Use this in C, C++, and Objective C Modes. Fixes bug#22246
* src/syntax.c (comment-end-can-be-escaped): New buffer local variable.
(forw-comment, back-comment): On encountering an end of comment character,
test whether it is escaped when `comment-end-can-be-escaped' is non-nil.
* doc/lispref/syntax.texi (Control Parsing): Describe
`comment-end-can-be-escaped'.
* etc/NEWS (Lisp Changes): Describe `comment-end-can-be-escaped'.
* lisp/progmodes/cc-langs.el: New c-lang-setvar `comment-end-can-be-escaped'.
This is a work-around to prevent the compiler from using semantic
knowledge about malloc for optimization purposes. E.g., gcc 5.2
with -O2 replaces most of calloc's definition by a call to calloc;
see Bug#22085.
* src/gmalloc.c [!HYBRID_MALLOC] (malloc, realloc, calloc)
(aligned_alloc, free): Do not undef. Instead, define these as
functions (perhaps renamed to gmalloc etc.) in terms of gmalloc etc.
* src/eval.c (unbind_to): Redo so that the FALLTHROUGH!! comment
becomes accurate again. This shouldn’t affect behavior.
(backtrace_eval_unrewind): Apply the recent unbind_to fix here, too.
* lisp/international/ucs-normalize.el (eval-when-compile): Make
sure char-code-property-alist includes elements that allow access
to 'decomposition' and 'canonical-combining-class' Unicode
properties, as compiling ucs-normalize.el requires that.
* lisp/loadup.el (featurep 'ns): Load ucs-normalize and ns-win
only of charprop.el was already loaded.
* src/Makefile.in ($(lispsource)/international/ucs-normalize.elc):
New order-only dependency.
* src/dired.c (file_name_completion): Reject false matches due to
file-name-coding-systems that decompose characters when encoding
file names, by comparing decoded file names as well. (Bug#22169)
(syms_of_dired) <Qdecomposed_characters>: New DEFSYM.
* lisp/international/ucs-normalize.el (utf-8-hfs): Give it a
non-nil 'decomposed-characters' property.
The coding system `utf-8-nfd', locally defined in ns-win.el,
didn't provide a :pre-write-conversion method, causing file name
completion of non-ASCII characters to fail. Solved by using the
`utf-8-hfs' coding system provided by `ucs-normalize'.
* lisp/loadup.el: Load international/ucs-normalize (when building
for ns).
* lisp/term/ns-win.el (utf-8-nfd): Made `utf-8-nfd' as alias for
`utf-8-hfs' and removed the old implementation. Set `utf-8-hfs'
as the file name coding system.
* src/nsfns.c (ns-convert-utf8-nfd-to-nfc): Removed.
Port recent fix for Bug#21968 to platforms like 'gcc -fno-common'.
* src/keyboard.c, src/keyboard.h (point_before_last_command_or_undo)
(buffer_before_last_command_or_undo):
Declare in keyboard.h, and define in keyboard.c,
instead of assuming the traditional Unix relaxed ref-def linkage.
* src/lisp.h: Document emacs-module.c assumptions about EQ and NILP.
* src/emacs-module.c (module_non_local_exit_get): Document that we
cannot use the current implementation.
(module_is_not_nil, module_eq): Document assumptions about EQ and
NILP.
* src/indent.c (Fvertical_motion): When moving from line beginning
to point under line truncation, assume overshoot by one line only
if point actually lies beyond the window's right margin.
(Bug#22194)
C11 threads are not needed for Emacs now, and their use is causing
hassles on FreeBSD 10.x. Problem reported by Ashish SHUKLA in:
http://lists.gnu.org/archive/html/emacs-devel/2015-12/msg00648.html
* configure.ac: Do not check for C11 threads. Remove unnecessary
fiddling with CPPFLAGS when configuring pthreads.
* src/emacs-module.c (main_thread, check_main_thread)
(module_init): Do not worry about C11 threads.
This fixes the smaller performance hit that I noted in:
https://lists.gnu.org/archive/html/emacs-devel/2015-12/msg00357.html
* src/alloc.c (macro_XPNTR_OR_SYMBOL_OFFSET, macro_XPNTR):
* src/puresize.h (puresize_h_PURE_P)
(puresize_h_CHECK_IMPURE):
New macros, with the old contents of the functions.
* src/alloc.c (XPNTR_OR_SYMBOL_OFFSET, XPNTR):
* src/puresize.h (PURE_P, CHECK_IMPURE):
Use the new macros. Also macros, if DEFINE_KEY_OPS_AS_MACROS.
* src/conf_post.h (ATTRIBUTE_UNUSED):
* src/lisp.h (DEFINE_KEY_OPS_AS_MACROS): New macros.
* doc/lispref/frames.texi (Parameter Access): Mention pitfalls
when simultaneously specifying multiple parameters for
`modify-frame-parameters' that all may change the frame's size.
* src/w32fns.c (x_set_menu_bar_lines): Don't set
windows_or_buffers_changed here.
(my_create_tip_window, Fx_show_tip): Call AdjustWindowRect
with third argument false.
* src/w32menu.c (set_frame_menubar): Set
windows_or_buffers_changed here.
* src/w32term.c (x_set_window_size): Determine third argument of
AdjustWindowRect from whether the frame has a menu bar and not
from whether it wants one.
* src/keyboard.c (command_loop_1): Restore the feature whereby C-u
was part of this-command-keys, but not of this-single-command-keys.
(Bug#22107)
* lisp/simple.el (internal-echo-keystrokes-prefix): Add
commentary about the function's return value.
* doc/lispref/frames.texi (Parameter Access): Clarify what "ignored
PARMs" mean for 'modify-frame-parameters'.
* src/frame.c (Fmodify_frame_parameters): Clarify what "ignored
PARMs" mean for this function. (Bug#22104)
* src/xdisp.c (redisplay_window): When someone forced
window-start, and honoring that failed to show the cursor, try
moving out of invisible text, before falling back to the middle of
the window. (Bug#22098)
When configured with --enable-locallisppath=no, which is the
default for OS X, the load-path incorrectly was populated with ".".
* src/lread.c (init_lread): Don't call `decode_env_path' when
PATH_SITELOADSEARCH is empty.
* doc/misc/calc.texi (Predefined Units): Use the bland modern
scientific style for spelling the units “ampere” and
“angstrom” rather than the older style “Ampere” and
“Ångstrom”. The latter spelling was wrong anyway (it should
have been “Ångström”).
* lisp/emacs-lisp/ert.el (ert--explain-equal-rec):
Fix misspelling of ‘atom’ in code.
* src/xdisp.c (redisplay_window): When scrolling fails to show
point, prefer using the desired matrix if possible for finding the
fallback glyph row for displaying the cursor. (Bug#22098)
(row_containing_pos): Exit the loop as soon as we hit the first
disabled glyph row. Otherwise we risk accessing garbled data and
departing to the no-no land.
* src/emacs-module.c (plain_values): New constant.
(module_nil): Now a constant.
(Finternal_module_call, value_to_lisp_bits, lisp_to_value_bits)
(syms_of_module): Use if, not #ifdef, so that both sides are
checked at compile-time, and so that GCC doesn’t complain
about an unused var in the typical case. Also, depend on
plain_values, not on WIDE_EMACS_INT; the code shouldn’t assume
that WIDE_EMACS_INT implies !USE_LSB_TAG.
(value_to_lisp_bits, lisp_to_value_bits): New functions.
Sign-extend integers rather than zero-extending them, as small
negative integers are more likely.
(value_to_lisp, lisp_to_value): Rewrite in terms of the new *_bits
functions.
(HAVE_STRUCT_ATTRIBUTE_ALIGNED): Define to 0 if not already defined.
(mark_modules): Remove. All uses removed.
(lisp_to_value): Don’t assume Fcons returns a pointer aligned
to GCALIGNMENT.
(syms_of_module): Check that module_nil converts to Qnil.
* src/lisp.h (lisp_h_XSYMBOL, XSYMBOL): Use signed conversion, since
we prefer signed to unsigned when either will do.
(TAG_PTR): Sign-extend pointers when USE_LSB_TAG, as this is
a bit better for emacs-module.c.
* src/emacs-module.c (value_to_lisp) [WIDE_EMACS_INT]: Use
unsigned data types to manipulate pointers, to avoid sign
extension coming after us with a vengeance.
* modules/mod-test/test.el (mod-test-sum-test): Add tests for
Emacs with wide ints that verify integer values near the critical
value that requires us to switch to a cons cell.
* src/emacs-module.c (lisp_to_value): Compare the produced value
with the original Lisp object, not with the one potentially
converted into a Lisp_Cons. Fixes assertion violations when
working with integers larger than fit into a 32-bit value.
* modules/mod-test/test.el (mod-test-sum-test): Add tests for
large integers, to test --with-wide-int.
* src/emacs-module.c (struct emacs_value_tag)
(struct emacs_value_frame, struct emacs_value_storage): Remove.
(value_frame_size): Remove constant.
(struct emacs_env_private): Use Lisp_Object for non_local_exit info.
(lisp_to_value): Remove first arg.
(module_nil): New constant.
Use it instead of NULL when returning an emacs_value.
(module_make_function): Adjust to new calling convention of
Qinternal_module_call.
(DEFUN): Receive args in an array rather than a list.
Use SAFE_ALLOCA rather than xnmalloc. Skip the lisp_to_value loop when
we don't have WIDE_EMACS_INT. Adjust to new type of non_local_exit info.
(module_non_local_exit_signal_1, module_non_local_exit_throw_1):
Adjust to new type of non_local_exit info.
(ltv_mark) [WIDE_EMACS_INT]: New constant.
(value_to_lisp, lisp_to_value): Rewrite.
(initialize_frame, initialize_storage, finalize_storage): Remove functions.
(allocate_emacs_value): Remove function.
(mark_modules): Gut it.
(initialize_environment): Don't initialize storage any more.
Keep the actual env object on Vmodule_environments.
(finalize_environment): Don't finalize storage any more.
(syms_of_module): Initialize ltv_mark and module_nil.
* src/emacs-module.h (emacs_value): Make it more clear that this type
is really opaque, including the fact that NULL may not be valid.
* modules/mod-test/mod-test.c (Fmod_test_signal, Fmod_test_throw):
Don't assume that NULL is a valid emacs_value.
* src/emacs-module.c (module_make_function): Use internal--module-call.
(Finternal_module_call): Renamed from Fmodule_call. Add safety
checks.
(syms_of_module): DEFSYM save-value-p and save-pointer-p. Do
defsubr internal--module-call.
(struct emacs_runtime_private): Keep an emacs_env instead.
(Fmodule_load, Fmodule_call): Declare emacs_env_private separately.
(initialize_environment): Split the arg in two. Adjust all callers.
Only store the private part in Vmodule_environments.
(finalize_environment): Change the arg to only be the private env.
Adjust all callers.
