GitLab

* frame.c (x_set_frame_parameters): Calculate default values of
new frame sizes only after all other frame parameters have been
processed (Bug#17142).

 src/w32term.c (x_draw_image_glyph_string): Fix computation of height
 and width of image background when it is displayed with a 'box'
 face.

* editfns.c (Fchar_equal): In unibyte buffers, assume values in
range 128-255 are raw bytes.  Suggested by Eli Zaretskii.

Fixes: debbugs:17011

* editfns.c (Fchar_equal): Do not use MAKE_CHAR_MULTIBYTE in
unibyte buffers, as we can't tell whether the characters are
actually unibyte.

Fixes: debbugs:17011

* simple.el (primitive-undo): Only process marker adjustments
validated against their corresponding (TEXT . POS).  Issue warning
for lone marker adjustments in undo history.  (Bug#16818)
(undo-make-selective-list): Add marker adjustments to selective
undo list based on whether their corresponding (TEXT . POS) is in
the region.  Remove variable adjusted-markers, which was unused
and only non nil during undo-make-selective-list.
(undo-elt-in-region): Return nil when passed a marker adjustment
and explain in function doc.

Have (MARKER . ADJUSTMENT) undo records always be immediately
after their corresponding (TEXT . POS) record in undo list.
(Bug#16818)
* lisp.h (record-delete): New arg record_markers.
(record_marker_adjustment): No longer needed outside undo.c.
* insdel.c (adjust_markers_for_delete): Move calculation of marker
adjustments to undo.c's record_marker_adjustments.  Note that
fileio.c's decide_coding_unwind is another caller to
adjust_markers_for_delete.  Because it has undo list bound to t,
it does not rely on adjust_markers_for_delete to record marker
adjustments.
(del_range_2): Swap call to record_delete and
adjust_markers_for_delete so as undo marker adjustments are
recorded before current deletion's adjustments, as before.
(adjust_after_replace):
(replace_range): Pass value for new record_markers arg to
delete_record.
* undo.c (record_marker_adjustment): Renamed to
record_marker_adjustments and made static.
(record_delete): Check record_markers arg and call
record_marker_adjustments.
(record_change): Pass value for new record_markers arg to
delete_record.
(record_point): at_boundary calculation no longer needs to account
for marker adjustments.

* undo-tests.el (undo-test-marker-adjustment-nominal):
(undo-test-region-t-marker): New tests of marker adjustments.
(undo-test-marker-adjustment-moved):
(undo-test-region-mark-adjustment): New tests to demonstrate
bug#16818, which fail without the fix.

* markers.texi (Moving Marker Positions): The 2014-03-02 doc
change mentioning undo's inability to handle relocated markers no
longer applies.  See bug#16818.
* text.texi (Undo): Expand documentation of (TEXT . POS) and
(MARKER . ADJUSTMENT) undo elements.

* w32term.c (x_set_window_size): Refine fix from 2014-03-14
(Bug#17077).

Fixes: debbugs:17073

overlay to make sure boundaries are visible.
* src/buffer.c (struct sortvec): Add field `spriority'.
(compare_overlays): Use it.
(sort_overlays): Set it.

Fixes: debbugs:15899

 src/xdisp.c (redisplay_window): If all previous attempts to find the
 cursor row failed, try a few alternatives before falling back to
 the top-most row of the window.  Use row_containing_pos.

* xterm.c (x_draw_image_relief): Respect Vtool_bar_button_margin.
* w32term.c (x_draw_image_relief): Likewise.

* w32fns.c (w32_wnd_proc): For WM_WINDOWPOSCHANGING don't
constrain frame size in SW_SHOWMAXIMIZED case so we can truly
maximize a frame for odd default fonts.
* frames.texi (Size and Position): In `frame-resize-pixelwise'
description drop remark about frame maximization.
* windows.texi (Display Action Functions): Add description for
`display-buffer-no-window' and explain use of `allow-no-window'
alist entries.

* doc.c (store_function_docstring): Fix pointer signedness mismatch.

the docstring of functions advised before dumping.
* src/doc.c (store_function_docstring): Warn when we don't know where to
put a docstring.
(Fsubstitute_command_keys): Don't advertise the fact that
text-properties are dropped, since we think it's a bug that we'll fix
in 24.5.

Fixes: debbugs:16993

* src/xterm.c (handle_one_xevent) <MapNotify>: Don't garbage the frame.
* src/frame.c (frame_garbaged): Make "docstring" more precise.

mention if EMACSDATA is set.

* doc/lispref/numbers.texi (Numeric Conversions, Rounding Operations):
Document that 'round' and 'fround' round to even.
* src/floatfns.c (emacs_rint) [!HAVE_RINT]: Round to even.
This way, the unusual !HAVE_RINT case acts like the usual
HAVE_RINT case, and we can fix the documentation accordingly.

 src/w32fns.c (reset_modifiers): Zero out keystate[] before using it.
 (w32_wnd_proc): Initialize the dwHoverTime member of
 TRACKMOUSEEVENT structure.

 src/search.c (find_newline): Speed up the function when using the
 newline cache, by halving the number of calls to
 region_cache_forward and region_cache_backward.

* w32term.c (x_set_window_size): When frame-resize-pixelwise is
nil, always resize character wise to avoid potential loss of the
mode line (Bug#16923 related).
* display.texi (Temporary Displays): Say that
with-temp-buffer-window makes its buffer current.
* frames.texi (Size and Position): Describe new option
`frame-resize-pixelwise'.  Rewrite descriptions of
`set-frame-size', `set-frame-height' and `set-frame-width'.

* frame.c (x_set_frame_parameters): Always calculate new sizes
pixelwise to avoid potential loss when rounding.
* window.el (fit-frame-to-buffer): Get maximum width from
display's width instead of height.

* xterm.c (x_free_frame_resources): Free all allocated cursors.

 src/w32.c (fstatat): Don't add an extra slash if the argument ends
 with a slash: this fails the subsequent call to stat_worker on
 Windows 9X.  Reported by oslsachem <oslsachem@gmail.com>.

* w32term.c (w32_read_socket): In SIZE_RESTORED case
unconditionally set visibility of frame to 1.

* xdisp.c (Fwindow_text_pixel_size): Adjust doc-string.
* elisp.texi (Top): Rename section "Width" to "Size of Displayed
Text".
* text.texi (Primitive Indent):
* strings.texi (String Basics):
* sequences.texi (Sequence Functions): Update references
accordingly.
* display.texi (Size of Displayed Text): Rename section from
"Width".  Add description for `window-text-pixel-size'.
(Window Dividers): Reword description of window dividers.
* frames.texi (Layout Parameters): Improve description of window
divider parameters.
* windows.texi (Window Sizes): Add descriptions of
`window-mode-line-height' and `window-header-line-height'.
(Coordinates and Windows): Mention window dividers.

* nsterm.m (constrainFrameRect:toScreen:): Constrain normally when frame
is only on one screen.

Fixes: debbugs:14713

Fixes: debbugs:16961

 src/xdisp.c (move_it_in_display_line_to): If word-wrap is ON, and
 there's a valid wrap point in the display line, the last glyph
 cannot "just barely fit" on this row, because display_line doesn't
 let it.  Instead, proceed as if the last glyph didn't fit, so that
 we eventually back up the iterator to the wrap point.  This avoids
 delusional behavior of move_it_to, whereby it proceeds to the next
 display line, but sets current_x to zero for all the glyphs that
 without word-wrap would fit on the previous display line.  One
 result was that visual-order cursor movement behaved erratically
 under word-wrap.
 (Fmove_point_visually): Add code to find the x coordinate of the
 last character before wrap point, under word-wrap on a TTY.