master da02b9edad: Fix coding systems used for X input methods

[Po Lu]

branch: master
commit da02b9edadbc809b25ac83eccf64089f1cf3b160
Author: Po Lu <luangruo@yahoo.com>
Commit: Po Lu <luangruo@yahoo.com>

    Fix coding systems used for X input methods
    
    * doc/emacs/mule.texi (International): Refer to X Coding as
    well.
    (Communication Coding): Document that locale-coding-system is
    not always used on X to decode keyboard input.
    (X Coding): New node.
    * etc/NEWS: Announce change to input method coding resolution.
    * lisp/term/x-win.el (x-get-input-coding-system): New function.
    * src/coding.c (syms_of_coding): Update doc string of
    locale-coding-system.
    * src/xfns.c (struct x_xim_text_conversion_data)
    (x_xim_text_to_utf8_unix_1, x_xim_text_to_utf8_unix_2)
    (x_xim_text_to_utf8_unix): Accept dpyinfo.  Use the coding
    system specified inside if possible.
    (xic_preedit_draw_callback): Pass dpyinfo.
    * src/xterm.c (handle_one_xevent): Use XIM coding system for IM
    input.
    (xim_open_dpy): Try to determine the input method coding system.
    (mark_xterm): Mark `xim_coding'.
    (syms_of_xterm): New variable `x-input-coding-system'.
    * src/xterm.h (struct x_display_info): New field `xim_coding'.
    (FRAME_X_XIM_CODING): New macro.
---
 doc/emacs/mule.texi | 40 ++++++++++++++++++++++++++++++----------
 etc/NEWS            | 13 +++++++++++++
 lisp/term/x-win.el  | 22 ++++++++++++++++++++++
 src/coding.c        |  6 +++---
 src/xfns.c          | 33 +++++++++++++++++++++++----------
 src/xterm.c         | 34 ++++++++++++++++++++++++++++++----
 src/xterm.h         |  9 +++++++++
 7 files changed, 130 insertions(+), 27 deletions(-)

diff --git a/doc/emacs/mule.texi b/doc/emacs/mule.texi
index 5f30341838..1bbd7440f3 100644
--- a/doc/emacs/mule.texi
+++ b/doc/emacs/mule.texi
@@ -61,7 +61,7 @@ can also be input by using the @kbd{C-x 8} prefix, see 
@ref{Unibyte Mode}.
 
 With the X Window System, your locale should be set to an appropriate
 value to make sure Emacs interprets keyboard input correctly; see
-@ref{Language Environments, locales}.
+@ref{Language Environments, locales}, and @ref{X Coding}.
 @end itemize
 
   The rest of this chapter describes these issues in detail.
@@ -79,6 +79,7 @@ value to make sure Emacs interprets keyboard input correctly; 
see
 * Text Coding::             Choosing conversion to use for file text.
 * Communication Coding::    Coding systems for interprocess communication.
 * File Name Coding::        Coding systems for file @emph{names}.
+* X Coding::                Coding systems for X input methods.
 * Terminal Coding::         Specifying coding systems for converting
                               terminal input and output.
 * Fontsets::                Fontsets are collections of fonts
@@ -1241,15 +1242,14 @@ current language environment.
   The variable @code{locale-coding-system} specifies a coding system
 to use when encoding and decoding system strings such as system error
 messages and @code{format-time-string} formats and time stamps.  That
-coding system is also used for decoding non-@acronym{ASCII} keyboard
-input on the X Window System and for encoding text sent to the
-standard output and error streams when in batch mode.  You should
-choose a coding system that is compatible
-with the underlying system's text representation, which is normally
-specified by one of the environment variables @env{LC_ALL},
-@env{LC_CTYPE}, and @env{LANG}.  (The first one, in the order
-specified above, whose value is nonempty is the one that determines
-the text representation.)
+coding system might also be used for decoding non-@acronym{ASCII}
+keyboard input on the X Window System and will also be used to encode
+text sent to the standard output and error streams in batch mode.  You
+should choose a coding system that is compatible with the underlying
+system's text representation, which is normally specified by one of
+the environment variables @env{LC_ALL}, @env{LC_CTYPE}, and
+@env{LANG}.  (The first one, in the order specified above, whose value
+is nonempty is the one that determines the text representation.)
 
 @node File Name Coding
 @section Coding Systems for File Names
@@ -1311,6 +1311,26 @@ C-w} to specify a new file name for that buffer.
 system.  This prompts for an existing file name, its old coding
 system, and the coding system to which you wish to convert.
 
+@node X Coding
+@section Coding Systems for X Keyboard Input
+@cindex X input method coding systems
+  Input methods under the X Window System specify their own coding
+systems that must be used to decode keyboard input.  By default, Emacs
+determines the coding system used for each input method automatically
+upon establishing the connection to the input method server, and uses
+that specific coding system to decode keyboard input.  However, that
+determination can sometimes fail; in that situation, the locale coding
+system (@pxref{Communication Coding}) is used instead.
+
+@cindex X input method coding systems, overriding
+@vindex x-input-coding-system
+  If the input method does not correctly announce the coding system it
+uses to encode text, then the coding system used by Emacs to decode
+text from input methods must be manually specified.  The value of the
+variable @code{x-input-coding-system}, when set to a symbol, is
+unconditionally used as the coding system used to decode keyboard
+input from input methods.
+
 @node Terminal Coding
 @section Coding Systems for Terminal I/O
 
diff --git a/etc/NEWS b/etc/NEWS
index db7c675edb..d7bc4b0e0c 100644
--- a/etc/NEWS
+++ b/etc/NEWS
@@ -186,6 +186,19 @@ of 'user-emacs-directory'.
 
 * Incompatible changes in Emacs 29.1
 
++++
+*** Emacs now picks the correct coding system for X input methods.
+Previously, Emacs would use the locale coding system for input
+methods, which could in some circumstances be incorrect, especially
+when the input method chose to fall back to some other coding system.
+
+Now, Emacs automatically detects the coding system used by input
+methods, and uses that to decode input in preference to the value of
+'locale-coding-system'.  This unfortunately means that users who have
+changed the coding system used to decode X keyboard input must adjust
+their customizations to 'locale-coding-system' to the variable
+'x-input-coding-system' instead.
+
 +++
 *** Bookmarks no longer include context for encrypted files.
 If you're visiting an encrypted file, setting a bookmark no longer
diff --git a/lisp/term/x-win.el b/lisp/term/x-win.el
index 9d3e780365..57c6b785e7 100644
--- a/lisp/term/x-win.el
+++ b/lisp/term/x-win.el
@@ -1613,6 +1613,28 @@ Users should not call this function; see `device-class' 
instead."
 (setq x-dnd-movement-function #'x-dnd-movement)
 (setq x-dnd-unsupported-drop-function #'x-dnd-handle-unsupported-drop)
 
+(defvar x-input-coding-function)
+
+(defun x-get-input-coding-system (x-locale)
+  "Return a coding system for the locale X-LOCALE.
+Return a coding system that is able to decode text sent with the
+X input method locale X-LOCALE, or nil if no coding system was
+found."
+  (if (equal x-locale "C")
+      ;; Treat the C locale specially, as it means "ascii" under X.
+      'ascii
+    (let ((locale (downcase x-locale)))
+      (or (locale-name-match locale locale-preferred-coding-systems)
+         (when locale
+           (if (string-match "\\.\\([^@]+\\)" locale)
+               (locale-charset-to-coding-system
+                (match-string 1 locale))))
+          (let ((language-name
+                 (locale-name-match locale locale-language-names)))
+            (and (consp language-name) (cdr language-name)))))))
+
+(setq x-input-coding-function #'x-get-input-coding-system)
+
 (provide 'x-win)
 (provide 'term/x-win)
 
diff --git a/src/coding.c b/src/coding.c
index 0ae8eb3282..ab73bda844 100644
--- a/src/coding.c
+++ b/src/coding.c
@@ -12014,9 +12014,9 @@ See also the function `find-operation-coding-system'.  
*/);
   Vnetwork_coding_system_alist = Qnil;
 
   DEFVAR_LISP ("locale-coding-system", Vlocale_coding_system,
-              doc: /* Coding system to use with system messages.
-Also used for decoding keyboard input on X Window system, and for
-encoding standard output and error streams.  */);
+    doc: /* Coding system to use with system messages.
+Potentially also used for decoding keyboard input on X Windows, and is
+used for encoding standard output and error streams.  */);
   Vlocale_coding_system = Qnil;
 
   /* The eol mnemonics are reset in startup.el system-dependently.  */
diff --git a/src/xfns.c b/src/xfns.c
index bb75ca5ad1..8cea93c669 100644
--- a/src/xfns.c
+++ b/src/xfns.c
@@ -3354,22 +3354,30 @@ struct x_xim_text_conversion_data
 {
   struct coding_system *coding;
   char *source;
+  struct x_display_info *dpyinfo;
 };
 
 static Lisp_Object
-x_xim_text_to_utf8_unix_1 (ptrdiff_t nargs,
-                          Lisp_Object *args)
+x_xim_text_to_utf8_unix_1 (ptrdiff_t nargs, Lisp_Object *args)
 {
   struct x_xim_text_conversion_data *data;
   ptrdiff_t nbytes;
+  Lisp_Object coding_system;
 
   data = xmint_pointer (args[0]);
+
+  if (SYMBOLP (Vx_input_coding_system))
+    coding_system = Vx_input_coding_system;
+  else if (!NILP (data->dpyinfo->xim_coding))
+    coding_system = data->dpyinfo->xim_coding;
+  else
+    coding_system = Vlocale_coding_system;
+
   nbytes = strlen (data->source);
 
   data->coding->destination = NULL;
 
-  setup_coding_system (Vlocale_coding_system,
-                      data->coding);
+  setup_coding_system (coding_system, data->coding);
   data->coding->mode |= (CODING_MODE_LAST_BLOCK
                         | CODING_MODE_SAFE_ENCODING);
   data->coding->source = (const unsigned char *) data->source;
@@ -3382,8 +3390,7 @@ x_xim_text_to_utf8_unix_1 (ptrdiff_t nargs,
 }
 
 static Lisp_Object
-x_xim_text_to_utf8_unix_2 (Lisp_Object val,
-                          ptrdiff_t nargs,
+x_xim_text_to_utf8_unix_2 (Lisp_Object val, ptrdiff_t nargs,
                           Lisp_Object *args)
 {
   struct x_xim_text_conversion_data *data;
@@ -3400,7 +3407,8 @@ x_xim_text_to_utf8_unix_2 (Lisp_Object val,
 
 /* The string returned is not null-terminated.  */
 static char *
-x_xim_text_to_utf8_unix (XIMText *text, ptrdiff_t *length)
+x_xim_text_to_utf8_unix (struct x_display_info *dpyinfo,
+                        XIMText *text, ptrdiff_t *length)
 {
   unsigned char *wchar_buf;
   ptrdiff_t wchar_actual_length, i;
@@ -3424,6 +3432,7 @@ x_xim_text_to_utf8_unix (XIMText *text, ptrdiff_t *length)
 
   data.coding = &coding;
   data.source = text->string.multi_byte;
+  data.dpyinfo = dpyinfo;
 
   was_waiting_for_input_p = waiting_for_input;
   /* Otherwise Fsignal will crash.  */
@@ -3441,18 +3450,21 @@ static void
 xic_preedit_draw_callback (XIC xic, XPointer client_data,
                           XIMPreeditDrawCallbackStruct *call_data)
 {
-  struct frame *f = x_xic_to_frame (xic);
+  struct frame *f;
   struct x_output *output;
-  ptrdiff_t text_length = 0;
+  ptrdiff_t text_length;
   ptrdiff_t charpos;
   ptrdiff_t original_size;
   char *text;
   char *chg_start, *chg_end;
   struct input_event ie;
+
+  f = x_xic_to_frame (xic);
   EVENT_INIT (ie);
 
   if (f)
     {
+      text_length = 0;
       output = FRAME_X_OUTPUT (f);
 
       if (!output->preedit_active)
@@ -3460,7 +3472,8 @@ xic_preedit_draw_callback (XIC xic, XPointer client_data,
 
       if (call_data->text)
        {
-         text = x_xim_text_to_utf8_unix (call_data->text, &text_length);
+         text = x_xim_text_to_utf8_unix (FRAME_DISPLAY_INFO (f),
+                                         call_data->text, &text_length);
 
          if (!text)
            /* Decoding the IM text failed.  */
diff --git a/src/xterm.c b/src/xterm.c
index d83b03d1ca..42335f0de0 100644
--- a/src/xterm.c
+++ b/src/xterm.c
@@ -19139,7 +19139,7 @@ handle_one_xevent (struct x_display_info *dpyinfo,
                                         &xkey, (char *) copy_bufptr,
                                         copy_bufsiz, &keysym,
                                         &status_return);
-             coding = Qnil;
+             coding = FRAME_X_XIM_CODING (f);
               if (status_return == XBufferOverflow)
                 {
                   copy_bufsiz = nbytes + 1;
@@ -22828,7 +22828,7 @@ handle_one_xevent (struct x_display_info *dpyinfo,
                                                &xkey, (char *) copy_bufptr,
                                                copy_bufsiz, &keysym,
                                                &status_return);
-                     coding = Qnil;
+                     coding = FRAME_X_XIM_CODING (f);
 
                      if (status_return == XBufferOverflow)
                        {
@@ -25488,9 +25488,10 @@ xim_destroy_callback (XIM xim, XPointer client_data, 
XPointer call_data)
 static void
 xim_open_dpy (struct x_display_info *dpyinfo, char *resource_name)
 {
+#ifdef HAVE_XIM
   XIM xim;
+  const char *locale;
 
-#ifdef HAVE_XIM
   if (use_xim)
     {
       if (dpyinfo->xim)
@@ -25513,6 +25514,14 @@ xim_open_dpy (struct x_display_info *dpyinfo, char 
*resource_name)
          destroy.client_data = (XPointer)dpyinfo;
          XSetIMValues (xim, XNDestroyCallback, &destroy, NULL);
 #endif
+
+         locale = XLocaleOfIM (xim);
+
+         /* Now try to determine the coding system that should be
+            used.  locale is in Host Portable Character Encoding, and
+            as such can be passed to build_string as is.  */
+         dpyinfo->xim_coding = safe_call1 (Vx_input_coding_function,
+                                           build_string (locale));
        }
     }
 
@@ -29843,7 +29852,7 @@ mark_xterm (void)
     }
 
 #if defined HAVE_XINPUT2 || defined USE_TOOLKIT_SCROLL_BARS \
-  || defined HAVE_XRANDR || defined USE_GTK
+  || defined HAVE_XRANDR || defined USE_GTK || defined HAVE_X_I18N
   for (dpyinfo = x_display_list; dpyinfo; dpyinfo = dpyinfo->next)
     {
 #ifdef HAVE_XINPUT2
@@ -29856,6 +29865,9 @@ mark_xterm (void)
 #endif
 #if defined HAVE_XRANDR || defined USE_GTK
       mark_object (dpyinfo->last_monitor_attributes_list);
+#endif
+#if defined HAVE_X_I18N
+      mark_object (dpyinfo->xim_coding);
 #endif
     }
 #endif
@@ -30385,4 +30397,18 @@ on the same display.
 In addition, when this variable is a list, only preserve the
 selections whose names are contained within.  */);
   Vx_auto_preserve_selections = list2 (QCLIPBOARD, QPRIMARY);
+
+  DEFVAR_LISP ("x-input-coding-system", Vx_input_coding_system,
+    doc: /* Coding system used for input from X input methods.
+If a symbol and non-nil, this is the coding system that will be used
+to decode input from X input methods.  It does not affect input from
+GTK native input methods enabled through `x-gtk-use-native-input'.  */);
+  Vx_input_coding_system = Qnil;
+
+  DEFVAR_LISP ("x-input-coding-function", Vx_input_coding_function,
+    doc: /* Function used to determine the coding system used by input methods.
+It should accept a single argument, a string describing the locale of
+the input method, and return a coding system that can decode keyboard
+input generated by said input method.  */);
+  Vx_input_coding_function = Qnil;
 }
diff --git a/src/xterm.h b/src/xterm.h
index f3791aa8df..b68a234faa 100644
--- a/src/xterm.h
+++ b/src/xterm.h
@@ -580,6 +580,9 @@ struct x_display_info
   XIMStyles *xim_styles;
   struct xim_inst_t *xim_callback_data;
   XIMStyle preferred_xim_style;
+
+  /* The named coding system to use for this input method.  */
+  Lisp_Object xim_coding;
 #endif
 
   /* A cache mapping color names to RGB values.  */
@@ -1348,6 +1351,12 @@ extern void x_mark_frame_dirty (struct frame *f);
 #define FRAME_X_XIM_STYLES(f) (FRAME_DISPLAY_INFO (f)->xim_styles)
 #define FRAME_XIC_STYLE(f) ((f)->output_data.x->xic_style)
 #define FRAME_XIC_FONTSET(f) ((f)->output_data.x->xic_xfs)
+#define FRAME_X_XIM_CODING(f)                          \
+  (SYMBOLP (Vx_input_coding_system)                    \
+   ? Vx_input_coding_system                            \
+   : (!NILP (FRAME_DISPLAY_INFO (f)->xim_coding)       \
+      ? FRAME_DISPLAY_INFO(f)->xim_coding              \
+      : Vlocale_coding_system))
 
 /* X-specific scroll bar stuff.  */