* src/w32heap.c (mmap_realloc): Only attempt extending a region if
the following region has the same allocation base. Also, use the
original allocation base and enlarged size to commit reserved
memory, to ensure that the allocation base stays at its original
value. This fixes several hard-to-debug problems whereby part of
buffer text was overwritten with binary nulls, because
mmap_realloc copied only part of buffer text when extending it.
See
http://lists.gnu.org/archive/html/emacs-devel/2016-04/msg00325.html
and http://debbugs.gnu.org/cgi/bugreport.cgi?bug=23223#55 for two
examples of the related problems.
Problem reported by Clément Pit--Claudel (Bug#23254).
* src/doc.c: Include intervals.h.
(Fsubstitute_command_keys): If the only substitutions are for
quotes, copy the source string’s text properties too, since no
substring lengths have changed.
* doc/lispref/os.texi (Time Zone Rules):
New section, mostly with material moved here from other sections.
* doc/emacs/cmdargs.texi (General Variables):
* doc/lispref/os.texi (Time Conversion, Time Parsing):
Xref new section.
* etc/NEWS, etc/PROBLEMS:
* lisp/org/org.el (org-timestamp-format):
* src/editfns.c (Fformat_time_string, Fdecode_time)
(Fencode_time, Fcurrent_time_string, Fcurrent_time_zone)
(Fset_time_zone_rule):
When documenting time zone rule strings, mention the TZ
environment variable in preference to mentioning the
sort-of-internal function set-time-zone-rule.
* src/xwidget.c (x_draw_xwidget_glyph_string): Use window_box
instead of calculating the clipping borders manually. Suggested
by YAMAMOTO Mitsuharu <mituharu@math.s.chiba-u.ac.jp>.
The arguments BEG and END were unused, and are now removed.
* display.texi (Xwidgets): Document the change
* xwidget.el (make-xwidget, xwidget-insert)
(xwidget-webkit-new-session): Reflect changed arguments
* xwidget.c (Fmake_xwidget, syms_of_xwidget): Reflect changed arguments
* src/xdisp.c (append_stretch_glyph, produce_xwidget_glyph)
(produce_image_glyph): Limit the pixel width of the produced glyph
to SHRT_MAX. (Bug#23178)
(append_composite_glyph, append_glyph, append_glyphless_glyph):
Add assertions to verify that the pixel width of the glyph will
never overflow a 'short'.
* src/term.c (append_composite_glyph): Add assertion to verify
that the pixel width of the glyph will never overflow a 'short'.
* src/xdisp.c (redisplay_window): Bind inhibit-redisplay non-nil
around the call to x_consider_frame_title, to prevent
resize_mini_window from undoing echo-area display. (Bug#23124)
* src/keyboard.c (requeued_events_pending_p, read_char)
(Finput_pending_p): Use CONSP instead of !NILP to see if there are
unread command events to be processed. (Bug#22976)
* src/xdisp.c (syms_of_xdisp) <truncate-partial-width-windows>:
Clarify in the doc string how the width of partial-width windows
is computed for the purposes of truncation decision. (Bug#4338)
* doc/emacs/windows.texi (Split Window): Clarify how the width of
windows is calculated for the purposes of truncation decision.
* doc/emacs/display.texi (Line Truncation): Remove a redundant
index entry.
* src/frame.c (DEFAULT_ROWS): Enlarge to 36, so that the initial
window displayed by "emacs -q" has enough space to show the whole
text even if it includes 2 lines talking about recovering crashes
sessions. (Bug#23074)
* lisp/startup.el (use-fancy-splash-screens-p): Fix off-by-one
error when computing the window-height from frame-height.
* etc/NEWS: Mention the change.
Problem reported by Mosè Giordano (Bug#18997).
* src/gtkutil.c (xg_get_pixbuf_from_pixmap): Remove.
(xg_get_pixbuf_from_pix_and_mask): Do not use
xg_get_pixbuf_from_pixmap, as it is poorly documented. Instead,
invoke XGetPixel directly. This is slow but speed is not
important here. Also, fail for unusual situations (not TrueColor,
or images that are not 8 bits per sample) instead of displaying
junk or crashing.
* src/alloc.c (make_save_ptr_ptr): Remove the !(defined
USE_X_TOOLKIT || defined USE_GTK) conditional. Reported by
Philipp Stephani <phst@google.com>. (Bug#23101)
* src/w32font.c (w32font_list_internal): Allow 'ascii-0' charset,
in addition to 'iso10646-1', 'unicode-bmp', and 'unicode-sip'.
This avoids rejecting many font families whose members are shown
by 'font-family-list', in particular 'courier' requested by
info.el. Without this change, many values of ':family' attribute
of a face have no effect on MS-Windows, because they are rejected
due to bogus mismatch of the charset.
* src/nsterm.m (ns_screen_margins): New function.
(ns_screen_margins_ignoring_hidden_dock): New function.
(ns_menu_bar_height): Reimplement in terms of `ns_screen_margins'.
([EmacsWindow zoom:]): Take all screen margins (except those
originating from a hidden dock) into account.
* src/xdisp.c (get_next_display_element): Handle the case when a
display string acquires the box face from an underlying string,
not from the buffer. (Bug#23091)
* src/bidi.c (bidi_copy_it):
* src/lisp.h (CHAR_TABLE_EXTRA_SLOTS):
Use only a single identifier as the second argument of offsetof.
Found by using clang -pedantic.
* src/gtkutil.c (xg_create_frame_widgets, xg_toggle_notify_cb):
Cast from function type to void * where the C standard requires this.
This works around a problem in the prototypes for
g_signal_handler_find and g_signal_handlers_block_by_func, which
use gpointer instead of GCallback. Found by using gcc -pedantic.
* src/emacsgtkfixed.c: Use workaround for GNOME bug 683906 only
in glib 2.35.6 and earlier, since the bug is fixed in 2.35.7.
* src/emacsgtkfixed.c (EmacsFixedPrivate):
* src/emacsgtkfixed.h (EmacsFixedClass):
Remove duplicate typedef, which strict C99 does not allow (Bug#23003).
* src/nsterm.m (EmacsBell): Save NSView when displaying the
visible bell and set `needsDisplay' when removed.
(hide_bell): Trace.
(ns_copy_bits): Trace.
* configure.ac: Use AS_IF so that gl_WARN_ADD’s prerequisites are
not done conditionally. This helps clang, which needs
-Wunknown-warning-option later when configured with warnings.
* src/editfns.c (invalid_time): Now _Noreturn, since clang isn’t
smart enough to figure this out on its own if warnings are enabled.
(lisp_time_struct): Redo for clarity, and to pacify clang.
* src/xfns.c (x_real_pos_and_offsets) [USE_XCB]: Don’t use
uninitialized locals. This avoids undefined behavior and pacifies
clang.
* src/xdisp.c (syms_of_xdisp) <resize-mini-windows>: Initialize to
nil.
* lisp/loadup.el <resize-mini-windows>: Set to 'grow-only' after
loading window.el. (Bug#22975)
* src/xdisp.c (syms_of_xdisp) <redisplay--inhibit-bidi>: New
boolean variable.
(init_iterator, reseat_to_string)
(Fcurrent_bidi_paragraph_direction)
(Fbidi_find_overridden_directionality): Use
redisplay--inhibit-bidi instead of purify-flag, to determine when
it's safe to reorder bidirectional text.
* lisp/loadup.el (redisplay--inhibit-bidi): Set to t at the
beginning of the file. Reset to nil when charprop.el is
successfully loaded, or when we are going to dump, whichever
happens last. (Bug#22975)
* src/xterm.h (struct x_display_info): Add fields to save XRANDR
version number.
* src/xfns.c (x_get_monitor_attributes): Save the version numbers
after querying the X server.
(x_get_monitor_attributes_xrandr): Don't use XRRGetOutputPrimary or
XRRGetScreenResourcesCurrent if the server doesn't support at least
RANDR version 1.3. Conditionalize the code blocks on compiling
against library version 1.3 or better, rather than feature tests for
each function.
* configure.ac: Stop testing for those two functions.
Work around Bug#22884 by rewording comments and strings to avoid ‘(’
at the start of a line unless it starts a function. This change
is a short-term hack; in the longer run we plan to fix cc-mode’s
performance for C files that have ‘(’ at the start of a line in a
comment or string.
* src/bidi.c (bidi_resolve_weak): Avoid entering a loop searching
for a character needed for resolving the type of a series of BN
and ET characters, as required by rule W5 of UAX#9, if the results
of the resolution are known in advance, because we are at level
zero, and the previous strong character was L.
(bidi_resolve_neutral): Partially resurrect the optimization for a
long series of control characters in an otherwise strictly L2R
text.
(bidi_level_of_next_char): Don't enter the loop that searches for
a paragraph separator if the current character is already at base
embedding level. (Bug#22739)
* src/bidi.c (bidi_resolve_explicit): Override the orig_type value
of FSI with either LRI or RLI, as determined by the first strong
directional character in the isolate. This prevents failure to
isolate when the FSI...PDI text is inside a directional override.
(Bug#22786)
