- Correct three pre-existing DEFVAR_BOOL doc strings (ns-use-native-
fullscreen, ns-use-mwheel-acceleration, ns-use-mwheel-momentum) that
were accidentally modified by an earlier rebase; restore original text
- Break @property line in nsterm.h to 79 chars
- Replace em-dash with --- and break 80-char comment line in patch 0004
- Issue 1: Add explicit ApplicationServices import for UAZoomEnabled/
UAZoomChangeFocus (was implicit via Carbon.h, now explicit)
- Issue 2: Rename FOR_EACH_FRAME variable 'frames' -> 'frame' (plural
was misleading; matches Emacs convention)
- Issue 3: Move unblock_input before ObjC calls in
postCompletionAnnouncementForBuffer: to avoid holding block_input
during @synchronized operations
- Issue 4: Fix DEFVAR_BOOL doc and Texinfo: initial value is nil,
not t; auto-detection sets it at startup
- Issue 5: Replace magic 10000 with NS_AX_MAX_COMPLETION_BUFFER_CHARS
constant with explanatory comment
- Issue 6: Add comment to lineStartOffsets loop explaining it is gated
on BUF_CHARS_MODIFF and never runs on the hot path
- Issue 8: Rewrite all 9 commit messages to GNU ChangeLog format with
'* file (symbol): description' entries
- Issue 9: Break 81-char @interface line in nsterm.h
- Issue 10: Add WINDOWP/BUFFERP guards before dereferencing
cf->selected_window and cw->contents in ns_zoom_find_child_frame_candidate
- Issue 11: Fix @pxref -> @xref at sentence start in macos.texi
accessibilityIndexForCharpos: walked composed character sequences
from run.ax_start up to the target charpos offset. For a run
covering an entire ASCII buffer, chars_in = pt - BUF_BEGV, making
each call O(cursor_position).
This method is called from ensureTextCache on EVERY redisplay frame
(as part of the cache validity check), making each frame O(position)
even when the buffer is completely unchanged. At line 34,000 of a
large file this is ~1,000,000 iterations per frame.
Fix: when ax_length == length for a run (all single-unit characters),
the ax_index is simply ax_start + chars_in. O(1) instead of O(N).
This is the symmetric counterpart to the charposForAccessibilityIndex:
fast path added in the previous commit. Both conversion directions
now run in O(1) for pure-ASCII buffers.
charposForAccessibilityIndex: walked composed character sequences
from the start of a visible run to the target AX index. For a run
covering an entire ASCII buffer, this is O(cursor_position): moving
to line 10,000 requires ~500,000 iterations per call.
The method is called on every SelectedTextChanged notification
response (accessibilityBoundsForRange: from the AX server for cursor
tracking), making cursor movement O(position) in large files.
Fix: when ax_length == length for a run (all characters are single
AX index units — true for all ASCII/Latin text), the charpos offset
is simply ax_idx - run.ax_start. O(1) instead of O(position).
Multi-byte runs (emoji, CJK, non-BMP) fall back to the sequence walk,
bounded by run length (visible window size), not total buffer size.
Root cause: the 50ms rate limit broke child-frame (Corfu) tracking.
When the Corfu child frame redraws, its ns_update_end fires first and
resets the rate-limit timer. When the parent frame's ns_update_end
fires immediately after, the timer has not expired, so
ns_zoom_track_completion returns early without scanning child frames.
Zoom focus stays on the first candidate.
Fix: remove the rate limit; add a FRAME_PARENT_FRAME(f) guard instead.
Child frames have no completion children to scan; their parent's
ns_update_end does the scan via FOR_EACH_FRAME. Returning early on
child-frame calls avoids the redundant scan and leaves the timer
problem moot. Overhead without the rate limit is ~40 Lisp evaluations
per redisplay (~5-20 µs), acceptable given ns_zoom_enabled_p() already
caches the UAZoomEnabled() IPC call.
Fget_char_property with a buffer as OBJECT checks text properties
only. Corfu highlights the selected candidate (corfu-current) via
an overlay, not a text property, so the scan always returned -1 and
Zoom focus stayed on the first line.
Pass cf->selected_window instead of cw->contents so that overlays
are included in the property lookup. Vertico uses text properties
and is unaffected; child-frame completion frameworks that use overlays
(Corfu, Company-box) now track correctly.
Changes:
- EmacsApp gets ns_update_accessibility_state and
ns_accessibility_did_change: methods (patch 0005)
- At startup: UAZoomEnabled() + AXIsProcessTrustedWithOptions()
determine initial ns_accessibility_enabled state
- com.apple.accessibility.api distributed notification updates it
whenever any AT connects or disconnects
- All Zoom call sites (UAZoomChangeFocus) now gated by
ns_accessibility_enabled in addition to ns_zoom_enabled_p()
- ns-accessibility-enabled docstring updated to describe auto-detect
Result: zero config needed; zero overhead when no AT is active;
single variable overrides auto-detection when needed.
500ms (2 Hz) was too aggressive — Zoom focus stopped updating during
keyboard navigation in Vertico/Corfu lists. 50ms (20 Hz) tracks
fast arrow-key navigation while still avoiding per-frame overhead.
UAZoomEnabled() is already cached so the main cost is the overlay
scan, which is cheap.
The Qnil initialization was in patch 0000 (Zoom) but the ivar
declaration is in patch 0008 (child frame tracking). Moved the
init to patch 0008 so each patch compiles independently.
Three fixes:
1. Patch 0000 now compiles standalone: replaced forward declaration
of ns_ax_face_is_selected (defined in VoiceOver patches) with
self-contained ns_zoom_face_is_selected in the Zoom patch.
2. ns_accessibility_enabled defaults to nil: eliminates ALL VoiceOver
overhead (text cache rebuild, AX notifications, Mach IPC to AX
server) when VoiceOver is not in use. Zero per-redisplay cost.
Enable with (setq ns-accessibility-enabled t).
3. UAZoomEnabled() cached for 1s + ns_zoom_track_completion rate-
limited to 2Hz: eliminates 150-600µs/frame of IPC overhead.
Root cause (per Opus analysis): UAZoomEnabled() is a synchronous
Mach IPC roundtrip to macOS Accessibility server, called 3x per
redisplay cycle. At 60fps = 180 IPC roundtrips/second blocking the
main thread. Combined with Emacs's inherent O(position) redisplay
cost, this compounded into progressive choppy behavior.
Fix 1: ns_zoom_enabled_p() caches UAZoomEnabled() for 1 second.
Fix 2: ns_zoom_track_completion() rate-limited to 2 Hz.
Also includes BUF_CHARS_MODIFF fix (patch 0009) for VoiceOver cache.
BUF_CHARS_MODIFF fix — the core performance regression:
ensureTextCache checked BUF_MODIFF which font-lock bumps on every
redisplay. Each cursor movement in a large file triggered full buffer
rebuild. Now uses BUF_CHARS_MODIFF (changes only on char insert/delete).
Performance issue: editing large files (>~10KB, >2000 lines) caused
progressive slowdown regardless of VoiceOver status.
Root causes:
1. ns_zoom_find_overlay_candidate_line: called Foverlays_in on the
entire visible buffer range on every redisplay when UAZoomEnabled().
In files with many overlays (font-lock, hl-line, show-paren etc.)
this was O(overlays) Lisp work per keystroke.
2. postAccessibilityNotificationsForFrame: when ns-accessibility-enabled
is non-nil, checked BUF_OVERLAY_MODIFF every redisplay. font-lock
bumps this on every redraw, triggering ns_ax_selected_overlay_text
(another O(overlays) scan) for non-minibuffer windows.
Fix: Both scans now guard with MINI_WINDOW_P check. Overlay completion
frameworks (Vertico, Icomplete, Ivy) only display candidates in
minibuffer windows --- no completion framework puts selected-face
overlays in normal editing buffers. For non-minibuffer windows both
functions return immediately with zero Lisp calls.
Additionally: ns_zoom_find_child_frame_candidate is skipped when
f->child_frame_list is nil (no child frames = no Corfu popup).
Zoom patch 0000 now tracks completion candidates:
- Overlay: Vertico, Icomplete, Ivy (face heuristic on before-string)
- Child frame: Corfu, Company-box (scan buffer text for selected face)
Also fixes duplicate lastCursorRect ivar when applied with VoiceOver.
Zoom (0000) declares lastCursorRect @public in EmacsView.
VoiceOver (0005) was re-declaring it, causing 'duplicate member'
compiler error when both applied together. Removed the duplicate.
VoiceOver patches 0001-0008 now apply cleanly on top of Zoom patch
0000. The full set (git am patches/000*.patch) works without
conflicts. Patch 0005 (integration) merges Zoom fallback and
VoiceOver postAccessibilityUpdates in ns_update_end.
The ivar was declared in patch 0001 but first used in patch 0005,
creating dead code in intermediate commits 0001-0004. Now each
commit only introduces declarations that are immediately used.
Patch 0007 bulk em-dash→triple-dash replacement accidentally changed
windowWillResize title format string and strstr search, introducing
a user-visible regression. Reverted those two lines to em-dash.
Fixes from Opus maintainer review:
1. [BLOCKER] Zoom code completely removed from ALL intermediate patches
(0005-0007 no longer have UAZoom/overlayZoom at any commit point)
2. [BLOCKER] Unified cursor rect ivar: lastCursorRect (was split
between lastZoomCursorRect and lastAccessibilityCursorRect)
3. [HIGH] Child frame static vars moved to EmacsView ivars
(childFrameLastCandidate/Buffer/Modiff — no cross-frame interference)
4. [HIGH] intern_c_string replaced with Qbefore_string/Qafter_string
5. [MEDIUM] Zoom fallback gated by zoomCursorUpdated flag (no double call)
Major changes:
1. Zoom separated into standalone patch 0000
- UAZoomChangeFocus in ns_draw_window_cursor
- Fallback in ns_update_end for window-switch tracking
- No overlayZoomActive (source of split/switch/move bug)
2. VoiceOver patches 0001-0008 are now Zoom-free
- All UAZoom*, overlayZoom*, kUAZoomFocus references removed
- lastAccessibilityCursorRect kept for VoiceOver bounds queries
- Commit messages cleaned of Zoom references
3. README.txt and TESTING.txt rewritten for new structure
Addresses reviewer (Stéphane Marks) feedback:
- Keep Zoom patch separate from VoiceOver work
- Design discussion needed for non-Zoom patches
- Performance: ns-accessibility-enabled=nil for zero overhead
BLOCKER fixes:
1. Remove duplicate ns_ax_face_is_selected, ns_ax_selected_overlay_text,
ns_ax_selected_child_frame_text definitions from patch 0002
(now defined only in 0007/0008 where they belong)
2. Fix idx → point_idx in accessibilityInsertionPointLineNumber (0002)
3. Remove stale 100K cap reference from documentation (0006)
Architecture fix:
- ns_ax_selected_child_frame_text moved from 0007 to 0008
(where it logically belongs)
Verified: all 8 patches apply cleanly on fresh emacs HEAD.
- 0001: remove NS_AX_TEXT_CAP (100K char cap), add lineStartOffsets/
lineCount ivars and method declarations to nsterm.h
- 0002: add lineForAXIndex:/rangeForLine: O(log L) helpers, build line
index in ensureTextCache, replace O(L) line scanning in
accessibilityInsertionPointLineNumber/accessibilityLineForIndex/
accessibilityRangeForLine, free index in invalidateTextCache/dealloc
- 0009 deleted (folded into 0001+0002)
- README.txt: remove NS_AX_TEXT_CAP references, update known
limitations, stress test threshold 50K lines
Move file-scope statics (lastChildFrameBuffer/Modiff/Candidate) and
ns_ax_reset_accessibility_updating before announceChildFrameCompletion.
Using them before their declaration was a forward reference (UB in C).
Remove two spurious blank lines at start of announceChildFrameCompletion
method body.
New patch 0009 fixes HIGH severity issues from Opus review:
- Announcement coalescing (50ms debounce)
- cachedText retain+autorelease in accessibilityValue
- EmacsView dealloc: nil out emacsView on all AX elements
- Nil guards on protocol methods + overlayZoomActive
0007 updated: revert accidental em-dash→triple-dash, add overlayZoomActive nil guards
0008 updated: specpdl exception safety for accessibilityUpdating, lastChildFrameBuffer staticpro
Series now 9 patches total (0001-0006 unchanged, 0007-0009 new/updated).
BLOCKER #1: accessibilityUpdating flag exception safety.
A Lisp signal (longjmp) during postAccessibilityUpdates left
the re-entrance flag permanently YES, suppressing all future
AX notifications → VoiceOver goes silent randomly.
Fix: specpdl unwind protection (record_unwind_protect_ptr)
resets the flag on any longjmp. All 3 exit points use unbind_to.
BLOCKER #2: static struct buffer *lastBuffer dangling pointer.
Raw C pointer to buffer struct has no GC protection. After
kill-buffer, the pointer dangles.
Fix: file-scope Lisp_Object lastChildFrameBuffer with staticpro.
EQ comparison instead of pointer equality.
Also: revert accidental em-dash → triple-dash in title bar (0007),
fix README factual error (BUF_OVERLAY_MODIFF cache key).
hl-line-mode (and similar) bumps BUF_MODIFF via text property
changes on every cursor movement. The else-if structure caused
the modiff branch to fire (skipping ValueChanged correctly) but
also blocked the cursor-move branch (SelectedTextChanged).
Fix: use textDidChange flag to decouple the two branches.
ValueChanged and SelectedTextChanged remain mutually exclusive
for real edits, but SelectedTextChanged now fires when only
text properties changed.
- childFrameCompletionActive flag: set by child frame handler,
cleared when no child frame visible on parent's next cycle
- Posts FocusedUIElementChanged on parent buffer element to
restore VoiceOver character echo after corfu closes
- corfu-auto-delay: 1.0 → 2.0 (reduce popup noise)
Fbuffer_substring_no_properties operates on current_buffer, not the
passed buffer. Added set_buffer_internal_1 + record_unwind_current_buffer
with unbind_to on every return path.
