Paul Eggert [Tue, 20 Aug 2013 00:51:35 +0000 (17:51 -0700)]
* image.c (SIGNATURE_DIGESTSIZE): Remove.
(struct animation_cache): Make signature a flexible array member.
All uses changed. This is a tad slower but may insulate us better
from future changes to ImageMagick.
Paul Eggert [Mon, 19 Aug 2013 20:47:27 +0000 (13:47 -0700)]
* image.c: Shrink memory needed for animation cache.
(SIGNATURE_DIGESTSIZE): New constant.
(struct animation_cache): Make 'signature' a fixed size array of bytes.
(imagemagick_create_cache): Copy the signature. This saves
several KB of memory that ImageMagick wastes per signature.
Don't bother updating the update_time, as the caller does that now.
(imagemagick_prune_animation_cache): Don't destroy the signature, as
it's a fixed size struct member now.
(imagemagick_get_animation_cache): Always destroy the signature,
as it's now imagemagick_create_cache's responsibility to copy it.
Avoid duplicate calls to strcmp and to imagemagick_create_cache,
and use memcmp rather than strcmp.
eassert that ImageMagick returns a signature of the specified length.
Fix some other minor performance problems while we're at it.
(imagemagick_create_cache): Clear just the members that
need clearing. Don't set update_time, as caller does that now.
(imagemagick_prune_animation_cache, imagemagick_get_animation_cache):
Simplify by using pointer-to-pointer instead of a prev pointer.
(imagemagick_prune_animation_cache): Use make_emacs_time rather
than EMACS_TIME_FROM_DOUBLE, and DestroyString rather than free.
(imagemagick_get_animation_cache): Don't xstrdup the image signature;
it's already a copy. Free the signature probe unless it's cached.
Ken Brown [Sun, 18 Aug 2013 17:37:31 +0000 (13:37 -0400)]
Change size of static heap on Cygwin; report maximum usage after dumping.
* src/sheap.c (STATIC_HEAP_SIZE): Adjust to current needs; use bigger
static heap if ENABLE_CHECKING is defined.
(max_bss_sbrk_ptr): New variable.
(bss_sbrk): Use it.
(report_sheap_usage): Report maximum static heap usage instead of
ending static heap usage.
Stefan Monnier [Thu, 15 Aug 2013 17:21:19 +0000 (13:21 -0400)]
* lisp/emacs-lisp/debug.el (debugger-setup-buffer): Put point on the
previous line.
(debugger-eval-expression, debugger-record-expression):
Use read--expression (bug#15102).
Paul Eggert [Thu, 15 Aug 2013 16:28:42 +0000 (09:28 -0700)]
Fix minor problems found by static checking.
* frame.c (delete_frame):
* xdisp.c (next_element_from_display_vector):
Avoid uninitialized local.
* image.c (imagemagick_compute_animated_image): Port to C89.
Prefer usual GNU indentation style for loops.
Be more careful about bizarrely large sizes, by using ptrdiff_t
instead of int.
(imagemagick_compute_animated_image): Implement a simple cache
(imagemagick_compute_animated_image): Fix some compilation
warnings. Implement a very simple cache to make the animation
usable at all, but it should be replaced with a per-image cache.
Eli Zaretskii [Thu, 15 Aug 2013 15:20:03 +0000 (18:20 +0300)]
Fix bug #15090 with redisplay under linum-mode and visual-line-mode.
src/xdisp.c (compute_window_start_on_continuation_line): When
WORD_WRAP is in effect, use move_it_to instead of move_it_by_lines
to make sure we end up setting the window start at the leftmost
visible character of the display line. This avoids funky
horizontal shifting because the window start is not kept on the
same position.
Dmitry Antipov [Thu, 15 Aug 2013 14:52:53 +0000 (18:52 +0400)]
* lisp.h (FOR_EACH_ALIST_VALUE): New macro
to do `for' loops over alist values.
* buffer.h (FOR_EACH_BUFFER):
* process.c (FOR_EACH_PROCESS): Use it.
(handle_child_signal, status_notify, Fget_buffer_process)
(kill_buffer_processes): Use FOR_EACH_PROCESS.
Dmitry Antipov [Thu, 15 Aug 2013 05:23:40 +0000 (09:23 +0400)]
* term.c (get_named_tty, create_tty_output, tty_free_frame_resources)
(tty_free_frame_resources, delete_tty): Prefer eassert to emacs_abort.
* image.c (make_image_cache): For struct image_cache, prefer xmalloc
to xzalloc and so avoid redundant call to memset.
* xterm.c (x_term_init): Avoid unnecessary initializations of dpyinfo
members because it is allocated with xzalloc and so already zeroed.
Dmitry Antipov [Wed, 14 Aug 2013 16:36:16 +0000 (20:36 +0400)]
Utility function and macro to copy Lisp string to C string.
* lisp.h (xlispstrdupa): New macro.
(xlispstrdup): New prototype.
* alloc.c (xlispstrdup): New function.
* callint.c (Fcall_interactively):
* fileio.c (Ffile_name_directory, Fexpand_file_name)
(Fsubstitute_in_file_name):
* frame.c (Fmake_terminal_frame): Use xlispstrdupa.
* image.c (x_create_bitmap_from_file):
* w32term.c (w32_term_init):
* xterm.c (x_term_init): Use xlispstrdup.
lisp/frameset.el: Allow orphaned minibufferless frames.
(frameset-filter-minibuffer): Deal with the case that the minibuffer
parameter was already set in FILTERED. Doc fix.
(frameset--record-minibuffer-relationships): Allow saving a
minibufferless frame without its corresponding minibuffer frame.
(frameset--reuse-frame): Accept a match from an orphaned minibufferless
frame, if the frame id matches.
(frameset--minibufferless-last-p): Sort non-orphaned minibufferless
frames before orphaned ones.
(frameset-restore): Warn about orphaned windows, instead of error out.
Martin Rudalics [Wed, 14 Aug 2013 09:14:32 +0000 (11:14 +0200)]
Fix some issues in window.el.
* window.el (window-make-atom): Don't overwrite parameter
already present.
(display-buffer-in-atom-window): Handle special case where we
split an already atomic window.
(window--major-non-side-window, display-buffer-in-side-window)
(window--side-check): Ignore minibuffer window when walking
window tree.
(window-deletable-p): Return 'frame only if no other frame uses
our minibuffer window.
(record-window-buffer): Run buffer-list-update-hook.
(split-window): Make sure window--check-frame won't destroy an
existing atomic window in case the new window gets nested
inside.
(display-buffer-at-bottom): Ignore minibuffer window when
walking window tree. Don't split a side window.
(pop-to-buffer): Don't set-buffer here, the select-window call
should do that.
(mouse-autoselect-window-select): Autoselect only if we are in the
text portion of the window.
Dmitry Antipov [Wed, 14 Aug 2013 07:25:45 +0000 (11:25 +0400)]
* xdisp.c (redisplay_window): If window_end_valid is cleared due to
non-zero windows_or_buffers_changed, clear current_matrix_up_to_date_p
and so do not call to try_cursor_movement for that window.