This fixes bug #23827.
* src/buffer.c (set-buffer-major-mode): Run `fundamental-mode' when the buffer
gets set to that mode, so that `run-mode-hooks', and thus
`hack-local-variables' get run.
* src/dispextern.h (FACE_OPT_FROM_ID): Don’t use FACE_FROM_ID,
since it is intended to be used only when it returns a non-null
pointer, and here the pointer might be null.
(IMAGE_OPT_FROM_ID): Don’t use IMAGE_FROM_ID, for similar reasons.
* src/lread.c (Fload): Don't overwrite the last character of the
file name in FOUND with 'c', unless the file name ended in ".elc"
to begin with. Don't treat empty files as byte-compiled. See
http://lists.gnu.org/archive/html/emacs-devel/2016-06/msg00463.html
for more details of the problem this caused.
* src/fileio.c (maybe_move_gap): New function to move the gap to
the end of a buffer, if it isn't there already.
(Finsert_file_contents): Call 'maybe_move_gap' before using
conversion_buffer's text as a C 'char' array. (Bug#23659)
* src/coding.c (decode_eol): Compute the byte increment before
calling del_range_2, because the latter can invalidate the pointer
to buffer text.
* src/composite.c (autocmp_chars):
* src/conf_post.h (DebPrint) [HAVE_NTGUI && !DebPrint && !EMACSDEBUG]:
Use simpler ((void) 0) for no-op expression returning void.
* src/dispextern.h [HAVE_WINDOW_SYSTEM]:
Include fontset.h, for face_for_char.
(FACE_SUITABLE_FOR_ASCII_CHAR_P, FACE_FOR_CHAR):
Now inline functions instead of macros. This avoids the need for
all those casts to void.
(FACE_SUITABLE_FOR_ASCII_CHAR_P): Omit 2nd (unused) arg.
All uses changed.
* src/frame.c (Ficonify_frame, Fset_frame_position):
* src/xdisp.c (Fmove_point_visually, show_mouse_face):
* src/xdisp.c (note_mode_line_or_margin_highlight)
(note_mouse_highlight):
Assume HAVE_WINDOW_SYSTEM for simplicity, since the code should
now work either way without generating warnings.
* src/frame.c (display_available) [HAVE_WINDOW_SYSTEM]: New function.
(window_system_available) [HAVE_WINDOW_SYSTEM]: Move to frame.h.
(decode_window_system_frame): Use check_window_system instead of
rolling the code ourself. Return needed only if HAVE_WINDOW_SYSTEM.
(decode_window_system_frame, check_window_system):
Merge the HAVE_WINDOW_SYSTEM and !HAVE_WINDOW_SYSTEM versions into one.
* src/frame.c (Ficonify_frame, Fset_frame_position):
* src/xdisp.c (show_mouse_face, define_frame_cursor1)
(note_mouse_highlight):
Narrow the scope of the HAVE_WINDOW_SYSTEM #ifdef;
this is a better way to pacify GCC.
* src/xdisp.c (x_set_left_fringe, x_set_right_fringe)
(x_set_right_divider_width, x_set_bottom_divider_width):
* src/xfns.c (x_set_internal_border_width):
Don’t use what are now function calls as lvalues.
* src/frame.h (WINDOW_SYSTEM_RETURN): New macro.
(decode_window_system_frame, check_window_system):
Use it, to avoid the need for duplicate declarations.
(window_system_available): Now an inline function.
(display_available): New decl.
(frame_dimension): New inline function.
(FRAME_FRINGE_COLS, FRAME_LEFT_FRINGE_WIDTH)
(FRAME_RIGHT_FRINGE_WIDTH, FRAME_TOTAL_FRINGE_WIDTH)
(FRAME_INTERNAL_BORDER_WIDTH, FRAME_RIGHT_DIVIDER_WIDTH)
(FRAME_BOTTOM_DIVIDER_WIDTH):
Use it, to avoid the need for duplicate definitions.
Now inline functions instead of macros.
* src/gnutls.c (gnutls_log_function2i): Remove.
* src/gnutls.h (GNUTLS_LOG2i): Use ‘message’ directly.
This avoids complaints about gnutls_log_function2i being defined
and not used on older platforms that do not need to call GNUTLS_LOG2i.
* src/image.c (DefaultDepthOfScreen) [0]: Remove unused macro.
* src/lisp.h (AUTO_STRING_WITH_LEN): Revert change from ‘type id =
expr’ to ‘type id; id = expr’, as this would suppress valid
jump-misses-init diagnostics. Let’s find a better way to address
the problem.
* src/vm-limit.c (__MALLOC_HOOK_VOLATILE):
Define only if needed.
* src/xdisp.c (handle_single_display_spec):
Simplify fringe_bitmap computation.
(define_frame_cursor1): Do nothing unless in a window system.
All callers changed and simplified.
* src/xfaces.c (realize_default_face):
Use a simpler way to pacify GCC when a return value is not used
on some platforms.
* src/xdisp.c (move_it_in_display_line_to): Reset the iterator's
'constrain_row_ascent_descent_p' flag after processing the
newline, similar to what 'display_line' does. Without this, the
flag remains set once set by x_produce_glyphs, and causes the
vertical layout calculations to go awry, because Emacs thinks the
screen lines have zero height. (Bug#23850)
* src/conf_post.h (DebPrint) [!EMACSDEBUG]: Add empty do-while
loop to avoid compiler warnings about empty body of 'if'
statement.
* src/image.c [HAVE_NTGUI] (DefaultDepthOfScreen): Disable unused
macro.
(x_create_bitmap_from_data): Declare 'frame'.
(x_create_bitmap_from_file): Don't declare unused variable
'dpyinfo'.
* src/lisp.h (AUTO_STRING_WITH_LEN): Avoid initialization to
prevent "jump-misses-init" compiler warnings.
* src/w32fns.c (check_w32_winkey_state): Define and use only if
WINDOWSNT.
(Fx_show_tip): Declare 'f'.
(Fx_file_dialog): Declare 'filter_a' only if not NTGUI_UNICODE.
(w32_strerror): Use format specifier %d for sprintf argument of
type int.
(emacs_abort): Cast sprintf argument of type DWORD to unsigned
int, and use format specifier %x, for compatibility with Cygwin.
(unicode_append_menu) [NTGUI_UNICODE]:
* src/w32menu.c [NTGUI_UNICODE] (get_menu_item_info)
(set_menu_item_info, unicode_append_menu)
(unicode_message_box):
* src/menu.c [NTGUI_UNICODE] (unicode_append_menu): Define as
functions rather than macros to avoid "address will always
evaluate as true" compiler warnings.
* src/w32font.c (w32_to_x_charset): Use format specifier %d for
sprintf argument of type int.
* src/w32term.c (x_draw_glyphless_glyph_string_foreground): Cast
sprintf argument of type int to unsigned int to match %X format
specifier.
(w32_scroll_bar_handle_click):
(w32_horizontal_scroll_bar_handle_click): Declare 'f'.
* src/w32term.h (FRAME_DISPLAY_INFO): Explicitly discard unused
argument.
* src/composite.c (autocmp_chars): Declare and set 'f' only if
HAVE_WINDOW_SYSTEM.
* src/dispextern.h (FACE_SUITABLE_FOR_ASCII_CHAR_P)
(FACE_FOR_CHAR): Explicitly discard unused arguments.
* src/font.c (font_open_entity): Declare 'min_width' where it is
first set, and only if HAVE_WINDOW_SYSTEM.
* src/frame.c [!HAVE_WINDOW_SYSTEM] (decode_window_system_frame):
Define _Noreturn version to avoid "suggest attribute noreturn"
compiler warning.
(check_window_system): Declare as _Noreturn.
(Ficonify_frame):
(Fset_frame_position): Don’t declare and set 'f'.
* src/frame.h [!HAVE_WINDOW_SYSTEM] (decode_window_system_frame)
(check_window_system): Add _Noreturn prototypes.
(FRAME_FRINGE_COLS, FRAME_TOTAL_FRINGE_WIDTH)
(FRAME_LEFT_FRINGE_WIDTH, FRAME_RIGHT_FRINGE_WIDTH)
(FRAME_INTERNAL_BORDER_WIDTH, FRAME_RIGHT_DIVIDER_WIDTH)
(FRAME_BOTTOM_DIVIDER_WIDTH): Explicitly discard unused argument.
* src/xdisp.c (handle_single_display_spec): Declare
'fringe_bitmap' only if HAVE_WINDOW_SYSTEM.
(append_space_for_newline): Declare 'g' where it is first set.
(Fmove_point_visually): Declare and set 'target_is_eol_p' only if
HAVE_WINDOW_SYSTEM.
(show_mouse_face): Declare and set 'f' and 'phys_cursor_on_p' only
if HAVE_WINDOW_SYSTEM.
(note_mode_line_or_margin_highlight):
(note_mouse_highlight): Declare and set 'cursor' and 'pointer'
only if HAVE_WINDOW_SYSTEM.
* src/xfaces.c (realize_default_face): Declare and set 'face' only
if HAVE_X_WINDOWS. Remove redundant #ifdef.
* src/alloc.c (ABLOCKS_BUSY): Rename arg to avoid potential clash
with member ‘abase’ in definiens.
(lisp_align_malloc, lisp_align_free): Use bool for boolean.
Avoid compiler warning with fewer casts.
(lisp_align_free): Check busy-field values; this can help the
compiler a bit when optimizing, too.
* src/chartab.c (char_table_set_range): Start the loop from the
first character of the block to which FROM belongs. (Bug#23797)
* test/src/chartab-tests.el: New test file.
d1efbaf Fix documentation of completion functions
65c96cc Clarify documentation of 'font-lock-maximum-decoration'
2ad3d01 * doc/misc/cl.texi (Usage): Add some more details.
b49cb0a Fbackward_prefix_chars: stay within buffer bounds
The commit 1fd3172d "(Fbackward_prefix_chars): Set point properly while
scanning" (1998-03-18), moved the check against of the position against the
buffer beginning out the loop condition so that we might end up checking
the syntax of characters before the beginning of the buffer. This can
cause segfaults or trigger a "Point before start of properties" error in
`update_interval' (called indirectly from `char_quoted').
* src/syntax.c (Fbackward_prefix_chars): Stop the loop when beginning of
buffer is reached (Bug #3552, Bug #17132, Bug #19379).
This fixes a five-year-old FIXME comment. Although I don’t know
of a platform where this is a problem in practice, better safe
than sorry.
* src/doc.c (Fdocumentation): If SUBRP, simply use doc as integer,
as it is now an integer, not char *.
(store_function_docstring): Offset is now EMACS_INT, not ptrdiff_t;
this is a file offset and EMACS_INT is better if --with-wide-int.
If SUBRP, simply store the offset rather than negating it and
converting it to char *.
* src/lisp.h (struct Lisp_Subr.doc): Now EMACS_INT, not char *.
Suggested by Eli Zaretskii in discussion of bug#18302.
* src/image.c [HAVE_NTGUI]: Remove the unused macro
PIXEL_ALREADY_TYPEDEFED. On Cygwin, include "noX/xpm.h" instead of
"X11/xpm.h".
* configure.ac [CYGWIN && HAVE_W32]: Change the xpm test to use
"noX/xpm.h". Don't add a -I switch to CPPFLAGS.
* src/editfns.c (tzlookup): Call tzset after setting TZ, so that
the setting change propagates into Emacs local time (Bug#23600).
(emacs_setenv_TZ): Clarify comments.
* src/conf_post.h (GNUC_PREREQ): New macro.
Change uses of __GNUC_MINOR__ and __GNUC_PATCHLEVEL__ to use this
macro instead, for clarity and consistency.
(PRINTF_ARCHETYPE): New macro. Define it to __gnu_printf__ only
if glibc, since non-GNU platforms don’t necessarily support GNU
printf formats.
(ATTRIBUTE_FORMAT_PRINTF): Use it.
* src/conf_post.h (ATTRIBUTE_FORMAT_PRINTF) [__MINGW32__]: Use
'__ms_printf__', not '__gnu_printf__', as the latter is not what
MS 'printf' supports, and causes bogus compilation warnings.
* src/lisp.h (pI) [__MINGW32__]: Define to "I64", as MS 'printf'
doesn't support the "ll" modifier.
* src/buffer.c, src/emacs.c, src/lread.c: Don’t include coding.h;
no longer needed, now that emacs_strerror is declared by lisp.h.
* src/coding.c (emacs_strerror): Remove; moved to emacs.c.
* src/coding.h (emacs_strerror) [emacs]: Remove decl; moved
to lisp.h.
* src/emacs.c (emacs_strerror): Move here from coding.c. Do not
convert result string; this is now the caller’s responsibility,
as some need conversion and others don’t.
* src/fileio.c (report_file_errno, report_file_notify_error):
Use emacs_strerror rather than rolling it ourselves.
* src/lisp.h (emacs_strerror): Move decl here from coding.h.
* src/lread.c (dir_warning): Just call emacs_strerror rather than
both strerror and emacs_strerror. Convert its result from
locale-coding-system, since it no longer does that conversion.
* src/sound.c (sound_perror):
* src/sysdep.c (emacs_perror, str_collate):
Use emacs_strerror, not strerror.
* src/dbusbind.c (Fdbus__init_bus): Use make_save_pointer to store
connection address. (Bug#23741)
(xd_lisp_dbus_to_dbus): Use XSAVE_POINTER to retrieve connection
address.
This should help insulate Emacs better from configuration screwups.
Future versions of the GNU C library are planned to deprecate
these functions, but will continue to support them in
already-built-and-dumped Emacs executables.
* src/alloc.c (malloc_initialize_hook) [DOUG_LEA_MALLOC]:
Abort if malloc_set_state fails.
(alloc_unexec_pre) [DOUG_LEA_MALLOC]:
Report malloc_get_state failure, and exit.
* admin/CPP-DEFINES: Remove HAVE_GETPEERNAME.
* configure.ac: Don’t check for getpeername.
* src/process.c (wait_reading_process_output) [!WINDOWSNT]:
Use getsockopt even if not GNU/Linux, as the platforms where
getsockopt used to hang are no doubt long dead.
* src/process.c (decode_status): 3rd arg is now Lisp_Object *,
not int *, and is not decoded. All uses changed.
(status_message): Do not assume ‘failed’ code is an integer.
* src/process.h: Document codes better.
* src/editfns.c (styled_format): Fix copying text properties from
the format specification to the produced string representation.
(Bug#23730)
(Fformat) Doc fix.
* doc/lispref/strings.texi (Formatting Strings): Document that
text properties from the format specifiers are also copied to the
produced string.
Inspired by a suggestion from RMS in: http://bugs.gnu.org/23640#58
* .dir-locals.el (c-mode): Adjust to macro changes.
* src/conf_post.h (NONVOLATILE, UNINIT): New macros (Bug#23640).
(IF_LINT): Remove. All uses replaced by the new macros.
* src/conf_post.h (__has_builtin, __builtin_assume_aligned):
New macros, for compilers not already defining them.
(__has_builtin___builtin_assume_aligned): New macro.
* src/lisp.h (lisp_h_XUNTAG): Use __builtin_assume_aligned.
This shrinks text space by 0.2% on x86-64 with GCC 6.1.
6e3adf8 Fix crash in syntax.c after GC
973ce5a Improve squiggly heredoc support in non-SMIE Ruby mode
9d5cceb Fix doc string quoting
0b33a23 Fix mouse dragging of vertical dividers with scroll bars on l...
a5d05f4 * etc/PROBLEMS: Mention the link-time problems on FreeBSD 11.
Problem reported by Vincent Belaïche (Bug#23704).
* src/syntax.c (skip_chars): Recompute pointers into the
buffer after every call to update_syntax_table_forward,
as it can GC.
* doc/lispref/os.texi (Time Zone Rules): Document MS-Windows
lack of support for numeric time zone abbreviations.
* src/w32.c (sys_putenv): Convert angle-bracket TZ syntax
to MS-compatible syntax if possible, and to "ZZZ" otherwise.
Problem reported by Kazuhiro Ito (Bug#23600).
