* lisp/emacs-lisp/bytecomp.el (byte-compile-form)
(bytecomp--actually-important-return-value-p):
Special handling of `sort` that takes into account that it may return
an important value depending on the :in-place keyword argument.
* lisp/sort.el (sort-on):
* doc/lispref/sequences.texi (Sequence Functions):
* etc/NEWS:
Remove the `sort-on` function which is now completely superseded by
the extended `sort` in features, ease of use, and performance.
Postpone the creation of a new list to after sorting which turns out to
be a lot faster (1.1x - 1.5x speedup).
* src/fns.c (sort_list, sort_vector, Fsort):
Create the new list when moving the data out from the temporary array.
This gives a 1.5x-2x speed-up when using the default :lessp value,
by eliminating the Ffuncall overhead.
* src/sort.c (order_pred_lisp, order_pred_valuelt): New.
(merge_state, inorder, binarysort, count_run, gallop_left, gallop_right)
(merge_init, merge_lo, merge_hi, tim_sort):
* src/fns.c (Fsort):
When using value<, call it directly.
Add the :key, :lessp, :reverse and :in-place keyword arguments.
The old calling style remains available and is unchanged.
* src/fns.c (sort_list, sort_vector, Fsort):
* src/sort.c (tim_sort):
Add keyword arguments with associated new features.
All callers of Fsort adapted.
* test/src/fns-tests.el (fns-tests--shuffle-vector, fns-tests-sort-kw):
New test.
* doc/lispref/sequences.texi (Sequence Functions): Update manual.
* etc/NEWS: Announce.
The original timsort code did provide for a key (accessor) function
along with the necessary storage management, but we dropped it because
our `sort` function didn't need it.
Now it's been put back since it seems that it will come in handy after all.
* src/fns.c (sort_list, sort_vector, Fsort): Pass Qnil as key function
to tim_sort.
* src/sort.c (reverse_slice, sortslice_copy)
(sortslice_copy_incr, sortslice_copy_decr, sortslice_memcpy)
(sortslice_memmove, sortslice_advance): New functions.
(sortslice): New type.
(struct stretch, struct reloc, merge_state)
(binarysort, merge_init, merge_markmem, cleanup_mem)
(merge_register_cleanup, merge_getmem, merge_lo, merge_hi, merge_at)
(found_new_run, reverse_sortslice, resolve_fun, tim_sort):
Merge back previously discarded parts from the upstreams timsort code
that dealt with key functions, and adapt them to fit in.
It's a general-purpose polymorphic ordering function, like `<` but
for any two values of the same type.
* src/data.c (syms_of_data): Add the `type-mismatch` error.
(bits_word_to_host_endian): Move...
* src/lisp.h (bits_word_to_host_endian): ...here, and declare inline.
* src/fns.c (Fstring_lessp): Extract the bulk of this function to...
(string_cmp): ...this 3-way comparison function, for use elsewhere.
(bool_vector_cmp, value_cmp, Fvaluelt): New.
* lisp/emacs-lisp/byte-opt.el (side-effect-free-fns, pure-fns):
Add `value<`, which is pure and side-effect-free.
* test/src/fns-tests.el (fns-value<-ordered, fns-value<-unordered)
(fns-value<-type-mismatch, fns-value<-symbol-with-pos)
(fns-value<-circle, ert-deftest fns-value<-bool-vector): New tests.
* doc/lispref/sequences.texi (Sequence Functions):
* doc/lispref/numbers.texi (Comparison of Numbers):
* doc/lispref/strings.texi (Text Comparison):
Document the new value< function.
* etc/NEWS: Announce.
* doc/emacs/android.texi (Android Windowing): Document proviso
on Android 7.0 and later.
* java/org/gnu/emacs/EmacsActivity.java (EmacsActivity)
<timeOfLastInteraction>: New field.
(onStop, onResume): Set and clear timeOfLastInteraction.
(isReallyFinishing): New function.
(onDestroy): Don't delete frame even in the event isFinishing
returns true if more than 4 hours have elapsed since the
activity last moved into the background.
* lisp/proced.el (proced-auto-update-flag): Document 'visible'
value and add it to the custom type.
(proced-auto-update-timer, proced-toggle-auto-update): Take
'visible' value into account.
Fail on files with no assertions, parser now accepts multiple
carets per line and face lists:
* lisp/emacs-lisp/ert-font-lock.el: Assertion parser fix.
* test/lisp/emacs-lisp/ert-font-lock-resources/no-asserts.js:
* test/lisp/emacs-lisp/ert-font-lock-tests.el
(test-parse-comments--no-assertion-error)
(test-syntax-highlight-inline--caret-negated-wrong-face)
(test-macro-test--file-no-asserts): New test cases.
* doc/misc/ert.texi (Syntax Highlighting Tests): More syntax examples.
The `pcase-mutually-exclusive-predicates` table was not very
efficient since it grew like O(N²) with the number of
predicates. Replace it with an O(N) table that's auto-generated
from the `built-in-class` objects.
* lisp/emacs-lisp/pcase.el (pcase-mutually-exclusive-predicates):
Delete variable.
(pcase--subtype-bitsets): New function and constant.
(pcase--mutually-exclusive-p): Use them.
* lisp/emacs-lisp/cl-preloaded.el (built-in-class): Don't inline.
This was just a typo: we were checking the modification time of
current-buffer instead of checking the modification time of the
passed-in buffer.
This caused matches to not be shown if they weren't present in
the current in-Emacs state of the buffer.
This was easily reproduced by writing a string to a file outside
Emacs, then searching for that string with
e.g. project-find-regexp. The string would seemingly not be
found, although in reality it was found, just not displayed.
* lisp/progmodes/xref.el (xref--find-file-buffer):
Check buf, not current-buffer (bug#70008).
* java/org/gnu/emacs/EmacsView.java (onCreateInputConnection):
Passwords might also be non-ASCII, and this flag apparently
requests an IME limited to ASCII characters, rather than just
capable of ASCII input.
The change caused type-check errors in auth-source where they use
`:type function` constraints on object slots and expect those to
be able to hold symbols.
* lisp/emacs-lisp/cl-preloaded.el (function): Revert last change.
* test/src/data-tests.el (data-tests--cl-type-of): Use `cl-functionp`
rather than `functionp` to test `function`.
* lisp/progmodes/cc-engine.el (c-after-conditional): Handle the
new keyword in place of a paren sexp after `if'.
* lisp/progmodes/cc-langs.el (c-negation-op-re)
(c-paren-clause-kwds, c-paren-clause-key)
(c-block-stmt-with-kwds, c-block-stmt-with-key): New
lang-consts/vars.
* if-11.cc, if-11.res: New test files.
We generally want types to form not just a DAG but a lattice.
If objects can be both `keyword` and `symbol-with-pos`, this
means there should be a more precise type describing this intersection.
If we ever find the need for such a refinement, we could add
such a `keyword-with-pos` type, but here I took the simpler
route of treating `keyword` not as a proper built-in type but
as a second-class type like `natnum`.
While fixing this problem, also fix the problem we had where
`functionp` was not quite adequate to characterize objects of type
`function`, by introducing a new predicate `cl-functionp` for that.
* lisp/emacs-lisp/cl-preloaded.el (cl-functionp): New function.
(function): Use it.
(keyword): Don't declare it as a built-in type.
(user-ptrp): Remove redundant declaration.
* lisp/emacs-lisp/cl-generic.el (cl--generic--unreachable-types):
Delete constant.
(cl-generic-generalizers): Remove corresponding test.
* lisp/emacs-lisp/cl-macs.el (cl-deftype-satisfies): Add entry for
`keyword` type.
* lisp/emacs-lisp/comp.el (comp-known-predicates): Fix type for
negative result of `characterp`. Remove duplicate `numberp` entry.
Fix types for `keywordp` now that `keyword` is not a built-in type any more.
* test/src/data-tests.el (data-tests--cl-type-of): Add a few cases.
Remove workaround for `function`.
* lisp/emacs-lisp/comp.el (comp-known-predicates)
(comp-known-predicates-h): Update.
(comp--pred-to-pos-cstr, comp--pred-to-neg-cstr): New functions.
(comp--add-cond-cstrs): Make use of them.
* test/src/comp-tests.el (comp-tests-type-spec-tests): Add a test.
* java/org/gnu/emacs/EmacsActivity.java (invalidateFocus1): New
argument resetWhenChildless.
(invalidateFocus): If a toplevel window has no focus window,
transfer focus to the toplevel itself.
* doc/lispref/commands.texi (Misc Events): Document new value of
text-conversion-style.
* java/org/gnu/emacs/EmacsService.java (EmacsService)
<IC_MODE_PASSWORD>: New constant.
* java/org/gnu/emacs/EmacsView.java (onCreateInputConnection):
Set TYPE_TEXT_VARIATION_PASSWORD and IME_FLAG_FORCE_ASII if mode
is IC_MODE_PASSWORD.
* lisp/subr.el (read-passwd): Set text-conversion-style to
`password'.
* src/androidgui.h (enum android_ic_mode): New value
ANDROID_IC_MODE_PASSWORD.
* src/androidterm.c (android_reset_conversion): Handle
`password'.
* src/buffer.c (syms_of_buffer)
<&BVAR (current_buffer, text_conversion_style)>: Update doc
string.
* src/textconv.c (syms_of_textconv) <Qpassword>: New DEFSYM.
<Vtext_conversion_edits>: Fix typos in doc string.
* lisp/bookmark.el (bookmark-buffer-file-name): Support Info
buffers.
(bookmark--remove-fringe-mark): Call 'bookmark-buffer-file-name'
instead of using 'buffer-file-name', which could be nil.
(Bug#69974)
Remove extra "\n" from the end of the "diff --git.*" part
of 'diff-outline-regexp' because "\n" is not used in outline-regexp
and causes problems in such cases like when killing hunks
in the diff buffer with outline-minor-mode that loses
the outline icons because outline--fix-buttons-after-change and
outline--fix-up-all-buttons are limited to the single line and
can't match an outline line with a regexp that ends with "\n".
Eliminate a case that matches very rarely and where the default
handling works just as well anyway.
* lisp/emacs-lisp/byte-opt.el (byte-optimize-form-code-walker): Remove
redundant case.
* java/org/gnu/emacs/EmacsService.java (getDisplayNameHash):
Always encode file names as modified UTF-8, as insurance against
future changes to undocumented behavior of the JVM.
Rework the code that "cleans" the backtrace for `edebug-pop-to-backtrace`.
The main changes are the following:
- Strip instrumentation from "everywhere" rather than trying to limit the
effect to "code" and leave "data" untouched. This is a worthy
goal, but it is quite difficult to do since code contains data
(so we ended up touching data anyway) and data can also
contain code.
The risk of accidentally removing something because it happens
to look like instrumentation is very low, whereas it was very common
for instrumentation to remain in the backtrace.
- Use a global hash-table to remember the work done, instead of
using separate hash-table for each element. By using a weak
hash-table we avoid the risk of leaks, and save a lot of work
since there's often a lot of subexpressions that appear
several times in the backtrace.
* lisp/emacs-lisp/edebug.el (edebug-make-enter-wrapper): Tweak code
layout so the comments are more clear.
(edebug-unwrap): Remove redundant patterns for `closure` and `lambda`.
Add `:closure-dont-trim-context` to the `edebug-enter` pattern,
so it also gets removed (this should have been done in commit
750bc57cbb).
(edebug--unwrap-cache): New var.
(edebug-unwrap*): Use it.
(edebug--unwrap1): Delete function. Merged into `edebug-unwrap*`.
Also apply unwrapping to the contents of byte-code functions since they
can refer to lambda expressions captured by the closure.
(edebug--symbol-prefixed-p): Rename from
`edebug--symbol-not-prefixed-p` and adjust meaning accordingly.
(edebug--strip-instrumentation): Adjust accordingly and simplify
a bit by unifying the "lambda" case and the "everything else" case.
(edebug--unwrap-frame): Use `cl-callf` and unwrap arguments even if
they've already been evaluated.
* lisp/erc/erc-goodies.el (erc-load-irc-script-lines): Move here from
main file and rework to always use `command-indicator' instead of only
partially, when available. Also use internal "send-action" and
"send-message" interfaces to defer command-handler output until
command lines have been inserted.
* lisp/erc/erc.el (erc-process-input-line): Redo doc string.
(erc-process-script-line): Fold exceptionally overlong line.
(erc-load-irc-script-lines): Move to erc-goodies.el. (Bug#67032)
