| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The seam's `query` now calls ZLS's analyser directly, on the worker thread,
in this process. There is no zls binary, no subprocess, no JSON-RPC, no
`initialize` handshake and no `Server` — `gd` is a function call whose answer
comes back as rows. ZLS's build.zig already publishes its guts as an
importable module (`b.modules.put("zls", ...)`), so this is a path dependency
on the local 0.16.x checkout plus one new file, `src/lsp_zls.zig`.
Construction is ZLS's own (tests/analysis_check.zig does exactly this):
InternPool.init -> DiagnosticsCollection -> DocumentStore struct literal ->
Analyser.init. `zig_exe_path` is null on purpose — shelling out to the
compiler is the thing this backend exists to avoid — and `zig_lib_dir` is
baked in from `b.graph.zig_lib_directory`, so `gd` on `std.mem.count` opens
the same mem.zig the compiler used, with ZIG_LIB_DIR overriding at runtime.
Offsets are `.@"utf-8"`, not ZLS's utf-16 default: `+Search` rows are byte
columns and we are not on a wire.
Seventeen probes, seventeen answering, no false claims, 5.9 MiB peak RSS.
The features that were already Server-free are calls (hover, document
symbols, code actions); the ones welded to `*Server` are reimplemented thin
on top of public primitives — goto is gotoHandler minus the protocol,
diagnostics is the in-process `std.zig.AstGen` branch of
getAstCheckDiagnostics, references is symbolReferences' algorithm from the
outside (offer every same-named identifier token back to the analyser and
keep the ones that resolve to the same decl, so a shadowed local is not a
false hit).
What it does not do, deliberately:
- Nothing is cached between queries. Each `query` builds a DocumentStore,
resolves imports and throws it away, because the arena dies on return and
`req.source` is a snapshot of a buffer the user is still typing into. So
cold IS warm — there is no index to warm up. It is also fast enough not to
need one: 124us for a local goto, 2.8ms into the stdlib, 8ms for
references over a 5000-line file. A cross-query cache is a real design
(a global, a mutex, an invalidation story), not a line of code, and it is
the obvious next step rather than something smuggled in here.
- References, rename and select-refs are THIS FILE only. Workspace-wide
means loading every project file into the store and running the analyser
over each; DocumentStore's own workspace iteration has the same limit
(it can only see handles already loaded). Workspace symbols and workspace
diagnostics DO walk the tree, because neither needs the analyser — a
parse and a tree walk each.
- Rename previews, format reports, code actions list. The seam hands back
rows, not edits, so there is no channel through which a backend could
rewrite the buffer. These answer the question the keypress asks and change
nothing.
- Without a zig binary, `@import("builtin")`, `@import("<pkg>")` and
`@cImport` resolve to nothing — silently, which is ZLS's behaviour, not a
bug introduced here. Relative imports and `std` work.
- Non-.zig files answer nothing. The core does not gate the keymap by file
type, so the gate is here: `gd` in a README must find nothing rather than
parse prose as Zig and confidently resolve a word out of it.
A whole-file report that ran and found nothing says so ("no diagnostics",
"already formatted", "no code actions") rather than returning zero rows,
because in this seam zero rows already means "no backend" — `lspResponse`
opens nothing for an empty answer, so silence cannot also mean "checked,
clean". Location queries keep the opposite rule: unresolvable is no rows.
test/snapshots/lsp.snap covers the round trip end to end — gd jumping on a
single result, SPC k opening +Hover, SPC s opening the +Search list that n
steps, and gd on a keyword answering nothing without opening anything. The
whole backend was also fuzzed at 20k queries over real, truncated and
byte-smashed sources across every kind; that found two crashes (a decl's
name token indexes its own file, not the requesting one, and is not
necessarily an identifier at all on a half-typed line) which are fixed.
emscripten does not get the backend: the web shell has no threads and
no-ops the lsp effect, so it keeps the empty one the base tree shipped.
DEPENDENCY: ZLS is FETCHED by the build system (build.zig.zon .url + .hash,
pinned to commit 3e0d0820 on the 0.16.x branch) rather than a path
dependency on the local genizah checkout, so it lands in zig-pkg/ like
every other dependency and the build is reproducible from the .zon alone.
Also passes -Dversion-string: ZLS's build.zig names itself by shelling out
to `git describe`, and a fetched package is an extracted tarball with no
.git, so every build printed a 'Failed to run git describe' warning. We
pin the commit, so we already know the answer.
|
| |
|
|
| |
Plus docs/lsp-evaluation.md: the three backends measured head to head.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The base every language backend plugs into. Three parts:
ASYNC. The core had no request/response shape - every effect was
fire-and-forget or instantaneous. A language query is the first thing
that answers later, so: Effect .lsp -> shell worker -> Event .lsp_resp.
tty.zig uses io.concurrent + the vaxis queue, gui.zig a detached thread
+ the mutex queue it already had for ptys; web no-ops it. The worker
never touches the core (path/source/arg are snapshotted into an LspJob),
one query in flight identified by a monotonic id so a second press makes
the first answer stale, and no rows is a legal answer.
KEYMAP. Helix's, verified against its default.rs rather than recalled.
gd/gD/gy/gi/gr and ]d/[d had no conflicts. The SPC letters did, so
pardes's own builtins moved instead of helix's: Kill k->q, Del d->wc
(closing a pane is a window op, and c is helix's own close), Dump/Restore
s?->f?, Tutor ht->T. A three-exception muscle-memory map is not a map.
RESULTS ARE +SEARCH ROWS. path:LINE:COL text, absolute. That is what
look.zig resolves and n/N step, so one row from a goto jumps and several
open a buffer - helix's multi-result picker needed no picker code.
Backends supply exactly one function (lsp.query) plus a supports set and
a name; the base has none on purpose. zig build lspbench scores them on
the same corpus: feature matrix (trusting results, not the supports
flag - a claimed-but-empty kind is reported as a false claim), cold and
warm latency, peak RSS.
Two snapshot scripts moved. leader.snap encoded the old key paths.
chordcut.snap's last two steps clicked column 5, which lands on a FILE
pane, so 'key c-b' toggled nothing and the typed text was being read as
normal-mode keys - the golden recorded no TTY pane and no cat -v output
anywhere. Pointing them at an actual shell makes both steps assert what
their comments claim, and the tty paste chord is now covered for the
first time.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
selects but never edits
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
asked, never into a column of their own
placeDoc classified any file pane as a doc, so a +Search opened from a lone
shell took layoutInsertColumn(0) - a whole new leftmost column, shoving every
existing column sideways - and an open results list was then eligible as the
split parent for a real file, stacking documents under the list. An output
buffer has no file behind it and no Save; it is the result list belonging to
the pane that asked for it. Now it is that on both sides of the rule: it
splits below from_id whatever kind of pane that is (a source that died falls
back to stackDocLeft, the same last resort a full column bar uses), and it is
skipped when a real doc hunts for its split parent, so a file opened from a
search row joins the docs instead of the results.
Real docs are untouched: fsearch/rsearch/look-file/tutor/lookpanes all pass
unchanged. psearch is the proof - its results snap was two columns (the shell
pushed to x=70, its gutter run at 70-71) and is now one 140-wide column with
the +Search stacked below at the same x origin.
|
| |
|
|
| |
most-recently-focused, search only when all fail
|
| |
|
|
|
|
|
|
| |
Colors lifted from the user's own helix theme (~/.config/helix/themes/pardes.toml,
pinned against a real hx render through a pty). Theme gains box/box_dim so a
restrained theme can turn the accent down; the hardcoded gray.* overlay greys
were the dark theme's own fields spelled twice and are gone. The old default is
still one NextColor away.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The image tag spelled out `Petscii <C64|Term> Ascii` and actOnSelection
matched those words by hand before the builtin dispatch. They are three
real Builtins now (SPC t p/l/a), so the by-word matching is gone, the tag
is a plain `img <path> Del`, and with nothing live left in it the image
early-returns in curTail, enterTagEdit and the tag-click guard go too — an
image tag edits like every other pane's.
Colors and Crt leave the topbar for SPC t c / SPC t r: set-once display
switches should not spend width in a bar you read every frame. NextColor
stays, it is the one you cycle repeatedly. Both remain full builtins.
The snap scripts that clicked those words now press the leader instead, and
images.snap finally exercises the palette and ascii toggles for real: its
old clicks at x=57 landed on the neighbouring shell's tag, so that toggle
had never been covered. The test PPM grew to 512x512 with LCG noise so the
ASCII bitmaps actually win cells and dropping them visibly changes the art.
|
| |
|
|
| |
painting over the UI
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Left/Down/Up/Right run the existing focusDir (now taking the pane the
builtin ran on, so an executed name acts on ITS pane); the Ctrl-w prefix
routes through the same four builtins instead of calling focusDir itself.
Toggleterm hops between the most recently focused document and terminal
off focus_hist, the MRU the core already rebuilds each sync.
Ctrl-w stays: a tty-mode pane never sees SPC, so it is the only keyboard
way out of one. Its snapshot step now moves focus somewhere (c-w l), which
the old c-w h in that layout did not.
|
| |
|
|
| |
Save moves w -> f s to free the w group for window navigation.
|
| | |
|
| | |
|
| | |
|
| |\ |
|
| | | |
|
| |/
|
|
| |
web snapshot tooling
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
the next left click instead of stretching to it (explicit v/x selections keep click-extend), and in a tty-mode shell 1-3 now pastes into the pty like a terminal emulator — the click is forwarded first (SGR or X10, only when the app listens for mouse) so mouse-aware programs paste at the click point, then the register arrives bracketed (mode 2004) or with \n->\r when unbracketed; 1-2 stays inert there. chordcut.snap pins both (a cat -v probe shows the exact wire bytes); tutor updated.
|
| |
|
|
| |
(file:LINE works) opens it focused in the left column with a greeting terminal on the launch dir at right, a DIRECTORY arg chdirs so shells spawn there. New argv.snap; boot/restore.snap pin -n 1 to keep the classic boot.
|
| |
|
|
| |
the new pane under 2 body rows, splitParent swaps in a qualifying pane (same column first), else the tallest pane anywhere; applied to Alt-n, look-dir, placeDoc and ttyForDir. New splitparent.snap.
|
| |
|
|
| |
BOX_H (the tag row doubles as the re-enlarge handle; the gutter move box wins over the handle so a tag-only pane stays movable), a hidden body's pty/vt is never resized so re-enlarging restores it intact, and cursor/modal-selection paints never land on the tag row. New shrinktag.snap; tutor updated.
|
| |
|
|
| |
middle-drag pane has no selection of its own (kept left sel, else explicit v/x n/N), the active pane is checked first (it holds the most recently made selection), then slot order. New selarg.snap covers borrowing a kept left selection and a v selection from another pane; tutor updated.
|
| |
|
|
| |
looked word (and its cwd-join) is matched against loaded file/image pane paths first, honoring :NN recentering — so the fs-less web build can focus dump-loaded panes on look, and native dedup no longer needs the file on disk. New lookloaded.snap: look at a path after rm'ing it still focuses/recenters the loaded pane.
|
| |
|
|
| |
blocked in every thread and consumed on a dedicated sigwait thread instead of vaxis's signal handler, which locked std.Io mutexes from signal context and panicked the Io.Threaded machinery; pane shells unblock it after fork so bash/vim still see resizes. winch.snap storms WINCH from a pane (crashed pre-fix) and checks a real resize still reflows.
|
| |
|
|
| |
(both in one hold = Snarf/copy; the chorded left release is inert), plus 2-3 / 3-1 / 3-2 cancel an execute or look drag mid-gesture (the 2-1 argument chord already existed). chordcut.snap covers cut/copy/paste and both cancels; tutor documents the chords.
|
| |
|
|
| |
scrolled view even when the pane isn't active (the old gate only protected the active pane; tag edits use the remembered body mode). New ttyfollow.snap: background job streams into a scrolled-up normal-mode pane without moving it, tty mode follows again.
|
| |
|
|
| |
longer force-drops to normal (it only restores the mode a tag edit hijacked, remembered in pane.tag_mode), and the left-release cursor pin keeps the terminal's mode (files still click into normal). New ttyclick.snap covers tty body click, tty tag round trip, insert body click.
|
| | |
|
| |
|
|
| |
(rust/cpp/python/...; e.g. opening agave cpi.rs or any tracy .cpp), Debug AND ReleaseSafe. Root cause: clang -fsanitize=function (in zig's default C UBSan set) traps at the runtime's indirect call of the scanner because grammars declare external_scanner_create() with EMPTY PARENS — a K&R non-prototype whose type hash differs from the void*(*)(void) pointer type. The ud1 trap lands on a bogus inlined line (stack.c:746), which cost the diagnosis a detour through rr (its gdbserver dies replaying past the task exit — core dump + coredumpctl worked; ud1 0x6(%eax) = SanitizerHandler kind 6 = function_type_mismatch; scanner-less c/zig grammars never crashed). Fix per review direction: -fno-sanitize=function on the grammar TUs in build.zig — uninstrumented callees make the runtime's call-site checks skip; the rest of UBSan stays live. Second half: fatal signals (SIGILL/SEGV/BUS/FPE) never run defers and bypassed the panic hook, leaving the terminal raw after a crash — root.debug.handleSegfault override now runs vaxis.recover() before std.debug.defaultHandleSegfault, verified in a raw pty (kill -ILL $PPID: rmcup + mouse resets precede the trace). Verified: rust/cpp/python opens work with real highlighting (snapstyle: keywords/strings/comments colored), agave cpi.js 2.7k-line open fine, suite 30/30, ReleaseSafe build opens rust identically.
|
| |
|
|
| |
mode. Surface.cursor grew a bar flag: the core sets it from the active pane mode (insert -> bar; normal/tty -> block) for both the modal/file cursor and the vt-tracking terminal cursor, and unconditionally for the tag cursor. tty shell: DECSCUSR via vaxis setCursorShape (beam for insert, the terminal default otherwise). gui shell: the insert cursor draws the left-eighth-block glyph in cell colors instead of the reverse-video block (ponytail: it hides the char under it; a dedicated quad if that ever matters). Tag editing: enterTagEdit goes straight to insert, Esc (or anything that exits the tag) restores the body normal mode; Enter looks / Tab executes the word at the cursor from inside the tag (tagChordText kept); the whole normal-mode tag machinery — handleTagKey motions, v/x selection, d/c/y/p, tagEnterInsert/tagDelete/tagYank/tagPaste — deleted as unreachable (~120 lines; multi-word tag execution stays the mouse middle-drag). tag.snap rewritten: INS on click, type/backspace straight in, Esc exit with the text persisting, Enter-look opens the file. Tutor documents it. Cursor SHAPE is not capturable by the harness (position only) — verify by eye: bar in insert/tags, block elsewhere, both shells. 30/30.
|
| |
|
|
| |
scroll): pane.hscroll cuts content columns after the lineno gutter in bodyText (post edit-splice, so typed runs shift with the text), no scrollbar. Drivers: horizontal wheel (new wheel_left/right Mouse.Buttons, SGR 66/67 via vaxis on the tty, SDL wheel.x on the gui, ±4 cols, clamped at 0, right overscroll allowed) and keyboard movement — ensureCursorVisible grew the horizontal mirror of the scrolloff reconcile, so $ pulls the view right with 3 columns of margin and 0 reels it home; the wheel never moves the cursor. Shifted alongside: syntax recolor pass, modal cursor paint (hidden when off-view), vsel highlight, click->content-col mapping, cursorWordSel. Byte columns like the rest of the file-pane code. New hscroll.snap golden: wheel shift with fixed gutter, left clamp, $ margin follow, 0 home. 30/30.
|
| |
|
|
| |
next/prev whitespace-separated token that look.resolve can open (several per line: the greet ls row hops big.txt -> plain.txt -> sub), park the cursor at the token start, open NOTHING. Enter's existing normal-mode handler does the look on the selection (dir -> shell+ls there, file -> pane), so navigation continues with n/N+Return inside whatever opened. Wraps when nothing lies in the motion direction (first n out of tty mode starts at the prompt row and lands on the first token — lookStep pins the cursor first). vsel anchors the token end, cursor at its start (vselText is end-inclusive). File-pane search n/N untouched. ttylook.snap rewritten for the new semantics: token hops with selection styling, insert-mode n still types, Return-opens-dir, empty-dir no-op, cursor-word Enter opens a file. 29/29.
|
| |
|
|
| |
into the live session. dump.outPath resolves $PARDES_DUMP verbatim (the snapshot harness pins /tmp/pardes-dump.zon so dump/load/restore goldens stay deterministic), else $XDG_DATA_HOME|~/.local/share /pardes/pardes-YYYYMMDD-HHMMSS.zon (dir mkdir-ed; wall clock via libc clock_gettime — std.time.timestamp is gone in 0.16). After write_dump the shell reports the written path back (core.setLastDump) and the topbar grows "Restore <path>": middle-clicking Restore reloads the last dump, and "Restore <path>" executed anywhere takes the path as its first argument (onSelect prefix match). Core stays sans-IO: the builtin stashes restore_req; each shell consumes takeRestore() after drainEffects and swaps in initFromDump at the current screen size (pre-sized: dump panes never greet), tearing down the live ptys (tty: cancel reader + close + gen bump; gui: SIGKILL, the gen-guarded eof closes the old fds). Web ignores it. New restore.snap golden: KEEPME survives the restore, GONESOON echoed after Dump does not, the topbar word appears then resets. 29/29.
|
| |
|
|
| |
gives the classic two-column boot), -n <count> CLI flag. The 26 legacy snap scripts pin -n 3 on their start line so their coordinates and goldens stay valid; boot.snap tests the new default (golden regenerated: one full-width pane). 28/28 green.
|
| |\ \ \ \ \ \ \ \ \ \ \
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | |
| | | | | | | | | | | | |
three merge-state fixes born of branch interactions.
Branch work merged: tty exit cleanup (krluslzz), drag-select effect-queue crash (zopqvlll), single-pane layout rule (tutmotnl), tty input coalescing + -Dtracy (olnvomwz), / file search (syqzvmkl) + generalized n/N look motion (nmzypsto), ghostty middle-click diagnosis (sqonpmsz), NOR/INS/TTY mode indicators (opmrortk), cursor-row gutter highlight (nymrwpwq), look path:NN centering (oktvtuss), scrolloff (wlqqyvvo), posix portability (pzwluyuw).
Merge-state fixes: (1) splitBelow caps keep at body-(BOX_H+3) so a content-full source still leaves the new pane a tag + a few body rows — Alt-n from a full shell was born 0 rows tall (caught by ttylook: OVER marker had no rows to render in). (2) spawn-effect pane-id reuse crash (user-reported panic at drainEffects assert): no close effect exists, so a deleted pane pty lingers in its slot until a respawn lands on it — both shells now reap the stale pty at spawn (tty: cancel reader + close; gui: SIGKILL, reader delivers the old fd through the eof event) and per-slot generations drop the dead shell late output/eof, which otherwise nuked the NEW pty on the reused id. New respawn.snap golden; fails on the pre-fix assert. (3) dedup re-look now lands the modal cursor on the target line (cur_row/cur_col/cur_pinned) — scrolloff cursor-anchored reconcile otherwise yanked the recentered view back to the stale cursor (caught auditing look-center: center-dedup stayed at 97-102 instead of 146-154); this is also the requested focus-moves-cursor behavior. look-center dedup/clamp clicks re-aimed for the merged layout geometry (typed row moved from 18 to 15).
Goldens reconciled by regenerating from the merged binary and auditing every diff against its origin branch; deltas are exactly: indicator text/width shift, layout-split pane sizing, cursor-row gutter styling, centering shifts, and the pane-edge hover dash now coinciding with new pane boundaries. Suite: 28/28 (27 + respawn).
|
| | | | | | | | | | | |/
| | | | | | | | | | |
| | | | | | | | | | |
| | | | | | | | | | | |
raw std.os.linux.* call sites converted with the errno-convention flip (libc returns -1 + thread-local errno, not -errno): open/close/read/write/rename in tty.zig, gui.zig, look.zig; harness too (snapshot.zig: getcwd/mkdir/chdir/fork+waitpid/getdents64->opendir-readdir, e2e_harness.zig: clock_gettime/write/close), so the parity suite itself is POSIX-portable. Per-OS branches kept only where platforms genuinely diverge: shellCwd in look.zig (linux /proc/<pid>/cwd readlink; darwin libproc proc_pidinfo PROC_PIDVNODEPATHINFO with a hand-declared extern; other POSIX degrades to null), bash path (/usr/bin/bash on linux, /bin/bash elsewhere), TIOCSWINSZ constant (absent from std.c.T on darwin). forkpty/execv/chdir externs were already libc and are portable (fork-safety preserved: absolute-path execv, no malloc between fork and exec, initial spawns before worker threads). build.zig: forward target/optimize to the ghostty dep for native/cross (web keeps its no-target fetch), disable ghostty simd for darwin cross-builds (apple_sdk needs xcrun; scalar fallback = web config), link libutil on the BSDs. Pinned ghostty needs two one-line zig-0.16 darwin fixes, applied in the zig-pkg cache and documented in build.zig. Verified: aarch64-macos + x86_64-macos + x86_64-freebsd cross-compiles green; linux native tty + gui + web green; snapshot suite 18/18 with goldens untouched. Out of scope: gui-on-macos needs Metal shaders (SDL GPU rejects our SPIR-V there).
|
| | | | | | | | | | |/
| | | | | | | | | |
| | | | | | | | | |
| | | | | | | | | | |
(vim scrolloff). One reconcile in Pane.ensureCursorVisible — every keyboard path (normal-mode motions, insert entry/edits, delete/yank/paste/undo/redo) already funnels through it; mouse click/wheel untouched. Clamps: line 0 at top, EOF-at-bottom-row for files (terminals let the vt clamp at the live bottom), margin shrinks on panes shorter than 2*off+1. New scrolloff.snap golden proves j-walk scroll at the margin, gg/top-row reachability, k-walk margin, G bottom clamp. 19/19 snapshots green; no existing goldens changed.
|
| | | | | | | | | |/
| | | | | | | | |
| | | | | | | | |
| | | | | | | | | |
body/2, clamped at 0): fresh opens lay out first (computeGeom after placeDoc) to center with the real body height; dedup re-looks recenter the existing pane. look-file golden shifts accordingly; new look-center snap proves fresh-open centering, dedup recentering, and the top clamp.
|
| | | | | | | | |/
| | | | | | | |
| | | | | | | |
| | | | | | | | |
tag style (tag_fg/tag_bg), resolved at render time from the theme like the rest of the chrome. Same row math as the cursor pass (pinned cur_row or emulator cursor + scroll), gated on active && !tag_edit so only the focused file pane shows it. New cursorline snap script + golden proves it follows j-movement; look-file/syntax style goldens regenerated (cursor-row gutter cells only). 19/19 snap green.
|
| | | | | | | |/
| | | | | | |
| | | | | | |
| | | | | | | |
+ dump-load tty detection (startsWith "TTY "), tutor.txt docs, exec/fileedit snap tag-click cols +1 for the wider prefix, all 18 goldens regenerated (indicator text + 1-col tag shift only)
|
| | | | | | |/
| | | | | |
| | | | | |
| | | | | | |
swallows middle press+release before mouse reporting when the desktop sets gtk-enable-primary-paste=false (apprt/gtk/class/surface.zig gcMouseDown/gcMouseUp; this machine has it false in gsettings). New snapshot ghostty-mid feeds pardes ghostty's byte-exact SGR middle-click stream (focus-in, per-cell hover motions, press M/release m) and proves the tty path executes it correctly. Comment at the setMouseMode call documents the trap.
|
