This is ugly, but it should fix the performance problem for older
GCC versions in the short run. I'll look into integrating the
Gnulib module for popcount, as a better fix.
See the thread starting in:
http://lists.gnu.org/archive/html/emacs-devel/2013-09/msg00474.html
(popcount_size_t_generic) [NEED_GENERIC_POPCOUNT]:
(popcount_size_t_msc) [USE_MSC_POPCOUNT]:
(popcount_size_t_gcc) [USE_GCC_POPCOUNT]:
(popcount_size_t): Use it.
Prefer signed to unsigned integers where either will do.
No need for 'const' on locals that do not escape.
Omit easserts with unnecessary and unportable assumptions about
alignment. Avoid unnecessary casts to char *.
This helps to avoid a few glitches dictated by C's aliasing rules.
* lisp.h (struct Lisp_Vector): Use union for next and
contents member. Adjust comment. Change related users.
* alloc.c (next_in_free_list, set_next_in_free_list): Remove.
Related users changed.
* buffer.c, bytecode.c, ccl.c, character.h, chartab.c, composite.c:
* composite.h, disptab.h, fns.c, fontset.c, indent.c, keyboard.c:
* lread.c, msdos.c, process.c, w32menu.c, window.c, xdisp.c:
* xfaces.c, xfont.c, xmenu.c: Related users changed.
* dispextern.h (struct glyph_row): Change layout of struct
glyph_row to help copy_row_except_pointers. Adjust comment.
* dispnew.c (null_row): Remove.
(clear_glyph_row): Use offsetof and memset to find and clear
just the members that need clearing. Adjust comment.
(copy_row_except_pointers): Likewise for copying.
* conf_post.h (assume): Always return void. Use lint version
only if GCC and MSC versions don't apply.
* conf_post.h (assume):
* data.c (USC_MSC_POPCOUNT, count_trailing_zero_bits):
Depend on _MSC_VER, not __MSC_VER, for consistency with
the rest of Emacs.
* data.c (bool_vector_spare_mask, popcount_size_t_generic)
(popcount_size_t_msc, popcount_size_t_gcc, popcount_size_t)
(bool_vector_binop_driver, count_trailing_zero_bits)
(size_t_to_host_endian): Now static, not static inline;
the latter isn't needed with modern compilers and doesn't
work with older compilers anyway.
src/xdisp.c (mouse_face_from_string_pos): Fix off-by-one error in
computing the end column of mouse-highlight that comes from
display or overlay strings.
* configure.ac: With clang, check for and use -Wno-switch,
-Wno-tautological-constant-out-of-range-compare and -Wno-pointer-sign.
* conf_post.h(assume): Use __builtin_unreachable for clang.
* src/filelock.c (lock_file_1): Rearrange to remove compiler warning
about excess arguments to snprintf.
warings.
* alloc.c (USE_VALGRIND): New macro; on by default
when ENABLE_CHECKING.
(mark_maybe_object,mark_maybe_pointer)
[USE_VALGRIND]: Mark conservatively-scanned regions valid for
valgrind purposes.
(valgrind_p) [USE_VALGRIND]: New variable.
(init_alloc) [USE_VALGRIND]: Initialize valgrind_p.
* nsterm.m (setMarkedText:selectedRange:):
(deleteWorkingText):
* nsmenu.m (addDisplayItemWithImage:idx:tag:helpText:enabled:):
* nsfont.m (ns_get_covering_families, ns_findfonts): Cast NSLog
argument to unsigned long to avoid warning.
(nsfont_draw): Use 0.25 instead of Fix2X (kATSItalicQDSkew).
* process.c (wait_reading_process_output): Change int pnamelen to
socklen_t.
http://lists.gnu.org/archive/html/emacs-devel/2013-09/msg00404.html
* data.c (Qbool_vector_p): New symbol.
(bool_vector_spare_mask,popcount_size_t_generic)
(popcount_size_t_msc,popcount_size_t_gcc)
(popcount_size_t)
(bool_vector_binop_driver)
(count_trailing_zero_bits,size_t_to_host_endian)
(Fbool_vector_exclusive_or)
(Fbool_vector_union)
(Fbool_vector_intersection,Fbool_vector_set_difference)
(Fbool_vector_subsetp,Fbool_vector_not)
(Fbool_vector_count_matches)
(Fbool_vector_count_matches_at): New functions.
(syms_of_data): Intern new symbol, functions.
* alloc.c (bool_vector_payload_bytes): New function.
(Fmake_bool_vector): Instead of calling Fmake_vector,
which performs redundant initialization and argument checking,
just call allocate_vector ourselves. Make sure we clear any
terminating padding to zero.
(vector_nbytes,sweep_vectors): Use bool_vector_payload_bytes
instead of open-coding the size calculation.
(vroundup_ct): New macro.
(vroundup): Assume argument >= 0; invoke vroundup_ct.
* casetab.c (shuffle,set_identity): Change lint_assume to assume.
* composite.c (composition_gstring_put_cache): Change
lint_assume to assume.
* conf_post.h (assume): New macro.
(lint_assume): Remove.
* dispnew.c (update_frame_1): Change lint_assume to assume.
* ftfont.c (ftfont_shape_by_flt): Change lint_assume
to assume.
* image.c (gif_load): Change lint_assume to assume.
* lisp.h (eassert_and_assume): New macro.
(Qbool_vector_p): Declare.
(CHECK_BOOL_VECTOR,ROUNDUP,BITS_PER_SIZE_T): New macros.
(swap16,swap32,swap64): New inline functions.
* macfont.c (macfont_shape): Change lint_assume to assume.
* ralloc.c: Rename ROUNDUP to PAGE_ROUNDUP throughout.
* xsettings.c (parse_settings): Use new swap16 and
swap32 from lisp.h instead of file-specific macros.
src/xdisp.c (try_window_id): Don't abort if cursor row could not be
found (which can legitimately happen when the glyph row at the
window start is disabled in the current_matrix.
syntax.h defined an extern inline function SYNTAX_ENTRY that was
conditionally compiled one way in some modules, and a different
way in others. This doesn't work with extern inline functions,
which must have the same definition in all modules, because the
defining code might be shared across modules, depending on the
implementation. Symptoms reported by Martin Rudalics in:
http://lists.gnu.org/archive/html/emacs-devel/2013-09/msg00414.html
* regex.c, syntax.c (SYNTAX_ENTRY_VIA_PROPERTY): Remove.
(SYNTAX, SYNTAX_ENTRY, SYNTAX_WITH_FLAGS): New macros,
overriding the corresponding functions in syntax.h.
* syntax.h (syntax_property_entry, syntax_property_with_flags)
(syntax_property): New inline functions.
(SYNTAX_ENTRY, SYNTAX_WITH_FLAGS, SYNTAX):
Rewrite in terms of these new functions.
src/dired.c (directory_files_internal): Use multibyte_chars_in_text,
not chars_in_text, whose result depends on the multibyteness of
the current buffer.
* xterm.c (x_send_scroll_bar_event, x_scroll_bar_to_input_event):
Don't assume that pointers are the same width as 'long'.
Add a compile-time check that a pointer fits into two X slots.
src/w32reg.c (w32_get_string_resource): Make the first 2 arguments
'const char *' to avoid compiler warnings due to similar change in
the prototype of x_get_string_resource.
Ref: http://lists.gnu.org/archive/html/emacs-devel/2010-02/msg00135.html
* lisp/simple.el (messages-buffer-mode): New major mode.
(messages-buffer): New function.
* lisp/startup.el (normal-top-level): Switch mode of *Messages* buffer.
* src/xdisp.c (message_dolog): If we create *Messages*,
switch it to messages-buffer-mode.
* lisp/emacs-lisp/ert.el (ert--force-message-log-buffer-truncation)
(ert-run-test): Use `message-buffer' function.
(ert--force-message-log-buffer-truncation): Ignore read-only.
* lisp/help.el (view-echo-area-messages): Use `message-buffer' function.
* lisp/mail/emacsbug.el (report-emacs-bug): Use `message-buffer' function.
* lisp/gnus/gnus-util.el (gnus-message-with-timestamp-1):
Use `message-buffer' function if available. Ignore read-only.
* etc/NEWS: Mention this.
* bidi.c (bidi_count_bytes):
* gtkutil.c, gtkutil.h (xg_event_is_for_menubar)
(xg_event_is_for_scrollbar):
* xselect.c (x_handle_property_notify)
(x_handle_selection_notify, x_handle_dnd_message):
* xsettings.c, xsettings.h (xft_settings_event):
* xterm.c (x_handle_net_wm_state, handle_one_event)
(x_menubar_window_to_frame, x_detect_focus_change)
(construct_mouse_click, note_mouse_movement)
(x_scroll_bar_to_input_event, x_scroll_bar_expose)
(x_scroll_bar_handle_click, x_scroll_bar_note_movement)
(handle_one_xevent, x_handle_net_wm_state):
* xterm.h (x_handle_property_notify, x_handle_selection_notify)
(x_handle_dnd_message):
Avoid unnecessary 'const', typically the second 'const' in
'const foo * const arg', a 'const' that does not affect the API
and doesn't significantly help the human reader.
require glib >= 2.28 for GTK3, glib >= 2.10 for GTK2,
glib >= 2.26 for GSettings and glib >= 2.7.0 for GConf, so
suitable glib should provide g_type_init unconditionally.
* src/image.c (fn_g_type_init) [WINDOWSNT]: Define and load
only if Glib < 2.36.0.
(fn_g_type_init) [!WINDOWSNT]: Define only if Glib < 2.36.0.
* src/xsettings.c (init_gconf, init_gsettings): Do not check
for g_type_init.
Wnen XEvent is processed, it is unlikely to be changed except
KeyPress case, so we can avoid copying and use const pointer to
const data to make sure that an event is not changed elsewhere.
* xterm.c (handle_one_xevent): Change 2nd arg to 'const XEvent *
const' and do not create local copy except for the KeyPress event.
Use casts to avoid a few glitches. Adjust formatting. Add comments.
(SET_SAVED_BUTTON_EVENT): Remove and move the code to the only user.
(x_handle_net_wm_state, x_menubar_window_to_frame)
(x_detect_focus_change, construct_mouse_click, note_mouse_movement)
(x_scroll_bar_to_input_event, x_scroll_bar_expose)
(x_scroll_bar_handle_click, x_scroll_bar_note_movement):
* gtkutil.c (xg_event_is_for_menubar, xg_event_is_for_scrollbar):
* xselect.c (x_handle_property_notify, x_handle_selection_notify)
(x_handle_dnd_message):
* xsettings.c (xft_settings_event):
Use 'const XEvent * const' where appropriate.
* xterm.h, gtkutil.h, xsettngs.h: Adjust related prototypes.
* frame.h (struct frame) [HAVE_X_WINDOWS]: New member wait_event_type.
* xterm.c (pending_event_wait): Remove. Adjust users.
(x_detect_focus_change): Pass frame arg.
(handle_one_xevent): Find related frame early and clear per-frame
wait_event_type only if this is an event for the relevant frame.
(x_wait_for_event): Use per-frame wait_event_type.
