Earlier versions of Emacs were limited to at most 32 subprocesses or
network connections on Windows due to the 64-object limit imposed by
WaitForMultipleObjects. To overcome this, a simple waiting thread
pool is implemented, allowing Emacs to efficiently wait on up to
2048 objects.
Each thread in the pool can wait on up to 63 objects, and a total of
32 threads are used, together with the main thread, to expand the
waiting capability. This enables Emacs to support approximately 1024
subprocesses, which is comparable to the 'pty' method on GNU/Linux
when using the default FD_SETSIZE of 1024.
To minimize overhead, the threads remain active instead of being
frequently created and destroyed, reducing unnecessary system
resource consumption. Idle threads can be terminated after a
period of inactivity to free up memory.
* src/w32.h (FD_SETSIZE): Change from 64 to 2048.
* src/w32.c (term_ntproc): Call 'free_wait_pool' to free waiting
threads.
* src/w32proc.c (WFO_ABANDONED, WFO_TIMEOUT, WFO_FAILED)
(WFO_MAX_WAIT): New macros.
(wait_objects_context, wait_objects_pool, wait_objects_info):
New structures for managing the thread pool.
(wait_objects_pool, wait_objects_info): New static variables for
managing the thread pool.
(wait_objects_thread, start_wait_objects, stop_wait_objects)
(end_wait_and_return, shrink_wait_pool, free_wait_pool):
New functions for waiting and managing the thread pool.
(wait_for_objects, msg_wait_for_objects): New functions as
replacements for WaitForMultipleObjects and
MsgWaitForMultipleObjects.
(wait_pid): Use 'wait_for_objects' and new macros.
(sys_select): Make some variables static to avoid stack
allocation. Use 'wait_for_objects', 'msg_wait_for_objects', and
the new macros.
Use `replace_range` rather than `delete+insert`.
* src/insdel.c (replace_range): Allow NEW to specify a chunk
of buffer text.
* src/editfns.c (Freplace_buffer_contents): Use it.
* test/src/editfns-tests.el (replace-buffer-contents-1): Remove
incorrect check which happened to succeed because point was not
preserved carefully enough. Make the replacement text share a bit
more content to make the test a bit more strict.
(editfns-tests--replace-region): Doesn't fail any more.
* src/frame.c (tty_child_pos_param): Handle additional position
values. New SIZE argument.
(tty_child_frame_rect): Process size before position.
(Fmake_terminal_frame): Do not store calculated size/position
values in parameters to avoid that a negative absolute position
is later interpreted as position relative to bottom/right edge
of parent.
(Fmodify_frame_parameters): For tty child frames process size
before position values so the latter can take a new size into
account.
* src/frame.h: Adjust declaration of tty_child_pos_param.
* src/gtkutil.c (xg_set_geometry): Handle negative frame
position specifications.
* src/msdos.c (IT_set_frame_parameters): For tty child frames
process size before position parameters.
* src/xfns.c (x_window): In the non-GTK toolkit variant leave
negative position values unchanged - the toolkit should know how
to handle them. Process child frame position parameters
separately. In the non-toolkit variant process child frame
parameters separately.
* src/xterm.c (x_fast_mouse_position): Clear the mouse_moved
flag whether or not mouse motion is reported, to avoid repeated
polling for mouse movement. (bug#76806)
* lisp/frame.el (frame--special-parameters): New const.
(frame--complete-parameter-value): New function.
(initial-frame-alist, minibuffer-frame-alist): Use them in
:type. (Bug#39143)
* lisp/cus-start.el (default-frame-alist): Use them here as well.
* src/frame.c (frame_parms): Add comment to try to keep
frame--special-parameters updated.
* configure.ac (ALIGNOF_INT, ALIGNOF_LONG, ALIGNOF_LONG_LONG):
New variables.
(emacs_cv_alignas_unavailable): Define if alignas and structure
alignment primitives are unavailable. In such an environment,
the MSB tagging scheme must be enabled, as must the GNU malloc.
* msdos/sed2v2.inp: Adjust correspondingly.
* src/alloc.c (union emacs_align_type): Remove types which
contain flexible array members. The address of a field
subsequent to an aggregate with flexible array members cannot
validly be taken.
(mark_memory) [!USE_LSB_TAG && !WIDE_EMACS_INT]: Strip type bits
before scanning memory.
* src/emacs.c (main):
* src/eval.c (Fautoload_do_load):
* src/fns.c (Frequire): Rename a number of illogically named
fields.
* src/lisp.h (ALIGNOF_EMACS_INT): Define to the natural
alignment of EMACS_INT.
(IDEAL_GCALIGNMENT): New macro.
(USE_LSB_TAG): Disable if no alignment specifiers are available,
WIDE_EMACS_INT is undefined, and the natural alignment of
EMACS_INT falls short of LSB tagging's requirements.
(gflags): Rename illogically named fields and don't define them
as bitfields, which runs afoul of certain compiler issues.
(will_dump_p, will_bootstrap_p, will_dump_with_pdumper_p)
(dumped_with_pdumper_p): Adjust accordingly.
* src/pdumper.c (VM_SUPPORTED): Define to 0 when !USE_LSB_TAG.
It is better to read dump files into the heap by hand than to be
supplied with an address that is not representable.
(_dump_object_start_pseudovector): Rename to
dump_object_start_pseudovector, to avoid encroaching on reserved
names.
(START_DUMP_PVEC): Adjust correspondingly.
(dump_mmap_contiguous_vm): Preserve errno around failure
cleanup.
(dump_bitset_bit_set_p): Work around certain compiler issues.
(pdumper_load) [!USE_LSB_TAG]: Reject dump file allocations
that are not representable as Lisp_Objects.
Tested on i386-unknown-solaris2.10, sparc-sun-solaris2.10.
* src/term.c (tty_write_glyphs_1): Renamed from tty_write_glyphs. Don't
check if writing the bottom-right corner.
(tty_write_glyphs): New function handling case of writing to the
bottom-right corner, and otherwise calling tty_write_glyphs_1.
* src/pdumper.c (dump_sort_copied_objects, drain_reloc_list)
(dump_do_fixups, dump_do_dump_relocation): Do not expand
Fnreverse multiple times in CALLN. Sun C 5.12 evaluates the
arguments to array initializers to which it expands, which
clobbers the lists being reversed.
* src/frame.c (delete_frame): Ignore tooltip frames when looking
for other frames on the same terminal. (Bug#76842)
(cherry picked from commit d2445c8c23)
lisp/term/w32-win.el contained definitions of variables which
were not used on Cygw32. It is now to split into two files,
common part and Windows native build specific part. The latter
is a new file, lisp/term/w32-nt.el.
* src/image.c (Qlibpng_version, Qlibgif_version)
(Qlibjpeg_version): Don't define on Cygw32 build.
* src/treesit.c (Qtree_sitter__library_abi): Ditto.
* lisp/term/w32-win.el (dynamic-library-alist, libpng-version)
(libgif-version, libjpeg-version, libgnutls-version)
(tree-sitter--library-abi, gui-backend-set-selection)
(gui-backend-get-selection, gui-backend-selection-owner-p)
(gui-selection-exists-p): Moved to lisp/term/w32-nt.el.
* lisp/term/w32-nt.el: New file, separated Windows native build
specific part from lisp/term/w32-win.el.
* lisp/loadup.el: Load term/w32-nt.el on Windows native build.
(Bug#75926)
* src/term.c (Ftty_frame_at):
* src/menu.c (syms_of_menu) <x-popup-menu-function>:
* src/dispnew.c (Fredisplay): Fix "doc:". The extra blank was
preventing 'make-docfile' from recognizing the functions and
variables and extracting their documentation to 'etc/DOC'.
(Bug#76722)
On Cocoa builds of NS Emacs, stipples are now rendered
using masked CGImages instead of patterned NSColors so that
stipples now render with color.
* src/nsimage.m ([EmacsImage stippleMask:]): Use a CGImageMask to
store the stipple mask when building for Cocoa.
* src/nsterm.m (ns_maybe_dumpglyphs_background): Perform a masked
fill to draw stipples when building for Cocoa.
(ns_draw_stretch_glyph_string): Perform a masked fill to draw
stipples when building for Cocoa.
Revert my commit 29a9fd4f4b
and the following commit 1ec0889e7b.
This fixes a bug where ‘emacs -nw’ would sometimes freeze when Emacs is
configured with ‘--with-pgtk --enable-link-time-optimization
--disable-gc-mark-trace’ on GNU/Linux x86-64 (Bug#76729).
As it is not yet clear whether this freeze is due to an Emacs bug that
I introduced, or due to GCC bug 117423
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117423>,
play it safe for now and revert to the previous state.
* src/androidterm.c (android_show_hourglass): Always define the
hourglass cursor, but...
(android_hide_hourglass): ... restore the invisible cursor
subsequently, if need be.
(android_toggle_visible_pointer): Do not hide the current cursor
if it is an hourglass.
* src/keyboard.c (kbd_buffer_store_selection_event_hold):
* src/keyboard.h (kbd_buffer_store_selection_event_hold): Declare
and define only for (HAVE_X11 || HAVE_PGTK).
