[COMMIT] Call XKeysymToString() much less, it leaks.

APPROVE COMMIT NOTE: This patch has been committed. # HG changeset patch # User Aidan Kehoe <kehoea(a)parhasard.net&gt; # Date 1326055297 0 # Node ID 1d1f385c9149cdf5cf013c286c8c073ee2c890b5 # Parent 4050bbef9bf650e5af74d87011d8779621594772 Call XKeysymToString() much less, it leaks. src/ChangeLog addition: 2012-01-08 Aidan Kehoe <kehoea(a)parhasard.net&gt; * device-x.c: * device-x.c (syms_of_device_x): Move #'x-keysym-on-keyboard{,-sans-modifiers}-p to Lisp, the hash table no longer stores the X keysyms, so we need to manipulate any strings we have been handed. * event-Xt.c (x_has_keysym): Don't call XKeysymToString() here, it leaks; trust x_keysym_to_emacs_keysym() instead. * event-Xt.c (x_keysym_to_emacs_keysym): No longer leak when looking up the strings for keysyms of the form UABCD. lisp/ChangeLog addition: 2012-01-08 Aidan Kehoe <kehoea(a)parhasard.net&gt; * x-init.el: * x-init.el (pseudo-canonicalize-keysym): New. * x-init.el: Move #'x-keysym-on-keyboard-sans-modifiers-p, #'x-keysym-on-keyboard-p here from device-x.c, some string manipulation it now needs to do is far easier in Lisp. * x-win-xfree86.el: * x-win-xfree86.el (x-win-init-xfree86): No longer call #'x-keysym-on-keyboard{,-sans-modifiers}-p, implement it ourselves cheaply. diff -r 4050bbef9bf6 -r 1d1f385c9149 lisp/ChangeLog --- a/lisp/ChangeLog Fri Jan 06 16:35:07 2012 +0000 +++ b/lisp/ChangeLog Sun Jan 08 20:41:37 2012 +0000 ＠＠ -1,3 +1,15 ＠＠ +2012-01-08 Aidan Kehoe <kehoea(a)parhasard.net&gt; + + * x-init.el: + * x-init.el (pseudo-canonicalize-keysym): New. + * x-init.el: Move #'x-keysym-on-keyboard-sans-modifiers-p, + #'x-keysym-on-keyboard-p here from device-x.c, some string + manipulation it now needs to do is far easier in Lisp. + * x-win-xfree86.el: + * x-win-xfree86.el (x-win-init-xfree86): + No longer call #'x-keysym-on-keyboard{,-sans-modifiers}-p, + implement it ourselves cheaply. + 2012-01-02 Aidan Kehoe <kehoea(a)parhasard.net&gt; * bytecomp.el (cl-compile-time-init): Add an autoload for this, diff -r 4050bbef9bf6 -r 1d1f385c9149 lisp/x-init.el --- a/lisp/x-init.el Fri Jan 06 16:35:07 2012 +0000 +++ b/lisp/x-init.el Sun Jan 08 20:41:37 2012 +0000 ＠＠ -31,8 +31,7 ＠＠ ;;; Code: (globally-declare-fboundp - '(x-keysym-on-keyboard-p - x-server-vendor x-init-specifier-from-resources init-mule-x-win)) + '(x-server-vendor x-init-specifier-from-resources init-mule-x-win)) (globally-declare-boundp '(x-initial-argv-list x-app-defaults-directory)) ＠＠ -79,6 +78,56 ＠＠ (interactive) (ow-find t)) +(labels + ((pseudo-canonicalize-keysym (keysym) + "If KEYSYM (a string or a symbol) might describe a keysym on +the current keyboard, return its canonical XEmacs form, a symbol; +otherwise return nil. + +Does not intern new symbols, since if a string doesn't correspond to a +keysym that XEmacs has seen, that string won't be in obarray." + (if (symbolp keysym) + keysym + (if (stringp keysym) + (or (intern-soft keysym) + (intern-soft (nsubstitute ?- ?_ (downcase keysym)))))))) + (declare (inline pseudo-canonicalize-keysym)) + + (defun x-keysym-on-keyboard-sans-modifiers-p (keysym &optional device) + "Return true if KEYSYM names a key on the keyboard of DEVICE. +More precisely, return true if pressing a physical key +on the keyboard of DEVICE without any modifier keys generates KEYSYM. +Valid keysyms are listed in the files /usr/include/X11/keysymdef.h and in +/usr/lib/X11/XKeysymDB, or whatever the equivalents are on your system. +The keysym name can be provided in two forms: +- if keysym is a string, it must be the name as known to X windows. +- if keysym is a symbol, it must be the name as known to XEmacs. +The two names differ in capitalization and underscoring." + (eq 'sans-modifiers (gethash (pseudo-canonicalize-keysym keysym) + (x-keysym-hash-table device)))) + + (defun x-keysym-on-keyboard-p (keysym &optional device) + "Return true if KEYSYM names a key on the keyboard of DEVICE. +More precisely, return true if some keystroke (possibly including modifiers) +on the keyboard of DEVICE keys generates KEYSYM. +Valid keysyms are listed in the files /usr/include/X11/keysymdef.h and in +/usr/lib/X11/XKeysymDB, or whatever the equivalents are on your system. +The keysym name can be provided in two forms: +- if keysym is a string, it must be the name as known to X windows. +- if keysym is a symbol, it must be the name as known to XEmacs. +The two names differ in capitalization and underscoring. + +This function is not entirely trustworthy, in that Xlib compose processing +can produce keysyms that XEmacs will not have seen when it examined the +keysyms available on startup. So pressing `dead-diaeresis' and then 'a' may +pass `adiaeresis' to XEmacs, or (in some implementations) even `U00E4', +where `(x-keysym-on-keyboard-p 'adiaeresis)' and `(x-keysym-on-keyboard-p +'U00E4)' would both have returned nil. Subsequent to XEmacs seeing a keysym +it was previously unaware of, the predicate will take note of it, though." + (and (gethash (pseudo-canonicalize-keysym keysym) + (x-keysym-hash-table device)) + t))) + (eval-when-compile (load "x-win-sun" nil t) (load "x-win-xfree86" nil t)) diff -r 4050bbef9bf6 -r 1d1f385c9149 lisp/x-win-xfree86.el --- a/lisp/x-win-xfree86.el Fri Jan 06 16:35:07 2012 +0000 +++ b/lisp/x-win-xfree86.el Sun Jan 08 20:41:37 2012 +0000 ＠＠ -39,9 +39,6 ＠＠ ;;; Code: -(globally-declare-fboundp - '(x-keysym-on-keyboard-p x-keysym-on-keyboard-sans-modifiers-p)) - (defun x-win-init-xfree86 (device) ;; We know this keyboard is an XFree86 keyboard. As such, we can predict ＠＠ -74,9 +71,6 ＠＠ nil nil nil nil nil ?/ nil nil nil nil nil nil nil nil nil nil nil nil nil ?=]) - (when (x-keysym-on-keyboard-p 'iso-left-tab device) - (define-key function-key-map 'iso-left-tab [(shift tab)])) - (loop for (key sane-key) in '((f13 f1) (f14 f2) ＠＠ -94,13 +88,17 ＠＠ with function-key-map = (symbol-value-in-console 'function-key-map (device-console device) function-key-map) + with x-keysym-hash-table = (x-keysym-hash-table device) + + initially + (when (gethash 'iso-left-tab x-keysym-hash-table) + (define-key function-key-map 'iso-left-tab [(shift tab)])) do - (when (and (x-keysym-on-keyboard-p key device) - (not (x-keysym-on-keyboard-sans-modifiers-p key device))) + (when (not (memq (gethash key x-keysym-hash-table) '(nil sans-modifiers))) ;; define also the control, meta, and meta-control versions. (loop for mods in '(() (control) (meta) (meta control)) do - (define-key function-key-map `[(,＠mods ,key)] `[(shift ,＠mods ,sane-key)]) - )))) + (define-key function-key-map `[(,＠mods ,key)] + `[(shift ,＠mods ,sane-key)]))))) ;;; x-win-xfree86.el ends here diff -r 4050bbef9bf6 -r 1d1f385c9149 src/ChangeLog --- a/src/ChangeLog Fri Jan 06 16:35:07 2012 +0000 +++ b/src/ChangeLog Sun Jan 08 20:41:37 2012 +0000 ＠＠ -1,3 +1,17 ＠＠ +2012-01-08 Aidan Kehoe <kehoea(a)parhasard.net&gt; + + * device-x.c: + * device-x.c (syms_of_device_x): + Move #'x-keysym-on-keyboard{,-sans-modifiers}-p to Lisp, the hash + table no longer stores the X keysyms, so we need to manipulate any + strings we have been handed. + * event-Xt.c (x_has_keysym): + Don't call XKeysymToString() here, it leaks; trust + x_keysym_to_emacs_keysym() instead. + * event-Xt.c (x_keysym_to_emacs_keysym): + No longer leak when looking up the strings for keysyms of the form + UABCD. + 2012-01-03 Didier Verna <didier(a)xemacs.org&gt; * faces.c (complex_vars_of_faces): Add missing foreback specifier diff -r 4050bbef9bf6 -r 1d1f385c9149 src/device-x.c --- a/src/device-x.c Fri Jan 06 16:35:07 2012 +0000 +++ b/src/device-x.c Sun Jan 08 20:41:37 2012 +0000 ＠＠ -1830,60 +1830,6 ＠＠ return DEVICE_X_DATA (d)->x_keysym_map_hash_table; } - -DEFUN ("x-keysym-on-keyboard-sans-modifiers-p", Fx_keysym_on_keyboard_sans_modifiers_p, - 1, 2, 0, /* -Return true if KEYSYM names a key on the keyboard of DEVICE. -More precisely, return true if pressing a physical key -on the keyboard of DEVICE without any modifier keys generates KEYSYM. -Valid keysyms are listed in the files /usr/include/X11/keysymdef.h and in -/usr/lib/X11/XKeysymDB, or whatever the equivalents are on your system. -The keysym name can be provided in two forms: -- if keysym is a string, it must be the name as known to X windows. -- if keysym is a symbol, it must be the name as known to XEmacs. -The two names differ in capitalization and underscoring. -*/ - (keysym, device)) -{ - struct device *d = decode_device (device); - if (!DEVICE_X_P (d)) - gui_error ("Not an X device", device); - - return (EQ (Qsans_modifiers, - Fgethash (keysym, DEVICE_X_KEYSYM_MAP_HASH_TABLE (d), Qnil)) ? - Qt : Qnil); -} - - -DEFUN ("x-keysym-on-keyboard-p", Fx_keysym_on_keyboard_p, 1, 2, 0, /* -Return true if KEYSYM names a key on the keyboard of DEVICE. -More precisely, return true if some keystroke (possibly including modifiers) -on the keyboard of DEVICE keys generates KEYSYM. -Valid keysyms are listed in the files /usr/include/X11/keysymdef.h and in -/usr/lib/X11/XKeysymDB, or whatever the equivalents are on your system. -The keysym name can be provided in two forms: -- if keysym is a string, it must be the name as known to X windows. -- if keysym is a symbol, it must be the name as known to XEmacs. -The two names differ in capitalization and underscoring. - -This function is not entirely trustworthy, in that Xlib compose processing -can produce keysyms that XEmacs will not have seen when it examined the -keysyms available on startup. So pressing `dead-diaeresis' and then 'a' may -pass `adiaeresis' to XEmacs, or (in some implementations) even `U00E4', -where `(x-keysym-on-keyboard-p 'adiaeresis)' and `(x-keysym-on-keyboard-p -'U00E4)' would both have returned nil. Subsequent to XEmacs seeing a keysym -it was previously unaware of, the predicate will take note of it, though. -*/ - (keysym, device)) -{ - struct device *d = decode_device (device); - if (!DEVICE_X_P (d)) - gui_error ("Not an X device", device); - - return (NILP (Fgethash (keysym, DEVICE_X_KEYSYM_MAP_HASH_TABLE (d), Qnil)) ? - Qnil : Qt); -} -  /************************************************************************/ /* grabs and ungrabs */ ＠＠ -2105,8 +2051,6 ＠＠ DEFSUBR (Fx_server_version); DEFSUBR (Fx_valid_keysym_name_p); DEFSUBR (Fx_keysym_hash_table); - DEFSUBR (Fx_keysym_on_keyboard_p); - DEFSUBR (Fx_keysym_on_keyboard_sans_modifiers_p); DEFSUBR (Fx_grab_pointer); DEFSUBR (Fx_ungrab_pointer); diff -r 4050bbef9bf6 -r 1d1f385c9149 src/event-Xt.c --- a/src/event-Xt.c Fri Jan 06 16:35:07 2012 +0000 +++ b/src/event-Xt.c Sun Jan 08 20:41:37 2012 +0000 ＠＠ -181,26 +181,17 ＠＠ for (j = 0; j < (upper_lower[0] == upper_lower[1] ? 1 : 2); j++) { - Extbyte *name; - keysym = upper_lower[j]; - - name = XKeysymToString (keysym); - if (name) - { - /* X guarantees NAME to be in the Host Portable Character Encoding */ - Lisp_Object sym = x_keysym_to_emacs_keysym (keysym, 0); - Lisp_Object new_value = with_modifiers ? Qt : Qsans_modifiers; - Lisp_Object old_value = Fgethash (sym, hash_table, Qnil); - - if (! EQ (old_value, new_value) - && ! (EQ (old_value, Qsans_modifiers) && - EQ (new_value, Qt))) - { - maybe_define_x_key_as_self_inserting_character (keysym, sym); - Fputhash (build_extstring (name, Qbinary), new_value, - hash_table); - Fputhash (sym, new_value, hash_table); - } + KeySym ks = upper_lower[j]; + Lisp_Object sym = x_keysym_to_emacs_keysym (ks, 0); + Lisp_Object new_value = with_modifiers ? Qt : Qsans_modifiers; + Lisp_Object old_value = Fgethash (sym, hash_table, Qnil); + + if (! EQ (old_value, new_value) + && ! (EQ (old_value, Qsans_modifiers) && + EQ (new_value, Qt))) + { + maybe_define_x_key_as_self_inserting_character (ks, sym); + Fputhash (sym, new_value, hash_table); } } } ＠＠ -757,6 +748,17 ＠＠ characters are the same as their ASCII codes. */ return make_char (keysym); + if (keysym >= 0x01000000 && keysym <= 0x0110FFFF) + { + /* These keysyms malloc with XKeysymToString(), *every time the + function is called.* Avoid leaking, construct the keysym string + ourselves. */ + Ascbyte buf [10]; + qxesprintf ((Ibyte *) buf, keysym & 0xff0000 ? "U%06X" : "U%04X", + (unsigned int) (keysym & 0xffffff)); + return KEYSYM (buf); + } + switch (keysym) { /* These would be handled correctly by the default case, but by -- ‘Iodine deficiency was endemic in parts of the UK until, through what has been described as “an unplanned and accidental public health triumph”, iodine was added to cattle feed to improve milk production in the 1930s.’ (EN Pearce, Lancet, June 2011) _______________________________________________ XEmacs-Patches mailing list XEmacs-Patches(a)xemacs.org http://lists.xemacs.org/mailman/listinfo/xemacs-patches