bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Scott Bell]

Under Mac OS X 10.6.2, Emacs 23.1.90.1, in an M-x term buffer
running /bin/bash, I get the following undesired behavior in
Cocoa emacs (Emacs.app/Contents/MacOS/Emacs -q) when typing 
`one two M-DEL' where meta is typed as the alt/option key:

$ one twoÿ

This occurs even when ns-alternate-modifier is set to 'meta.
Running Emacs in a terminal window (Emacs.app/Contents/MacOS/
Emacs -q -nw) works as I expect:

$ one 

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Ken Hori]

[-- Attachment #1: Type: text/plain, Size: 961 bytes --]

It happens on Linux as well
(For example, Emacs 23.1.91 bzr revision 99310)

So M-d (delete-next-word) on term-mode does not work, even though term-mode
is supposed to be emulating a terminal as it is.

I can confirm this bug has existed for at least 1 month.
Perhaps reporting it again with a more noticeable title might get more
attention that it deserves. This is a major bug that keeps me from using M-x
term.


On Wed, Jan 6, 2010 at 3:35 PM, Scott Bell <sctb@me.com> wrote:

> Under Mac OS X 10.6.2, Emacs 23.1.90.1, in an M-x term buffer
> running /bin/bash, I get the following undesired behavior in
> Cocoa emacs (Emacs.app/Contents/MacOS/Emacs -q) when typing
> `one two M-DEL' where meta is typed as the alt/option key:
>
> $ one twoÿ
>
> This occurs even when ns-alternate-modifier is set to 'meta.
> Running Emacs in a terminal window (Emacs.app/Contents/MacOS/
> Emacs -q -nw) works as I expect:
>
> $ one
>
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 1265 bytes --]

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Ken Hori]

[-- Attachment #1: Type: text/plain, Size: 959 bytes --]

It happens on Linux as well
(For example, Emacs 23.1.91 bzr revision 99310)

So M-d (delete-next-word) on term-mode does not work, even though term-mode
is supposed to be emulating a terminal as it is.

I can confirm this bug has existed for at least 1 month.
Perhaps reporting it again with a more noticeable title might get more
attention that it deserves. This is a major bug that keeps me from using M-x
term.

On Wed, Jan 6, 2010 at 3:35 PM, Scott Bell <sctb@me.com> wrote:

> Under Mac OS X 10.6.2, Emacs 23.1.90.1, in an M-x term buffer
> running /bin/bash, I get the following undesired behavior in
> Cocoa emacs (Emacs.app/Contents/MacOS/Emacs -q) when typing
> `one two M-DEL' where meta is typed as the alt/option key:
>
> $ one twoÿ
>
> This occurs even when ns-alternate-modifier is set to 'meta.
> Running Emacs in a terminal window (Emacs.app/Contents/MacOS/
> Emacs -q -nw) works as I expect:
>
> $ one
>
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 1337 bytes --]

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Ken Hori]

[-- Attachment #1: Type: text/plain, Size: 1357 bytes --]

I identified the cause of this bug.

It happens when PS1 environment variable contains certain \e escape code
like '\[\e[01;32m\][\h:\w]\$\[\e[00m\] '.

In such a case, term.el gets messed up somehow and becomes
no longer able to handle meta (alt) keyboard input properly.


On Sun, Jan 17, 2010 at 6:09 AM, Ken Hori <fplemma@gmail.com> wrote:

> It happens on Linux as well
> (For example, Emacs 23.1.91 bzr revision 99310)
>
> So M-d (delete-next-word) on term-mode does not work, even though term-mode
> is supposed to be emulating a terminal as it is.
>
> I can confirm this bug has existed for at least 1 month.
> Perhaps reporting it again with a more noticeable title might get more
> attention that it deserves. This is a major bug that keeps me from using M-x
> term.
>
> On Wed, Jan 6, 2010 at 3:35 PM, Scott Bell <sctb@me.com> wrote:
>
>> Under Mac OS X 10.6.2, Emacs 23.1.90.1, in an M-x term buffer
>> running /bin/bash, I get the following undesired behavior in
>> Cocoa emacs (Emacs.app/Contents/MacOS/Emacs -q) when typing
>> `one two M-DEL' where meta is typed as the alt/option key:
>>
>> $ one twoÿ
>>
>> This occurs even when ns-alternate-modifier is set to 'meta.
>> Running Emacs in a terminal window (Emacs.app/Contents/MacOS/
>> Emacs -q -nw) works as I expect:
>>
>> $ one
>>
>>
>>
>>
>>
>

[-- Attachment #2: Type: text/html, Size: 2045 bytes --]

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Scott Bell]

On 2010-01-17, at 7:18 AM, Ken Hori wrote:

> I identified the cause of this bug.
> 
> It happens when PS1 environment variable contains certain \e escape code
> like '\[\e[01;32m\][\h:\w]\$\[\e[00m\] '.

I can't reproduce this -- I still encounter the problem
when PS1='> ' (no escape codes).

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Ken Hori]

[-- Attachment #1: Type: text/plain, Size: 1857 bytes --]

Please ignore my last post.

You can instead apply the following patch for a quick fix:

+++ term.el    2010-01-18 09:27:33.085376730 -0800
@@ -1208,8 +1208,17 @@

 (defun term-send-raw-meta ()
   (interactive)
-  (let* ((keys (this-command-keys))
-         (char (aref keys (1- (length keys)))))
+  (let ((char last-input-event))
+    (when (symbolp last-input-event)
+      ;; Convert `return' to C-m, etc.
+      (let ((tmp (get char 'event-symbol-elements)))
+    (when tmp
+      (setq char (car tmp)))
+    (when (symbolp char)
+      (setq tmp (get char 'ascii-character))
+      (when tmp
+        (setq char tmp)))))
+    (setq char (event-basic-type char))
     (term-send-raw-string (if (and (numberp char)
                    (> char 127)
                    (< char 256))

Or, just replace your term-send-raw-meta with:

(defun term-send-raw-meta ()
  (interactive)
  (let ((char last-input-event))
    (when (symbolp last-input-event)
      ;; Convert `return' to C-m, etc.
      (let ((tmp (get char 'event-symbol-elements)))
    (when tmp
      (setq char (car tmp)))
    (when (symbolp char)
      (setq tmp (get char 'ascii-character))
      (when tmp
        (setq char tmp)))))
    (setq char (event-basic-type char))
    (term-send-raw-string (if (and (numberp char)
                   (> char 127)
                   (< char 256))
                  (make-string 1 char)
                (format "\e%c" char)))))

Let me know if it succeeds or not.

On Sun, Jan 17, 2010 at 3:46 PM, Scott Bell <sctb@me.com> wrote:

> On 2010-01-17, at 7:18 AM, Ken Hori wrote:
>
> > I identified the cause of this bug.
> >
> > It happens when PS1 environment variable contains certain \e escape code
> > like '\[\e[01;32m\][\h:\w]\$\[\e[00m\] '.
>
> I can't reproduce this -- I still encounter the problem
> when PS1='> ' (no escape codes).
>
>
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 2414 bytes --]

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Scott Bell]

On 2010-01-18, at 2:17 PM, Ken Hori wrote:

> Please ignore my last post.
> 
> You can instead apply the following patch for a quick fix:
> 
> +++ term.el    2010-01-18 09:27:33.085376730 -0800
> @@ -1208,8 +1208,17 @@
>  
>  (defun term-send-raw-meta ()
>    (interactive)
> -  (let* ((keys (this-command-keys))
> -         (char (aref keys (1- (length keys)))))
> +  (let ((char last-input-event))
> +    (when (symbolp last-input-event)
> +      ;; Convert `return' to C-m, etc.
> +      (let ((tmp (get char 'event-symbol-elements)))
> +    (when tmp
> +      (setq char (car tmp)))
> +    (when (symbolp char)
> +      (setq tmp (get char 'ascii-character))
> +      (when tmp
> +        (setq char tmp)))))
> +    (setq char (event-basic-type char))
>      (term-send-raw-string (if (and (numberp char)
>                     (> char 127)
>                     (< char 256))
> 
> Or, just replace your term-send-raw-meta with:
> 
> (defun term-send-raw-meta ()
>   (interactive)
>   (let ((char last-input-event))
>     (when (symbolp last-input-event)
>       ;; Convert `return' to C-m, etc.
>       (let ((tmp (get char 'event-symbol-elements)))
>     (when tmp
>       (setq char (car tmp)))
>     (when (symbolp char)
>       (setq tmp (get char 'ascii-character))
>       (when tmp
>         (setq char tmp)))))
>     (setq char (event-basic-type char))
>     (term-send-raw-string (if (and (numberp char)
>                    (> char 127)
>                    (< char 256))
>                   (make-string 1 char)
>                 (format "\e%c" char)))))
> 
> Let me know if it succeeds or not.

This works, at least in my common use cases (M-d, M-RET, etc.)

- Scott

bug#5330: M-d broken in term.el [Was: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER]

[Chong Yidong]

> It happens on Linux as well
> (For example, Emacs 23.1.91 bzr revision 99310)
>
> So M-d (delete-next-word) on term-mode does not work, even though term-mode
> is supposed to be emulating a terminal as it is.

Hi Stefan,

This bug started with your 2009-11-13 change to term.el:

2009-11-13  Stefan Monnier  <monnier@iro.umontreal.ca>

  * term.el (make-term, term-exec, term-sentinel, term-read-input-ring)
  (term-write-input-ring, term-check-source, term-start-output-log):
  (term-display-buffer-line, term-dynamic-list-completions):
  (term-ansi-make-term, serial-term):
  ...
  Use with-current-buffer.

Could you please fix it?  I'm not sure Ken Hori's patch is right,
because it affects for another part of term.el.  Better to find out how
your change broke term.el first.

Thanks.

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Chong Yidong]

I narrowed the problem down to this specific change:

2009-12-04  Stefan Monnier  <monnier@iro.umontreal.ca>

   Minor cleanup.
   * term.el (term-send-raw, term-send-raw-meta): Use read-key-sequence's
   key decoding rather than do it manually via last-input-event +
   ascii-character.

I'm not sure if the change to term-send-raw breaks anything, but the the
change to term-send-raw-meta definitely prevents the process from
getting M-d correctly.  I've reverted it in the repository for now
(change attached), but you might want to double-check the entire change.


*** lisp/term.el	2010-01-13 08:35:10 +0000
--- lisp/term.el	2010-01-20 17:12:01 +0000
***************
*** 1208,1215 ****
  
  (defun term-send-raw-meta ()
    (interactive)
!   (let* ((keys (this-command-keys))
!          (char (aref keys (1- (length keys)))))
      (term-send-raw-string (if (and (numberp char)
  				   (> char 127)
  				   (< char 256))
--- 1208,1224 ----
  
  (defun term-send-raw-meta ()
    (interactive)
!   (let ((char last-input-event))
!     (when (symbolp last-input-event)
!       ;; Convert `return' to C-m, etc.
!       (let ((tmp (get char 'event-symbol-elements)))
! 	(when tmp
! 	  (setq char (car tmp)))
! 	(when (symbolp char)
! 	  (setq tmp (get char 'ascii-character))
! 	  (when tmp
! 	    (setq char tmp)))))
!     (setq char (event-basic-type char))
      (term-send-raw-string (if (and (numberp char)
  				   (> char 127)
  				   (< char 256))

bug#5330: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Scott Bell]

On 2010-01-20, at 10:18 AM, Chong Yidong wrote:

> I narrowed the problem down to this specific change:
> 
> 2009-12-04  Stefan Monnier  <monnier@iro.umontreal.ca>
> 
>   Minor cleanup.
>   * term.el (term-send-raw, term-send-raw-meta): Use read-key-sequence's
>   key decoding rather than do it manually via last-input-event +
>   ascii-character.
> 
> I'm not sure if the change to term-send-raw breaks anything, but the the
> change to term-send-raw-meta definitely prevents the process from
> getting M-d correctly.  I've reverted it in the repository for now
> (change attached), but you might want to double-check the entire change.

The latest repository seems to work for me with
this change, thanks!

- Scott

bug#5449: Term mode in Cocoa does not respect NS-ALTERNATE-MODIFIER

[Lars Magne Ingebrigtsen]

Scott Bell <sctb@me.com> writes:

>> I'm not sure if the change to term-send-raw breaks anything, but the the
>> change to term-send-raw-meta definitely prevents the process from
>> getting M-d correctly.  I've reverted it in the repository for now
>> (change attached), but you might want to double-check the entire change.
>
> The latest repository seems to work for me with
> this change, thanks!

Reading the thread, it seems like the bug was fixed, so I'm closing this
report.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/