| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
The message duplicated the modeline directly above it: the buffer name and the eat state icons already announce which agent is focused, so the echo just lingered as clutter. The no-other-agents echo stays, since it reports something the modeline can't.
|
| |
|
|
|
|
| |
The modeline now follows one layout rule: the left side is Emacs information (mode icon, eat state, modified/read-only, buffer name, @host, Narrow, VC branch, position, MACRO, process) and the right side is a systray for package indicators (recording, flycheck counts, misc-info). The VC branch and mode-line-process moved left to fit the rule.
eat buffers trade the [semi-char]:run text for two icons beside the mode icon: a keyboard glyph for the input mode (quiet when semi-char, warning otherwise, hover text explains where keys go, mouse-1/2/3 switch modes mirroring eat's own bindings) and a green play / red power-off for the process state. eat-config clears eat's buffer-local mode-line-process so nothing renders twice.
|
| |
|
|
| |
cj/markdown-preview now brings up the simple-httpd listener itself instead of signaling a user-error pointing at cj/markdown-preview-server-start. The two-command flow guarded against a surprise network listener, but the preview is the only reason the listener exists, so the guard was just an extra step. I kept the start command for manual use.
|
| |
|
|
| |
A grazed F2 followed by a stray 2, s, or b invokes two-column mode, which replaces the buffer's mode-line-format with its own retro layout and spawns a 2C/ companion buffer. That's what mangled the modeline in an agent terminal: eat semi-char buffers don't forward F2, so the prefix was armed everywhere. Nothing of mine uses global F2 (markdown-mode's binding lives in its own map), so I unbound both routes to 2C-command. cj/modeline-reset repairs a hijacked buffer.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
I rebuilt the custom modeline as pure segment helpers with thin :eval wiring:
- The nerd-icons mode icon replaces the mode-name text (cached per buffer, plain name on terminal frames), with the full mode name in the help-echo.
- New left-side segments: modified dot / read-only lock (file buffers only), remote @host tag, Narrow tag that widens on click, point-based percentage, region selection info, and a MACRO tag while a keyboard macro records.
- New right-side segments: mode-line-process (eat and compilation state was invisible) and flycheck per-severity counts with click-through to the error list, replacing the stock status text. Glyphs are nerd-icons private-use codepoints so emojify can't rewrite them, with text fallbacks when icons are unavailable.
- cj/modeline-reset kills a hijacked buffer-local mode-line-format (two-column mode, ediff, calc).
- Optional taller bar via cj/modeline-height-factor, a display height property on the padding space so the theme's mode-line faces stay untouched.
- Housekeeping: the dead user-constants require and stale commentary are gone, cj/modeline-vc-faces left the risky-local-variable list, and the cache-key defun is cj/--modeline-vc-cache-key so it no longer shadows the same-named defvar.
Percent signs from :eval strings go through the mode-line %-construct pass, so the position segment emits %% for a literal percent.
|
| |
|
|
|
|
|
|
| |
ai-term.el had grown to ~1,215 lines mixing project/tmux session discovery, window display policy, the EAT terminal backend, and the public commands, so a change to any one risked coupling to the others.
I extracted three layers, following the calendar-sync split shape: ai-term-sessions.el (discovery, tmux naming and parsing, launch command, picker ordering), ai-term-display.el (display-buffer actions and rule, toggle state, server-window routing), and ai-term-backend-eat.el (terminal create/reattach, pty send, EAT keymap). The backend file is named for its backend so a future one lands as a sibling. ai-term.el stays the public face (options, commands, keymap, shutdown), every name unchanged, so existing (require 'ai-term) callers and all 30 test files work as before.
The extracted layers forward-declare the face's defcustoms rather than requiring it, keeping the graph acyclic. I dropped the unused cl-lib and host-environment requires and added the three modules to the header-contract list.
|
| |
|
|
| |
Delete file (D) ran with no confirmation at all; erase, clear-to-top/bottom, and revert were a single keystroke from destroying unsaved edits; and raw revert-buffer prompted even when there was nothing to lose. Policy now: delete always confirms, naming the file (the VC path keeps vc-delete-file's own prompt); erase/clear/revert confirm only when a file-visiting buffer has unsaved edits, and stay fast otherwise. The delete workhorse is split into an unconfirmed internal so its existing tests keep exercising the file mechanics; 13 new tests cover the policy.
|
| |
|
|
| |
treesit-auto-install was t, so opening a file could silently trigger a network download and compiler build mid-edit. It now prompts, and cj/install-treesit-grammars is the deliberate fresh-machine bootstrap that installs everything in one command.
|
| |
|
|
| |
cj/org-babel-toggle-confirm landed on C-; k as a placeholder. It's an org-babel concern, so it now lives on the org menu as C-; O b, and C-; k is free again. The binding registers after org-config loads so a standalone load of this module still works.
|
| |
|
|
| |
With a single agent open and focused, the rotation wrapped back to the same agent and echoed a misleading "Agent: <name>" as if it had swapped. Now it says there are no other ai-terms to switch to. A sole agent that is displayed but not selected still gets selected, and the no-agents picker fallback is unchanged.
|
| |
|
|
| |
The wf-recorder integration tests passed video-recordings-dir to the real capture pipeline, and the test file's defvar stub of that variable is clobbered by user-constants' defconst on load — so every suite run left ~0.5s screen captures in the real recordings directory. Pass an explicit make-temp-file dir and delete it in the unwind, matching what the capture probe already did.
|
| |
|
|
| |
At speed 3 the native compiler emits direct calls for functions in the same compilation unit, bypassing the symbol's function cell. Any cl-letf mock of a module's own helper then silently runs the real code: the recording tests' mocked wayland check and device validation were bypassed, and make test launched real wf-recorder screen captures. Speed 2 is the highest level that preserves redefinition semantics. A meta test now pins the setting; the local eln cache needs one flush so stale speed-3 artifacts recompile.
|
| |
|
|
|
|
| |
EAT 0.9.4's parser accepts more charset-designation final bytes than its store step maps. A designation like ESC ( A (UK) isn't one of the two it handles ("0" and "B"), so it stores nil as that slot's charset. The next character written then fails (cl-assert charset) in eat--t-write. Since writes run off the output-queue timer, it repeats once per output chunk. An agent terminal that emits one of these bytes throws "cl-assertion-failed (charset)" hundreds of times and stops rendering.
I added filter-args advice on eat--t-set-charset that coerces a nil charset to us-ascii before it's stored, so an unmapped designation falls back to plain ASCII instead of wedging. Patching the vendored pcase would be cleaner, but a package update reverts it. The advice loads with eat, since the target is an internal function.
|
| |
|
|
|
|
| |
Audio-only recordings were written as AAC in an MP4/.m4a container. The stop path SIGINTs ffmpeg, and if the MP4 muxer doesn't write its moov trailer before exit, the file has no moov atom and won't decode. ffmpeg and AssemblyAI both reject it. Three recordings were lost that way and had to be rebuilt with untrunc. The video path already avoids this by using Matroska, which needs no finalize pass.
I switched the audio-only path to FLAC. FLAC frames are self-contained, so an abruptly stopped recording still decodes, with no trailer to miss at close. It's also lossless, dropping the 64k AAC encode that degraded speech before transcription. AssemblyAI recommends a lossless source and accepts FLAC directly. The transcription path passes audio files through untouched.
|
| |
|
|
|
|
| |
A timer tick that fired while a calendar's previous fetch was still running launched a second concurrent sync for that calendar, wasting work and racing to write the same org file. The dispatcher now skips a calendar whose status is already syncing and logs the skipped tick.
The sentinel resets the status on process exit, so the skip clears on its own. load-state also clears a stale syncing status left by a crash, so a calendar can't be skipped forever.
|
| |
|
|
| |
cj/make-script-executable runs from a global after-save-hook and set +x on any saved file whose first line was a shebang, in every buffer. A downloaded script you were reading, a template, or a shebang in a text or org file silently became executable. I gated it on derived-mode-p prog-mode, so it only acts on actual script buffers. Real scripts (sh-mode, python-mode) still get the fast path.
|
| |
|
|
|
|
| |
The live markdown preview pulled strapdown.js from http://ndossougbe.github.io over plain HTTP. That broke the preview with no network, loaded third-party JS over an unencrypted connection (mixed content, MITM), and trusted an unmaintained github.io page against the localhost preview.
I vendored the self-contained bundle (jQuery, marked, bootstrap themes) into assets/strapdown.js and embed it inline. The whole preview now serves from localhost and works offline. cj/markdown-html reads the file once and caches it.
|
| |
|
|
|
|
|
|
| |
Emacs crashed at launch with wrong-number-of-arguments on cj/save-some-buffers, down the startup path (dashboard-only to kill-all-other-buffers to save). Two modules defined a function by that name: custom-buffer-file.el's legible save prompt (arg + pred), installed as an override on save-some-buffers, and undead-buffers.el's older one-arg wrapper that called save-some-buffers internally. custom-buffer-file loads first, undead-buffers second. The one-arg version won the shared symbol, so the override re-entered it with two args.
I removed undead-buffers.el's wrapper. cj/kill-all-other-buffers-and-windows now calls the standard save-some-buffers with the undead predicate, which routes through the override when loaded and the built-in otherwise, so undead-buffers no longer depends on custom-buffer-file. The legible override keeps the cj/save-some-buffers name.
A regression test loads both modules in launch order and guards the call and the arity so a one-arg shadow can't return.
|
| |
|
|
|
|
| |
Declining one occurrence of a recurring meeting left it on the agenda. Google emits that decline as a RECURRENCE-ID override carrying the user's PARTSTAT=DECLINED. But calendar-sync--parse-exception-event never read the override's attendee block, so the occurrence kept the series' inherited "accepted" status and the declined filter never dropped it. The apply side already re-derives status from an override's attendees. The parse side just wasn't supplying them.
The fix parses the override's ATTENDEE lines into :attendees, the same way parse-event does. A unit test covers the extraction. An integration test runs the full parse/apply/filter chain on a declined week.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
disk-changed saves
Replace the cryptic single-key save prompts with read-multiple-choice menus whose options are labeled on screen instead of recalled as keys.
save-some-buffers (C-x s and save-on-exit) is overridden with a reimplementation that reuses the stock candidate selection and abbrev-save tail but swaps map-y-or-n-p's terse key list for a labeled menu, and adds a clean-whitespace-and-save action.
C-x C-s gains a wrapper that fast-paths normal saves to save-buffer; only when the buffer has unsaved edits and the file changed on disk does it show a labeled menu (save / diff / clean / revert / cancel) instead of the bare "Save anyway?" yes/no.
Both menus share a diff toggle, and whitespace-only diffs route to a plain unified diff with trailing whitespace highlighted, since difftastic normalizes trailing whitespace and renders such changes blank.
The interactive wrappers stay thin over pure, tested helpers: the save-loop planner, the key-to-action maps, the whitespace detector, and the renderer choice.
|
| |
|
|
|
|
| |
Font size now carries across books and sessions. The +/-/= keys write the text-scale offset to data/nov-reading-text-scale and the offset is restored when a book opens, so a size I set sticks instead of resetting to the base height on every reopen. The = key returns to the base height and persists that reset.
Each palette grows from a single bg/fg face into a bundle: :face plus optional :heading and :link. When a palette is active, its heading and link faces remap shr's h1-h6 and link faces buffer-local, so the EPUB hierarchy reads in the palette's accent. The remap stays buffer-local to the nov buffer, so HTML mail and eww keep the theme's normal shr colors.
|
| |
|
|
|
|
|
|
| |
EPUB reading prefs were scattered: a hardcoded Merriweather/180 font-remap in calibredb-epub-config's nov hook, no color control (the old sepia foreground had been stripped), and a frame-global EBook fontaine preset as the only way to size up. That preset resized the font in every buffer in the frame, not just the book.
I pulled the reading view into its own layer, modules/nov-reading.el, on top of stock nov (no fork). It owns three things, all buffer-local: a reading palette (sepia/dark/light, each a face the dupre theme owns, sepia the default), the serif typography (family plus a defcustom base height replacing the hardcoded 180), and page font sizing (+/- bump the size live, = resets to the base). Width moves to { }. calibredb-epub-config keeps the library and width/centering layout. Its nov hook now calls into the layer.
The three palette faces register as a nov-reading app in theme-studio (face_data.py), so they're tunable there like any other app. I dropped the EBook fontaine preset, since reading size is buffer-local now.
|
| |
|
|
|
|
| |
calendar-sync ran calendar-sync-start at load, which syncs immediately and then every interval. Every configured calendar resolves its .ics feed URL from a :secret-host in authinfo.gpg, so both the immediate sync and each timer tick decrypt authinfo.gpg. On a cold gpg-agent (after a reboot) that surfaced as a GPG passphrase prompt right after startup, before I'd asked for anything needing a secret.
I deferred the whole start, immediate sync and recurring timer alike, to the first org-agenda use via a one-shot org-agenda-mode-hook. The unlock now happens when I actually open the agenda. A manual calendar-sync-start or calendar-sync-now still works on demand.
|
| |
|
|
| |
nerd-icons already rendered every icon in the config (dashboard, dirvish, ibuffer, completion). all-the-icons survived only as scaffolding: a font-install helper, the all-the-icons-nerd-fonts bridge, and a terminal-blanking advice block the nerd-icons one beside it already duplicated. I removed all of it and pointed the font-install helper at nerd-icons (Symbols Nerd Font Mono), keeping the auto-install-on-first-GUI-frame convenience. I updated the font-config tests to the renamed helper.
|
| |
|
|
|
|
|
|
|
|
| |
The Super+N quick-capture popup frame is named "org-capture" and was torn
down by deleting the selected frame on capture exit. When the daemon's
selected frame was something else at finalize (common with multiple frames),
the real capture frame survived and lingered, showing whatever buffer was
behind it. Reap by frame name across all frames instead, sparing any popup
still mid-capture (*Org Select* or a CAPTURE-* buffer), and expose
cj/org-capture-reap-popup-frames for manual cleanup.
|
| |
|
|
|
|
|
|
|
|
| |
Eight completing-read pickers listed bare file basenames, so marginalia had no directory to resolve and couldn't annotate them. Add cj/completion-file-annotator to system-lib — an annotation-function factory that takes a candidate->path resolver and yields a size + modification-date suffix (or "dir" for directories, nil for missing files). Wire each picker through cj/completion-table-annotated with a per-site category and resolver: timer sounds, drill flashcards, Info files, the test-runner focus add/remove, vc clone dirs, hugo drafts, and agenda projects (the project's todo.org mtime). music-config's existing completion table gains the category and annotator inline, keeping its sort metadata.
The candidate strings and every return value are unchanged — this only adds completion metadata — so all downstream logic is untouched. The six modules that didn't already pull in system-lib now require it.
Tests: cj/completion-file-annotator gets Normal/Boundary/Error coverage (file, directory, nil path, missing file). Full suite green at 5394.
Claude-Session: https://claude.ai/code/session_014fyKMTTqLrZpL3rDF3dYc3
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Break the 1025-line video-audio-recording.el into a thin public face plus two layered libraries, moving every function verbatim so behavior and public names are unchanged:
- video-audio-recording-devices.el — base layer: PulseAudio source and sink discovery, the pactl output parsers, device labeling and sort/status helpers for completing-read, and the lookup predicates. Pure string and shell-query helpers with no dependency on recording state, config, or the engine. This is the heavily-tested core.
- video-audio-recording-capture.el — engine: ffmpeg/wf-recorder command construction, the recording process lifecycle (sentinel, producer-first shutdown, exit polling), the modeline indicator, dependency checks, device acquisition and validation, and the start/stop entry points. Requires the devices layer.
video-audio-recording.el keeps configuration and the recording process-handle state, the device-diagnostic and device-test commands, the toggle commands, and the C-; r keymap, and requires the two layers. The engine reads and updates the config and process-handle variables, which the top module owns, through forward declarations, so no layer requires the top module back. No function call crosses from a layer up into the top module, so the split needs no forward-declared functions.
Every public name is preserved, so all 323 existing video-audio-recording tests pass unchanged through the require chain. The two new modules carry the load-graph and package headers and join the header-contract allowlist.
Claude-Session: https://claude.ai/code/session_014fyKMTTqLrZpL3rDF3dYc3
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Break the 1724-line calendar-sync.el into a thin public face plus four layered libraries, moving every function verbatim so behavior and public names are unchanged:
- calendar-sync-ics.el — base parsing: RFC 5545 text cleaning, VEVENT property extraction, attendee/organizer/URL parsing, timezone and timestamp conversion, date arithmetic, single-event parsing. Depends on neither of the other new modules.
- calendar-sync-recurrence.el — RRULE/EXDATE/RECURRENCE-ID expansion.
- calendar-sync-org.el — Org rendering and atomic file output.
- calendar-sync-source.el — sync state and persistence, async .ics fetch, the batch conversion worker, and the Google Calendar API path.
calendar-sync.el keeps configuration, the parse orchestrator, sync dispatch, the user commands, the timer, and the C-; g keymap, and requires the four layers. Each layer forward-declares the config defvars it reads, so no layer requires the top module back. The batch worker loads the whole graph, so source forward-declares the two functions it calls there.
Every public name is preserved, so all 574 existing calendar-sync tests pass unchanged through the require chain. The four new modules carry the load-graph and package headers and join the header-contract allowlist.
Claude-Session: https://claude.ai/code/session_014fyKMTTqLrZpL3rDF3dYc3
|
| |
|
|
|
|
| |
test-system-cmd-restart-emacs-no-service-aborts mocked cj/system-cmd--emacs-service-available-p directly, but cj/system-cmd-restart-emacs reaches that helper through a native-comp intra-file direct call that bypasses the symbol-function redefinition. The mock silently no-opped, the real check ran, and on any machine that has emacs.service it returned non-nil, so restart proceeded and the should-error saw no error.
I mock executable-find to nil instead, driving the real helper to nil at a boundary its subr trampoline honors, matching the passing sibling tests. The test now passes whether or not emacs.service is installed.
|
| |
|
|
|
|
| |
custom-misc.el was an incoherent grab-bag, so anything small defaulted to landing there. I split its eight commands by concern. Three moved into new modules: custom-format (region/buffer reformat), custom-counts (word and character counts), and custom-text-transform (fraction glyphs). The other three went to existing homes: the previous-buffer toggle to custom-buffer-file, the delimiter jump to custom-line-paragraph, and the align-regexp space advice with its alignment and fill bindings to custom-whitespace.
The C-; bindings, which-key labels, and the six test files moved with their functions, and custom-misc.el is deleted. No behavior change: every command keeps its name and its C-; key.
|
| |
|
|
|
|
| |
Two owned helpers carried unprefixed generic names that risk colliding in the single Emacs namespace: car-member (local-repository.el) and unpropertize-kill-ring (system-defaults.el). I renamed them to localrepo--car-member and cj/--unpropertize-kill-ring and updated their callers and tests. Both are non-interactive and contained, so no alias was needed.
docs/design/naming-audit.org records the rest of the scan: the allowlist of deliberate module prefixes, the foreign forward-declarations that aren't owned definitions, and a deferred list (keybound commands, the with-timer macro, the ui-theme defcustoms, the user-constants paths) that each want a focused pass rather than an unattended rename.
|
| |
|
|
|
|
| |
The first-line header on 33 modules named the file without its .el extension (;;; font-config --- ... rather than ;;; font-config.el --- ...), the form checkdoc and package-lint expect and the other modules already use. I normalized all 33 to the canonical ;;; name.el --- summary shape. The change is line 1 only.
A new test, test-meta-package-headers.el, locks the convention. It checks every module for the canonical first line, Commentary before Code, a provide footer, and no BOM, and unit-tests the checker against each malformed shape so the guard itself is proven.
|
| |
|
|
| |
The route-colors pass dropped the literal cj/music-* face definitions but left the playlist header referencing them, so every header render spammed "Invalid face reference". I restored the five as deffaces that inherit themed base faces, so the theme still owns their colors. A test asserts each referenced face is defined.
|
| |
|
|
|
|
| |
F1 (cj/dashboard-only) kills every other buffer, burying only those on the undead list. Agent buffers were never registered, so the sweep killed live agents and detached their sessions.
Agent buffers are a dynamic family ("agent [<project>]") that an exact-name list can't pre-enumerate, so undead-buffers gains a regexp list (cj/undead-buffer-regexps) and a centralized cj/--buffer-undead-p predicate. Both kill paths route through it. ai-term registers the "agent [" pattern, so every agent -- current or future, however created -- is buried rather than killed.
|
| |
|
|
|
|
| |
Every calibredb launch (the dashboard "b", M-x, anywhere) now opens filtered to the in-progress books rather than the whole library, via an :after advice on calibredb. Clear with L or x to see everything.
The filter scopes to the tag field (calibredb-tag-filter-p), not a bare keyword search. A bare keyword matches any field, which surfaced books that only mention "in-progress" in their description.
|
| |
|
|
|
|
| |
Bookmark reformatting to "Author, Title" was nov-only, so PDF bookmarks kept the raw filename. PDFs open in pdf-view-mode, whose pdf-view-bookmark-make-record carried no advice.
I added a parallel :filter-return advice there, reusing the same extension-agnostic filename parser, and renamed the helpers off the nov- prefix to reading- since they now serve both EPUB and PDF. New tests cover a PDF filename and a PDF-shaped record.
|
| |
|
|
|
|
| |
A Weather launcher joins the dashboard's top row after Agenda, on key w, drawn with the nf-weather-day_sunny_overcast Weather Icons glyph. It opens the wttrin forecast through call-interactively so wttrin's location prompt runs. A bare (wttrin) call errors, since the command takes the location as a required argument that its interactive form supplies.
Row sizes move from 4-4-3-3 to 5-4-3-3. The launcher table stays the single source for both the navigator icons and the keymap.
|
| |
|
|
|
|
| |
Summoning the agent (M-SPC) into a single-window frame docked it at the default fraction even when it was last fullscreen, because the size-memory model only captured split geometry at toggle-off and a sole window has no dock size to record. A window-configuration-change-hook tracker now records whether the displayed agent fills its frame (cj/--ai-term-last-fullscreen), and display-saved restores it in place when that flag and a single-window frame both hold.
The tracker records only that flag, not dock geometry: re-capturing the dock size on every window change fed a capture/replay loop that drifted the dock height a couple rows per cycle. The restore guard uses (one-window-p t) so an active minibuffer (a picker prompt mid-summon) isn't counted as a second window, which otherwise misfired the restore into a dock and cascaded.
|
| |
|
|
|
|
| |
M-<arrow> now mirrors C-; b <arrow>. From a sole window it pulls a sliver split open on the opposite edge, revealing the previous buffer. In a multi-window layout it nudges the divider via windsize.
cj/window-resize-sticky derives the arrow with event-basic-type, so one command serves both chords. M-up/down were unbound, and M-left/right shed word-motion, which stays on C-left/right. org and other modes that own M-arrow still shadow it, so C-; b remains the universal binding.
|
| |
|
|
| |
More of the commentary/comment audit. The custom-* command modules, weather-config, and mousetrap-mode had empty package summary lines and verbose summary paragraphs. I filled the summaries and condensed the prose while keeping each module's load-contract metadata. dwim-shell-config and auth-config had malformed ;; headers (now ;;;), local-repository had a leading BOM, and two test files had blank summaries.
|
| |
|
|
| |
C-; ! was a prefix map with per-command leaf keys (s/r/e/l/L/E/S) I rarely use. I bound C-; ! directly to the completing-read menu instead and removed the leaf keys, reclaiming the real-estate. Every command stays reachable through the menu. Updated the commentary and the two keymap tests to the new contract.
|
| |
|
|
|
|
| |
make-process :stderr was a file path, but :stderr takes a buffer, so Emacs made a buffer named after the path instead of writing the file. The "Errored. Logs in <file>" notification pointed at a log with no error text, and that hidden buffer leaked one per transcription.
I now route stderr through an explicit, erased buffer: passed to :stderr, threaded to the sentinel, drained into the log, then killed. That keeps stderr off the stdout :buffer, so the transcript stays clean.
|
| |
|
|
| |
Claude Code and similar inline TUIs sometimes truncate a colored span without emitting a reset; the unterminated color then bleeds onto every following line in the EAT buffer. Advise eat-term-process-output to inject an SGR reset before each newline, containing the bleed to its own line. Validated in real EAT and via an ansi-color proxy: it contains the bleed and leaves per-line coloring intact, since programs re-open their color on each line. Gated by cj/eat-reset-sgr-at-newline (default on) so it can be disabled if a program ever carries one color across newlines without re-opening it.
|
| |
|
|
| |
Claude Code (and any Ink-style inline TUI) moves the terminal cursor up to redraw its whole block and back to the bottom on every tick; EAT follows the cursor with point, so the Emacs window chases it up and down -- the bounce. Add cj/--eat-tame-scroll on eat-mode-hook: scroll-conservatively 101, scroll-margin 0, and auto-window-vscroll nil, so the window line-scrolls minimally instead of recentering. It doesn't remove the bounce (the inline redraw is the root) but makes each jump gentler.
|
| |
|
|
| |
dired's d was dired-flag-file-deletion and D was dired-do-delete, while the ediff diff sat on e. Bind d to the ediff diff and keep D as delete, so the d=diff / D=delete pair is consistent with C-; b and ibuffer. d no longer flags for deletion -- mark with m for batch deletes, then D. Fix the stale commentary (it called d "delete" and D "duplicate") and add which-key labels for the pair.
|
| |
|
|
| |
C-/M-left/right were in EAT's default eat-semi-char-non-bound-keys, so they fell through to Emacs and ran left-word/right-word, moving point in the EAT buffer instead of being sent to the program. The terminal's own cursor never moved, so the next keystroke snapped point back to the real cursor -- the "cursor jumps back" symptom when editing claude's input. Bind them to eat-self-input so they forward as word motion, the way ghostel did. Window arrows (S-, C-M-) still reach Emacs for windmove and buffer-move.
|
| |
|
|
| |
EAT's semi-char mode left the bare escape key unbound and treated ESC only as the Meta prefix, so a lone Escape never reached the pty. That is why C-<up>'s tmux copy-mode could not be exited with Escape: tmux's own Escape=cancel binding never saw the key. Bind <escape> to forward ESC to the terminal, so it cancels tmux copy-mode and still works in TUIs like vim. Also bind <escape> in eat-mode-map to return to semi-char, so the same key exits EAT's own emacs and char modes. One exit key for both copy views; q is no longer required.
|
| |
|
|
| |
In the ibuffer buffer list, d now diffs the buffer at point against its saved file (ibuffer-diff-with-file, was on =) and D marks it for deletion (was on d; x still executes the marks).
|
| |
|
|
| |
Remove the dead ghostel app from theme-studio: the GHOSTEL_FACES/SEED data, the registry row, the renderGhostelPreview previewer, and the package_seed test, then regenerate the tool. ansi-color stays since eat inherits it. Rename testutil-ghostel-buffers to testutil-terminal-buffers and drop make-fake-ghostel-buffer; the toggle-filter test now uses the eat fixture, since agents are eat. Fix the comments that still called the agent buffers ghostel (they're eat now) in eat-config and the ai-term and auto-dim test docstrings. I also package-deleted the unused ghostel ELPA package. Full suite green; the remaining ghostel mentions are accurate migration history.
|
| |
|
|
| |
Opening a video from dirvish routed through cj/xdg-open to the OS default handler, which plays it once. I pulled the video extensions into their own cj/video-extensions list and route them through a new cj/open-video-looping, which launches mpv with --loop-file=inf (the player and its args are both customizable) detached, so the video plays on repeat. Audio and office docs still use the OS default handler. The find-file advice already covers dirvish RET, so this applies wherever a video is opened.
|
