answer and newline.

Index: src/gdb/event-top.c
===================================================================
--- src.orig/gdb/event-top.c	2006-07-21 15:46:53.000000000 +0100
+++ src/gdb/event-top.c	2006-11-30 12:51:46.000000000 +0000
@@ -501,7 +501,10 @@ command_handler (char *command)
      but GDB is still alive. In such a case, we just quit gdb
      killing the inferior program too. */
   if (command == 0)
-    quit_command ((char *) 0, stdin == instream);
+    {
+      printf_unfiltered ("quit\n");
+      execute_command ("quit", stdin == instream);
+    }
 
   time_at_cmd_start = get_run_time ();
 
Index: src/gdb/utils.c
===================================================================
--- src.orig/gdb/utils.c	2006-11-30 12:23:34.000000000 +0000
+++ src/gdb/utils.c	2006-11-30 12:24:09.000000000 +0000
@@ -1230,6 +1230,7 @@ defaulted_query (const char *ctlstr, con
       clearerr (stdin);		/* in case of C-d */
       if (answer == EOF)	/* C-d */
 	{
+	  printf_filtered ("%c\n", def_answer);
 	  retval = def_value;
 	  break;
 	}

Re: [PATCH] Ctrl-D

[Andrew STUBBS]

Any chance of getting this nice-to-have into 6.6? I'm away from Friday 
onwards.

Andrew STUBBS wrote:
> Hi,
> 
> In response to the recent discussion of Ctrl-D behaviour on the GDB 
> list, here is a patch which attempts to address the issues identified.
> 
> This example shows two presses of Crtl-D:
> 
> Old behaviour:
> 
> (gdb) <C-d>The program is running.  Exit anyway? (y or n) <C-d>bash$
> 
> New behaviour:
> 
> (gdb) <C-d>quit
> The program is running.  Exit anyway? (y or n) <C-d>Y
> bash$
> 
> 
> In addition to these cosmetic changes, in have adjusted the code such 
> that hook-quit will work, as will trace-commands:
> 
> (gdb) define hook-quit
> Type commands for definition of "hook-quit".
> End with a line saying just "end".
>  >set confirm 0
>  >end
> (gdb) set trace-commands on
> (gdb) <C-d>quit
> +quit
> ++set confirm 0
> bash$
> 
> 
> :ADDPATCH CLI:
> 
> Andrew Stubbs
> 
> 
> ------------------------------------------------------------------------
> 
> 2006-11-30  Andrew Stubbs  <andrew.stubbs@st.com>
> 
> 	* event-top.c (command_handler): On EOF, print 'quit' and run quit
> 	command via execute_command such that hooks and trace work.
> 	* utils.c (defaulted_query): On EOF, print default answer and newline.
> 
> Index: src/gdb/event-top.c
> ===================================================================
> --- src.orig/gdb/event-top.c	2006-07-21 15:46:53.000000000 +0100
> +++ src/gdb/event-top.c	2006-11-30 12:51:46.000000000 +0000
> @@ -501,7 +501,10 @@ command_handler (char *command)
>       but GDB is still alive. In such a case, we just quit gdb
>       killing the inferior program too. */
>    if (command == 0)
> -    quit_command ((char *) 0, stdin == instream);
> +    {
> +      printf_unfiltered ("quit\n");
> +      execute_command ("quit", stdin == instream);
> +    }
>  
>    time_at_cmd_start = get_run_time ();
>  
> Index: src/gdb/utils.c
> ===================================================================
> --- src.orig/gdb/utils.c	2006-11-30 12:23:34.000000000 +0000
> +++ src/gdb/utils.c	2006-11-30 12:24:09.000000000 +0000
> @@ -1230,6 +1230,7 @@ defaulted_query (const char *ctlstr, con
>        clearerr (stdin);		/* in case of C-d */
>        if (answer == EOF)	/* C-d */
>  	{
> +	  printf_filtered ("%c\n", def_answer);
>  	  retval = def_value;
>  	  break;
>  	}

Re: [PATCH] Ctrl-D

[Joel Brobecker]

Hi Andrew,

> Any chance of getting this nice-to-have into 6.6? I'm away from Friday 
> onwards.

My rationale is that this patch is only a nice-to-have. Also, it is
not completely obvious for me either (the effect of changing the call
to quit_command into a call to execute_command needs to be looked at).

-- 
Joel

Re: [PATCH] Ctrl-D

[Daniel Jacobowitz]

On Thu, Nov 30, 2006 at 01:10:09PM +0000, Andrew STUBBS wrote:
> New behaviour:
> 
> (gdb) <C-d>quit
> The program is running.  Exit anyway? (y or n) <C-d>Y
> bash$

How'd you feel about "EOF [assumed %c]" or something like that?  I'd
look at that and think I typed the Y.

> 2006-11-30  Andrew Stubbs  <andrew.stubbs@st.com>
> 
> 	* event-top.c (command_handler): On EOF, print 'quit' and run quit
> 	command via execute_command such that hooks and trace work.
> 	* utils.c (defaulted_query): On EOF, print default answer and newline.

OK for HEAD with or without that change, as you prefer.  I'd recommend
this for 6.6 now that it's in HEAD; Joel, is that OK?

:REVIEWMAIL:

-- 
Daniel Jacobowitz
CodeSourcery

Re: [PATCH] Ctrl-D

[Joel Brobecker]

> > New behaviour:
> > 
> > (gdb) <C-d>quit
> > The program is running.  Exit anyway? (y or n) <C-d>Y
> > bash$
> 
> How'd you feel about "EOF [assumed %c]" or something like that?  I'd
> look at that and think I typed the Y.

I really like Daniel's suggestion.

> > 2006-11-30  Andrew Stubbs  <andrew.stubbs@st.com>
> > 
> > 	* event-top.c (command_handler): On EOF, print 'quit' and run quit
> > 	command via execute_command such that hooks and trace work.
> > 	* utils.c (defaulted_query): On EOF, print default answer and newline.
> 
> OK for HEAD with or without that change, as you prefer.  I'd recommend
> this for 6.6 now that it's in HEAD; Joel, is that OK?

Sure. I was a bit nervious about any unexpected side-effect of the change
in event-top.c (changing the call to quit_command() into a call to
execute_command ("quit")) But if you think it's low risk, then I don't
have any problem.

-- 
Joel

Re: [PATCH] Ctrl-D

[Andrew STUBBS]

[-- Attachment #1: Type: text/plain, Size: 262 bytes --]

Daniel Jacobowitz wrote:
> OK for HEAD with or without that change, as you prefer.  I'd recommend
> this for 6.6 now that it's in HEAD; Joel, is that OK?

Committed to both HEAD and gdb_6_6-branch with the suggested change.

The final patch is attached.

Andrew

[-- Attachment #2: ctrl-d.patch --]
[-- Type: text/plain, Size: 1304 bytes --]

2006-12-06  Andrew Stubbs  <andrew.stubbs@st.com>

	* event-top.c (command_handler): On EOF, print 'quit' and run quit
	command via execute_command such that hooks and trace work.
	* utils.c (defaulted_query): On EOF, print default answer and newline.

Index: src/gdb/event-top.c
===================================================================
--- src.orig/gdb/event-top.c	2006-12-06 10:49:25.000000000 +0000
+++ src/gdb/event-top.c	2006-12-06 13:52:46.000000000 +0000
@@ -501,7 +501,10 @@ command_handler (char *command)
      but GDB is still alive. In such a case, we just quit gdb
      killing the inferior program too. */
   if (command == 0)
-    quit_command ((char *) 0, stdin == instream);
+    {
+      printf_unfiltered ("quit\n");
+      execute_command ("quit", stdin == instream);
+    }
 
   time_at_cmd_start = get_run_time ();
 
Index: src/gdb/utils.c
===================================================================
--- src.orig/gdb/utils.c	2006-12-06 10:49:26.000000000 +0000
+++ src/gdb/utils.c	2006-12-06 13:54:01.000000000 +0000
@@ -1230,6 +1230,7 @@ defaulted_query (const char *ctlstr, con
       clearerr (stdin);		/* in case of C-d */
       if (answer == EOF)	/* C-d */
 	{
+	  printf_filtered ("EOF [assumed %c]\n", def_answer);
 	  retval = def_value;
 	  break;
 	}