* src/xdisp.c (Fmove_point_visually): When lines are truncated,
simulate display in a window of infinite width, to allow move_it_*
functions reach positions outside of normal window dimensions.
Remove code that tried to handle a subset of these situations by
manual iteration of buffer text. (Bug#17777)
Crowe's patch in:
https://lists.gnu.org/archive/html/emacs-devel/2015-07/msg00106.html
A problem in this area has been reported by several users; see
Bug#16737, Bug#17101, Bug#17026, Bug#17172, Bug#19320, Bug#20283.
This fix differs from Mike Crowe's patch in that it should avoid a
race condition that could lose SIGIO signals. ignore_sigio dates
back to the 1980s when some platforms couldn't block signals, and
could only ignore them, which led to races when signals arrived
while being ignored. We shouldn't have to worry about those old
platforms now.
* src/dispextern.h, src/sysdep.c (ignore_sigio): Remove.
* src/emacs.c (shut_down_emacs):
Don't call ignore_sigio; unrequest_sigio should suffice.
* src/keyboard.c (kbd_buffer_store_buffered_event):
Use unrequest_sigio, not ignore_sigio.
(kbd_buffer_get_event):
Call request_sigio when getting the ball rolling again.
Improve the heuristic for distinguishing stack overflows from
other SIGSEGV causes (Bug#21004). Corinna Vinschen explained that
the getrlimit method wasn't portable to Cygwin; see:
https://www.cygwin.com/ml/cygwin/2015-07/msg00092.html
Corinna suggested pthread_getattr_np but this also has problems.
Instead, replace the low-level system stuff with a simple
heuristic based on known good stack addresses.
* src/eval.c, src/lisp.h (near_C_stack_top): New function.
* src/sysdep.c: Don't include <sys/resource.h>.
(stack_direction): Remove. All uses removed.
(stack_overflow): New function.
(handle_sigsegv): Use it instead of incorrect getrlimit heuristic.
Make SEGV fatal in non-main threads.
After stack overflow, command_loop calls init_eval, and this needs to
clear gcprolist and byte_stack_list (Bug#20996).
* src/alloc.c (init_alloc):
Move gcprolist and byte_stack_list initialization from here ...
* src/eval.c (init_eval): ... to here.
* src/macfont.m (macfont_family_cache): New variable.
(syms_of_macfont): Initialize it.
(macfont_available_families_cache): New variable.
(macfont_invalidate_family_cache, macfont_get_family_cache_if_present)
(macfont_set_family_cache, macfont_invalidate_available_families_cache)
(macfont_handle_font_change_notification)
(macfont_init_font_change_handler)
(macfont_copy_available_families_cache): New functions.
(macfont_create_family_with_symbol): Use font family caches.
(macfont_list, macfont_list_family): Use
macfont_copy_available_families_cache instead of
mac_font_create_available_families.
* src/process.c (wait_reading_process_output): Extend the behavior of
not breaking due to not finding output when a timer has lowered the
timeout to include when SIGIO lowers the timeout.
* src/process.c (wait_reading_process_output):
The patch for Bug#17647 returns too fast sometimes when reading
from any processes. Revert part of it, and limit the timeout more
sensibly (Bug#20978).
* src/process.c (wait_reading_process_output): Rename inner nsecs to
adaptive_nsecs. There is already an nsecs, and this function is
confusing enough (Bug#20978).
* src/process.c (wait_reading_process_output, status_notify):
Previously the function wait_reading_process_input was renamed to the
more logical wait_reading_process_output. Make its local variables
consistent with that change (Bug#20978).
* src/process.c (make-process, make-pipe-process, deactivate_process)
(wait_reading_process_output, read_process_output, send_process)
(init_process_emacs): ifdef ADAPTIVE_READ_BUFFERING was originally
added in case there was an operating system in which it was not
useful. That was 11 years ago and it hasn't happened. Make
development easier by not considering the effect of changes on a
theoretical OS where this is disabled (Bug#20978).
* src/w32fns.c (w32_wnd_proc): Don't change the mouse pointer
shape while a menu is in use. This started happening since we now
send WM_EMACS_SHOWCURSOR messages when the mouse moves.
* src/frame.c (x_set_font): If font_spec_from_name returns nil,
don't barf; instead, request a new fontset to be generated. This
avoids unnecessarily rejecting fonts named against XLFD rules. See
http://lists.gnu.org/archive/html/help-emacs-windows/2015-06/msg00001.html,
for the description of the original problem.
* lisp/faces.el (set-face-attribute): Don't be fooled too easily
by a hyphen in a font's name.
* src/keyboard.c (Fposn_at_x_y, Fposn_at_point): Allow X pixel
coordinate of -1, for a newline in a right-to-left line that
overflowed into the left fringe.
* etc/NEWS:
* doc/lispref/frames.texi (Mouse Tracking): Document the special
effect of setting 'track-mouse' to 'dragging'.
* lisp/textmodes/artist.el (artist-mouse-draw-continously):
* lisp/ruler-mode.el (ruler-mode-mouse-drag-any-column-iteration):
* lisp/mouse-drag.el (mouse-drag-throw):
* lisp/mouse.el (mouse-drag-line): Set 'track-mouse' to 'dragging'
to avoid changes in the shape of the mouse pointer.
* src/xdisp.c (define_frame_cursor1): Don't change the mouse
pointer shape when do_mouse_tracking has the value of 'dragging',
not just any non-nil value. (Bug#20934)
(syms_of_xdisp): DEFSYM 'dragging'.
* src/gnutls.c (Fgnutls_boot): Use the %x conversion specifier in
the call to 'error', instead of the unsupported %u. Reported by
lo2net <fangtao0901@gmail.com>. (Bug#20908)
* src/w32fns.c: Include windowsx.h.
(w32_wnd_proc): If the mouse moved and the mouse pointer is
invisible, make it visible again even when the main (Lisp)
thread is busy.
* src/w32term.c (w32_toggle_invisible_pointer): Rather then
garbaging the frame have the input thread call SetCursor.
* src/w32fns.c (w32_wnd_proc): Handle f->pointer_invisible
for WM_SETCURSOR and WM_EMACS_SETCURSOR cases.
* src/w32term.c (w32_hide_hourglass): Handle
f->pointer_invisible.
(w32_toggle_invisible_pointer): New function.
(w32_create_terminal): Add w32_toggle_invisible_pointer as
toggle_invisible_pointer_hook for this terminal.
* src/Makefile.in ($(libsrc)/make-docfile$(EXEEXT)):
Depend on $(lib)/libgnu.a, so that we build $(lib)/*/*.h
before the submake in $(libsrc) would spin off a subsubmake
for $(lib) in parallel with our submake for $(lib) (Bug#20894).
* doc/lispref/help.texi (Keys in Documentation):
* lisp/international/mule-cmds.el (set-locale-environment):
* lisp/term/w32console.el (terminal-init-w32console):
* src/doc.c (Fsubstitute_command_keys, Vhelp_quote_translation):
If ‘ is not displayable, transliterate it to `, not to '. See:
http://lists.gnu.org/archive/html/emacs-devel/2015-06/msg00542.html
Without this fix, lib/fcntl.h isn't built in time (Bug#20894).
* lib-src/Makefile.in (../lib/libgnu.a):
* src/Makefile.in ($(lib)/libgnu.a): Build all, not libgnu.a.
Use the same alignment for the !USE_LSB_TAG case as for the
more-typical USE_LSB_TAG case. The attempt to support arbitrary
alignments with !USE_LSB_TAG had subtle bugs in garbage collection
once we changed the representation of symbols so that Qnil == 0.
Problem reported by Eli Zaretskii (Bug#20862).
* src/alloc.c (XMALLOC_HEADER_ALIGNMENT) [XMALLOC_OVERRUN_CHECK]:
* src/alloc.c (vector_alignment, union aligned_Lisp_Symbol)
(union aligned_Lisp_Misc, maybe_lisp_pointer, pure_alloc):
Use same alignment for !USE_LSB_TAG as for USE_LSB_TAG.
* src/alloc.c (POINTERS_MIGHT_HIDE_IN_OBJECTS): Remove.
This optimization in the !USE_LSB_TAG case is no longer valid when
symbols are represented via offsets. Change the only use to
assume that pointers might hide in objects.
* src/lisp.h (alignas) [!USE_LSB_TAG]:
Require support in this case, too.
(TAG_SYMOFFSET, XSYMBOL) [!USE_LSB_TAG]: Do not shift the offset.
This is OK, because the !USE_LSB_TAG case now applies only when
Lisp_Object is wider than void *, so there's no longer any need
to shift the offset. Not shifting the offset means that
symbol representations have the same alignment as pointers,
which the GC assumes.
On some optimizing C compilers, copying a structure did not
copy the padding bytes between elements, and the type punning
between struct input_data and struct selection_input_data did
not work. Change the C code to use a proper union type instead.
Problem reported by YAMAMOTO Mitsuharu (Bug#20756).
* src/keyboard.c (kbd_buffer, kbd_fetch_ptr, kbd_store_ptr)
(readable_events, discard_mouse_events, kbd_buffer_events_waiting)
(kbd_buffer_get_event, process_special_events, stuff_buffered_input)
(mark_kboards):
Use union buffered_input_event, not struct input_event.
(clear_event, deliver_input_available_signal, process_special_events):
Remove unnecessary forward decls.
(kbd_buffer_store_buffered_event): New function, mostly just the
old kbd_buffer_store_event_hold, except its argument is of type
union buffered_input_event, not struct input_event.
(kbd_buffer_unget_event): Define only if HAVE_X11, since it's
not needed otherwise. Argument is now of type
struct selection_input_event *, not struct input_event *.
All callers changed.
(clear_event): Arg is now of type union buffered_input_event *,
not struct input_event *. All callers changed.
* src/keyboard.h [HAVE_X11]: Include "xterm.h".
(union buffered_input_event): New type.
(kbd_buffer_store_event_hold): Now an inline function,
defined here.
* src/termhooks.h (EVENT_KIND_WIDTH): New constant.
(struct input_event): Use it.
* src/xselect.c (struct selection_event_queue):
Make elements be of type struct selection_input_event,
not struct input_event.
(selection_input_event_equal): New static function.
(x_queue_event): Use it.
(x_queue_event, x_decline_selection_request)
(x_selection_current_request, x_reply_selection_request)
(x_handle_selection_request, x_handle_selection_clear)
(x_handle_selection_event): Use struct selection_input_event,
not struct input_event. All callers changed.
(x_convert_selection): Omit unused first arg. All callers changed.
(Fx_disown_selection_internal): Omit unnecessary union.
* src/xterm.c (handle_one_xevent): Use new union buffered_input_event
rather than rolling our own equivalent. Prefer sie.kind when
setting up that kind of structure.
Call kbd_buffer_store_buffered_event, not kbd_buffer_store_event_hold.
* src/xterm.h (struct selection_input_event: Use EVENT_KIND_WIDTH.
(SELECTION_EVENT_DISPLAY, SELECTION_EVENT_DPYINFO)
(SELECTION_EVENT_REQUESTOR, SELECTION_EVENT_SELECTION)
(SELECTION_EVENT_TARGET, SELECTION_EVENT_PROPERTY)
(SELECTION_EVENT_TIME, x_handle_selection_event):
Arg is now of type struct selection_input_event *)
not struct input_event *. All callers changed.
