* configure.ac: Stop relaying --with-ndk-cxx-shared to the
nested invocation of configure.
* build-aux/ndk-build-helper-1.mk (SYSTEM_LIBRARIES):
* build-aux/ndk-build-helper-2.mk (SYSTEM_LIBRARIES): Insert all
of the C++ libraries available on Android.
* configure.ac: Call ndk_LATE and ndk_LATE_EARLY within if
statement at toplevel, averting needless calls to AC_PROG_CXX.
* cross/ndk-build/Makefile.in (NDK_BUILD_CXX_STL)
(NDK_BUILD_CXX_LDFLAGS):
* cross/ndk-build/ndk-build.mk.in (NDK_BUILD_CXX_STL)
(NDK_BUILD_CXX_LDFLAGS): New variables.
* cross/ndk-build/ndk-resolve.mk (NDK_SYSTEM_LIBRARIES):
Introduce several other C++ libraries sometimes present on
Android.
(NDK_SO_EXTRA_FLAGS_$(LOCAL_MODULE)): Insert NDK_BUILD_CXX_STL
when any of these new C++ libraries are requested.
* m4/ndk-build.m4: Completely rewrite C++ compiler and library
detection.
* java/org/gnu/emacs/EmacsNative.java (EmacsNative): Attempt to
load more libraries from static initializer.
* java/INSTALL: Remove obsolete information.
We can have several active `handler-bind`s that all want to invoke the
debugger, in which case we can have the following sequence:
- The more deeply nested handler calls the debugger.
- After a while the user invokes `debugger-continue`.
- `signal_or_quit` propagates the error up the stack to the
second handler, which calls the debugger again.
- The user thus ends up right back at the same place, as if
`debugger-continue` had not be processed.
Fix this by remembering the last processed error and skipping
the debugger if we bump into it again.
* lisp/emacs-lisp/debug.el (debugger--last-error): New var.
(debugger--duplicate-p): New function.
(debug): Use them.
Suggested by Michael Heerdegen <michael_heerdegen@web.de>.
* src/window.c (window_restore_killed_buffer_windows): In
doc-string write 'symbol' instead of 'constant'.
* lisp/window.el (display-buffer--lru-window)
(display-buffer-use-least-recent-window): In doc-strings write
'symbol' instead of 'constant'.
* doc/lispref/windows.texi (Window Configurations): Write
'symbol' instead of 'constant'.
* etc/NEWS: Announce 'window-restore-killed-buffer-windows'.
* src/buffer.h (struct buffer) : New field last_name_.
* src/buffer.c (Fbuffer_last_name): New function to return last
name of buffer before it was killed or renamed.
(bset_last_name, Fget_buffer_create, Fmake_indirect_buffer)
(Frename_buffer, Fkill_buffer, init_buffer_once): Set buffer's
last_name_ field accordingly.
* src/window.c (window_restore_killed_buffer_windows): New
variable replacing Vwindow_kept_windows_functions.
(Fset_window_configuration): Use
window_restore_killed_buffer_windows instead of
Vwindow_kept_windows_functions.
* lisp/window.el (window--state-put-2, window-state-put): Use
'window-restore-killed-buffer-windows' instead of
'window-kept-windows-functions'.
* doc/lispref/windows.texi (Window Configurations): Describe
'window-restore-killed-buffer-windows' which replaces
'window-kept-windows-functions'.
Previously, if minibuffer-visible-completions was non-nil, we bound RET
whenever the *Completions* buffer was visible. This meant that RET in
completion-in-region would not enter a newline, which is a somewhat
annoying behavior change from minibuffer-visible-completions=nil.
Now, we only bind RET when a completion is selected. This means
RET will newline in completion-in-region.
So that completion help continues to suggest the correct keys,
we also add minibuffer-visible-completions--always-bind. When
let-bound to a non-nil value, it makes the
minibuffer-visible-completions binds always active. We let-bind
it around substitute-command-keys.
* lisp/minibuffer.el (minibuffer-visible-completions--always-bind)
(minibuffer-visible-completions--filter): Add.
(minibuffer-visible-completions-bind): Use
minibuffer-visible-completions--filter. (bug#68801)
* lisp/simple.el (minibuffer-visible-completions--always-bind)
(completion-setup-function): Let-bind
minibuffer-visible-completions--always-bind so the completion
help is correct.
The 'shr-fill-...' functions handle both hard-filling (adding
newlines to break long lines) and indentation. Setting
'shr-fill-text' to nil currently causes these functions to be
short-circuited completely, so e.g. blockquotes are no longer
indented, whereas the intent of this user option is only to
prevent hard-filling to let visual-line-mode reflow text.
* lisp/net/shr.el (shr-fill-lines): Document that the function
handles more than just filling; move the 'shr-fill-text' check...
(shr-fill-line): ... here, after indentation has been taken care
of.
* test/lisp/net/shr-resources/blockquote.html:
* test/lisp/net/shr-resources/blockquote.txt: New test resources.
* test/lisp/net/shr-tests.el (shr-test--rendering-check): Rename
from 'shr-test', to make the relationship with the 'rendering'
testcase clearer; prefer 'file-name-concat' to 'format'; raise
ERT failure if need be, calling (ert-fail ...) directly instead
of (should (not (list ...))).
(shr-test--rendering-extra-configs): New variable to easily check
that user customizations do not degrade rendering.
(rendering): Consult that new variable; delegate failure-raising
to reduce duplication.
* lisp/progmodes/project.el (project--value-in-dir):
Ensure that the global value of the variable is still honored,
when the variable is not in dir-locals.
The old code tested if the mode function is `fboundp` but in an
inconsistent way and without paying attention to
`major-mode-remap`.
* lisp/files.el (set-auto-mode-0): Return `:keep` rather than nil if
the mode was already set. And emit a warning when the mode function
doesn't exist.
(set-auto-mode): Remove checks that the mode function exists now that
`set-auto-mode-0` does it for us. Adjust to the new return values of
that function, and simplify the code using a big `or` instead of
a sequence of steps each setting&testing `done`.
(hack-local-variables--find-variables): Use `major-mode-remap`
when skipping the "mode:" entries that specify modes we don't have.
Also, when (eq handle-mode t), don't bother building a list of results
only to return a single element in the end.
(flyspell--mode-on): Add flyspell-check-changes to post-command-hook
when flyspell-check-changes is non-nil.
(flyspell--mode-off): Remove flyspell-check-changes from post-command-hook.
(flyspell-check-changes): New function (bug#61874).
(project--value-in-dir): Use 'alist-get' on 'file-local-variables-alist'
to avoid calling 'hack-local-variables-apply' via
'hack-dir-local-variables-non-file-buffer' because it might enable
undesirable modes such as flyspell-mode in a temporary buffer (bug#69740).
* lisp/menu-bar.el (menu-bar-project-item): New variable from
'project-menu-entry'.
(menu-bar-tools-menu): Use 'menu-bar-project-item'.
* lisp/mouse.el (context-menu-functions): Add
'context-menu-project' to choice.
(context-menu-project): New function.
* lisp/progmodes/project.el (project-menu-entry): Remove variable.
(project-mode-line-map): Use 'menu-bar-project-item' instead of
'project-menu-entry'.
* lisp/progmodes/python.el
(python-shell-comint-watch-for-first-prompt-output-filter):
Detect the readline support.
(python-shell-readline-completer-delims): Update docstring.
(python-shell-completion-native-setup): Move the readline
detection code to ...
(python-shell-readline-detect): ... new function.
(python-shell-completion-native-turn-on-maybe): Skip if Python
has no readline support.
(python-shell-completion-at-point): Respect the delimiter of
readline completer in non-native completion.
* test/lisp/progmodes/python-tests.el
(python-shell-completion-at-point-1)
(python-shell-completion-at-point-native-1)
(python-completion-at-point-1, python-completion-at-point-2)
(python-completion-at-point-pdb-1)
(python-completion-at-point-while-running-1)
(python-completion-at-point-native-1)
(python-completion-at-point-native-2)
(python-completion-at-point-native-with-ffap-1)
(python-completion-at-point-native-with-eldoc-1): Skip tests if
Python has no readline support.
(python-shell-completion-at-point-jedi-completer): Add test for
non-native Python shell completion. (bug#68559)
* lisp/thingatpt.el (thing-at-point-decimal-regexp)
(thing-at-point-hexadecimal-regexp): Extract regexps from...
(number-at-point): ...here. Use them in 'number-at-point'.
(number): Add 'bounds-of-thing-at-point' property as
`forward-word' does not always return the right boundary,
e.g., in latex-mode buffers. (Bug#69239)
