| Commit message (Collapse) | Author | Age |
| |
|
|
|
|
|
|
|
|
| |
of ~100 KB once file texts are pinned) and what a fork stalls the authority for, which settle open points 3 and 5
`zig build rollback-bench -Dplatform=tty -Doptimize=ReleaseFast` (test/rollback_bench.zig) builds a typical session (a shell with 1 MB of scrollback, three 60 KB files and a 3 MB one, 40 edits) and a big one (14 panes, 18 MB, 4x3 MB and 8 small files, 8 MB of scrollback, 200 edits) in process, then measures the state the digest walk visits, the part of it that is a file's text and its undo snapshots (immutable once made: an edit puts a new buffer in place, a snapshot is a copy, so a copy pins them), and the walk again with those taken out, plus a memcpy of what is left. Typical: 36.3 MB walked, 80 KB to copy, about 0.02 ms; big: 282.5 MB, 96 KB, about 0.045 ms. Then a fork of the big session (431 MB RSS: 3.4 to 8.3 ms) and 64 MB written by the parent under the child (~28 ms of copy-on-write).
test/fork_stall.py does the fork alone, on any system with a Python (the FreeBSD box): N MB written, then fork() timed and 64 MB written under the child, with Linux's THP as set and with 4 KiB pages. 1 GB: 13.3 to 19.6 ms here, 20.4 ms on FreeBSD 15.1; the copy-on-write after it ~1.3 us a page on both.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
-L` can forward, refuses another build's hello in words and hands the connection to its session, which welcomes it and answers pings
The daemon binds `pardes-replica.sock` beside its own (SOCK_STREAM, so ssh can forward it; the SEQPACKET socket that carries fds stays local) once it holds its lock. A replica's first message is a hello (detached/wire.zig, now version 14): the protocol version at offset 0, the client's semantics hash and the session's name. The daemon reads exactly that much and no further, so whatever the client sends next is the session's, keeps up to 16 hellos still on their way (ponytail: one that never comes holds its slot until it hangs up), and refuses in words another protocol, a session nobody has, or another build: every session now says its own hash in its registration (daemon protocol 2), so a mismatch never reaches it. An accepted connection goes to the session over its link (SCM_RIGHTS) and the daemon keeps no copy; the session welcomes it with its pid and name on a thread of its own and answers each ping with a pong, until bye or a hangup. ponytail: that stub is all a session does with a replica until phase 5 proper; it is what times a round trip through the handoff and `ssh -L`.
The semantics hash (wire.semantics) is phase 1's undone item: the protocol, the release and commit, the platform, the build options, every dependency the manifest pins by its content hash (grammars, ghostty, the Unicode tables: build.zig's new `deps` option) and the walker codec's assumptions (pointer width, byte order). Not the OS: a Linux client of a FreeBSD session is the main case. ponytail: an unstamped build has no commit, so two dev trees of one release with the same options are not told apart. `pardes --version` prints it on a second line. The hello lost its grid (a replica's view takes a resize like any other input).
test/daemon.py: a hello of this build welcomed by the session itself (its pid, one fd more in the session and none in the daemon, and back once it hangs up), 50 pings, another build's hello and an unknown session's refused, and `--ping <socket> <session> <pardes> [n]`, which times a hello and n pings on any replica socket (a forwarded one, say).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
| |
blocked for good while keys still work
vaxis's event queue woke one blocked writer, and only on the pop from full. The tty's loop has two writers on it: a pane's pty reader, and the input reader posting the wake for a frame's answered status request (FrameAck). Under `yes` both block on the full queue; the loop then takes a whole batch, its first pop wakes one of them, the rest wake no one, and the pty reader waits for good: the program in the pane blocks writing to its pty, the pane stops moving, and every key still works. It happened from about 400 KB of output, every time at 1 MB, through the snapshot harness's pty as through a terminal.
build.zig patches vaxis's src/queue.zig at build time, count-checked as its Parser.zig and Vaxis.zig are: a pop that makes room wakes every blocked writer. A test (src/vaxis_input.zig) empties a full queue in one batch with two writers waiting and requires both to land. `yes | head -c N` through a bash pane in tty mode now finishes at 1 MB, 10 MB, 100 MB (16 s) and 1 GB (179 s, test/perf.zig --live).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
-c 1G` through a real pty, and a big file's open
The baseline the daemon work is measured against (docs/next-daemon.md). `pardes-perf --live <pardes>` runs the real tty binary in a pty (test/e2e_harness.zig, as the snapshot harness does): FrameStats.keys's gestures (key-down, key-right, edit-char, page-down, wheel) on the 50k-line fixture, a phase each, each frame's latency from the shell taking the input off its queue to the end of its present, read back from the session's PARDES_FRAMES dump; `yes | head -c 1G` typed into a bash pane in tty mode, wall time to the mark after it and the frames drawn; and a 1,000,000-line file opened on the command line, wall time to its first line on screen. `--live-gui <pardes-gui>` runs the same gestures as the GUI's own timedemo (PARDES_FRAMES_DEMO=keys) in a hidden test window. p50 and p99 per phase, pooled over --reps runs; text or --json.
Each run has a private home (bash with a `$ ` prompt), XDG_RUNTIME_DIR and fixtures under $TMPDIR, all removed after. The frames dump now says the monotonic time of its first frame (`# t0_ns`), so a driver lines its own phase times up with the frames'.
The perf gate's baselines (test/perf-baseline-{tty-debug,tty-fast,gui-debug}.json) are recorded again: the harness id is perf.zig's hash, and the medium cells came out within about 10% of the old ones (edit-char and edit-twin faster).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
daemon included
Cross-compiled with -Dtarget=x86_64-freebsd.15.0 (15 ships libutil.so.10;
Zig's default 14.0 links .so.9) and run on the FreeBSD 15.1 box:
- 9p_io serves there: fstatat, not Linux's statx. madvise is Linux's.
- WINCH reaches the sigwait thread through a handler that does nothing:
FreeBSD drops a signal ignored by default as it is sent.
- A dead terminal answers a write with ENXIO, which std does not expect: a
Debug build traced it to stderr, that same terminal, and recursed until
the stack ran out (a session whose ssh dropped hung). No tracing there.
- waitid's FreeBSD numbers and 64-bit id: a command pane's exit is seen.
- No /proc: a shell's directory and program by sysctl, ttyTaken by the
foreground group, the foreground job's name by its program.
- File watching by FreeBSD 15's inotify.
- ghostty's kitty loader: off Linux a stat comes before a t=f open (a FIFO
hung it); FreeBSD's shm_open refuses O_NONBLOCK.
- vaxis's test terminal has no read off Linux: no input thread for it.
- test-install: the test-build programs installed into <prefix>/test, to
run where the build cannot.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
hangs pardes or swells it (ghostty's loader patched at build time)
A program in a pane that named a FIFO (`mkfifo /tmp/f`, then t=f, t=t, or t=s through /dev/shm) hung the whole editor: ghostty's loader opens the path and only then checks that it is a regular file, and an open for reading waits for a writer. A device was opened too before being refused (an open may do something), and any image file up to ghostty's 400 MB was read and held before a screen's 64 MiB storage could refuse it.
build.zig's patchGhosttyKitty patches src/terminal/kitty/graphics_image.zig with count-checked anchors, as vaxis and mvzr are: on Linux a path is probed with O_PATH (which opens nothing), must be a regular file, and that same inode is reopened through /proc/self/fd; shared memory must be a regular file (stat of /dev/shm/<name> first, then shm_open non-blocking and fstat); max_size is TerminalImages.storage_limit. Which paths are accepted stays ghostty's (kitty's) policy, which the user accepts.
Tests: a FIFO by each medium and through a /dev/shm symlink does not block (a watchdog thread opens it for writing after 1.5 s and the test fails if it had to); /proc, /sys, devices and directories are refused; t=t leaves a tty-graphics-protocol file outside the temp directory, and one behind a /tmp symlink; a 100 MiB file is refused.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
linked by ghostty-vt), stb_image without them, its writer and its inflate, SDL without its private stb_image; a symbol-table guard; docs/shared-libs.md
ReleaseFast before, both binaries carried stb_image's PNG, JPEG and inflate beside wuffs's (whose function tables keep its decoders linked whatever pardes calls), stb_image_write and its deflate (unused), and the GUI a second stb_image of SDL's (SDL_stb.c, for SDL_LoadPNG and camera MJPEG). Now:
- image.zig decodes PNG and JPEG with wuffs (`wuffs_images`, wherever ghostty-vt is in the build), sized from the IHDR or the first SOF before decoding, as stb's info call sized them; other formats and the builds without ghostty-vt keep stb_image.
- build.zig compiles zstbi with -DSTBI_NO_PNG -DSTBI_NO_JPEG and patches its stb_image_write include out (one count-checked anchor; a no-op stbi_flip_vertically_on_write stays for zstbi.deinit), and SDL with SDL_DISABLE_STB.
Text: tty -74,488 bytes, gui -154,392.
`zig build shared-libs-check` (in unit-test, ELF) reads the linked symbol table and fails on a family the build chose against: stb's PNG/JPEG/inflate beside wuffs, a second stb_image, stb_image_write's deflate, a second zlib or FreeType. Checked against the old binaries: it names every one.
docs/shared-libs.md: the inventory (implementation, users, copies and bytes before and after, allocator hook) and where one copy was not possible: zlib and wuffs both inflate (FreeType's and MuPDF's zlib API vs ghostty's kitty o=z), MuPDF's libjpeg beside wuffs (filter-dct.c is libjpeg's API), MuPDF's unreached fitz image loaders, base64 in Zig std/simdutf/MuPDF, per-library Unicode code, std.debug's flate.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
`zig build kitty-test`, in unit-test) and an end-to-end harness (`zig build kitty-harness`, tty and gui); four gaps it found fixed, one left
The core suite feeds a pane's output as pty reads do and checks what the program reads back, ghostty's image storage and the ImagePlaces a frame draws: queries and their errors, q=0/1/2, f=24/32/100 and o=z, chunked m=1 split at every read size, t=f/t=t/t=s, a=T and C=1 cursor movement, a=p by placement id with source rects and re-puts, image numbers, every d= delete (a/A i/I n/N c/C p/P q/Q x/X y/Y z/Z r/R) and a delete abandoning a chunked load, Unicode placeholders over two rows with implicit columns and a placement named in the underline colour, scroll/resize/reflow, the alternate screen, clear (ED 2 + ED 3), replies per pane and never in the cells, many replies from one read, kitty before DA, CAN/SUB, an oversized image, 300 images against the storage limit, animations and relative placements.
The harness runs test/kitty_client.zig in a pane's bash (query, a 16-chunk 8x4-cell place, the three media, quiet levels, CAN, show then delete) and kitten icat (stream and detect) and chafa when installed. Outside, the tty arm is a kitty terminal (a ghostty-vt emulator): the image pardes re-emits lands at the cell after ANCHOR>, 8x4 cells a row at a time, the client's own commands and replies never pass through, and d=I frees what was sent. The gui arm reads the GPU capture.
Fixed: kitty file, temporary-file and shared-memory media (libghostty defaults to direct only; now allWithTempDir like ghostty and kitty); replies past 256 bytes of one read were cut mid-sequence (4 KiB now, an answer that does not fit dropped whole and counted); animations never played (TerminalImages draws renderData(), ticks ImageStorage.animationTick on the core clock, nextWake wakes for the next frame); relative placements (P=/Q=/H=/V=) were not drawn (resolveChain, as ghostty's renderer). image.decode no longer jumps through stb_image's null allocator before image.start.
Known gap (a knownGap test): no z-index; every image is drawn over the cells in storage order.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
| |
ghostty, libvaxis and uucode move to ghostty's 0.17 pins, cloud9 to its 0.17
main, SDL to castholm's release-3.4.x (3.4.18), zstbi and tree-sitter to
forks with 0.17 build files (see build.zig.zon). A bare zig build copies the
binaries into ~/.local/bin (-Dhome-install=false leaves it out): 0.17 keeps
the prefix from build scripts. A terminal's edited copy goes when the program
clears the screen. tinywin.snap is gone; perf baselines are re-recorded.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
the virtual filesystem is drawn as a tree
#word draws a builtin as a chip on the tag's ground over a rule (Look and
Exec in acme's look green and exec red), a <code class="word"> in HTML,
and links it to its entry in a new Builtins chapter wherever that
chapter is in the same output (the book, the site). The chapter is
drawn from docs/typ/builtins-tty.json and builtins-gui.json, which
`zig build builtins-json -Dplatform=tty|gui` writes from the registry
and the Help rows (src/builtins_json.zig, tools/gen_builtins_json.zig;
summaryOf in ninep/ctl.zig is now pub); style.typ merges them, marking
the window-only words. The doc steps refresh their build's file first,
and each build's unit-test fails when its file is stale. Everyday words
first (`everyday` in style.typ), then groups by leader prefix as Help
has them, then settings, the words with no path, debugging and
internals, and last an A-Z index of every word. `zig build book` joins
the doc steps.
#fstree draws a directory tree with its connectors, a description
column and links to the sections that document each file; reference.typ
draws the served tree with it, each pane, pty and column file on its
own line, and a test in config.zig checks it against ninep/tree.zig's
files both ways. The cheatsheet has a two-level version. On the site
the tree reads as `tree` text in w3m and wraps descriptions under the
names on a phone.
The site's tutor draws its B1, B2, B3 and 1-2 chords as mouse glyphs;
the landing credits Plan 9 and acme, as does the book's title page;
install.typ no longer lists the mirrors (the footer does).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
recordings, one Typst bundle run
docs/site/site.typ writes every page in one bundle compile: the landing
page (the tagline, the recordings listed in media/clips.toml, links to
start from), a page per entry of `chapters` in style.typ, and the tutor.
Every page shares a tag-line header with the chapter words; #doc links
resolve to the right page and anchor; a chapter with four or more
sections gets an "On this page" list. Links are relative and pages flat,
so it reads from file:// or any static server. `zig build site` installs
it to <prefix>/share/doc/pardes/site and fails plainly without typst.
The look is 0x4200.cafe's lapis layout (framed panels, hard shadows, a
tag-line trail, Departure Mono headings over Crimson Pro) in orchard's
greens and amber, forestbones_light when the reader prefers light; both
fonts are OFL and shipped. In HTML a key is one pill with its + and ›,
and a mouse button is an inline SVG in currentColor. The tagline lives
once in style.typ for the sheet, the book and the site, and the single
page html.typ gives way to the site.
The landing leads with one plain sentence (columns of panes, click to
run or open, helix keys, scripts through files), then the tagline and a
line on where it runs; a hero image is the first ready clip's poster;
"Start: the guide" leads the links, "Hands on: the tutor" next, the rest
a plain row; the captions use no undefined words. docs/typ/install.typ
holds the install lines once: setup.typ has an Install section that
includes it, and the landing includes it too. Building says up front it
is for contributors and points at setup to install. In HTML the guide's
glossary is a closed <details> (#glossary, the same heading and table on
paper), each drawn mouse carries hidden text (left-click, hold left,
click middle) for screen readers, w3m and copying, and the footer takes
a source link and a licence from `project` in style.typ once decided.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |\
| |
| |
| | |
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |/
|
|
|
|
|
|
|
|
|
| |
check reads the target asked for, not the board's pinned one
The refusal compared the resolved target, which build.zig pins to the
board for esp32p4, so it could never fire and -Dtarget=x86_64-linux
built firmware anyway, silently. It now reads -Dtarget as given and
refuses any target but riscv32-freestanding.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
docs/typ/cheatsheet.typ holds the facts as semantic calls (#key, #btn,
#chord, #word, #tag, #addr, #file, #cmd, #doc); docs/typ/style.typ
defines those once, drawing them for paged output and emitting HTML
elements under --features html; docs/typ/cheatsheet-a4.typ is the page.
`zig build cheatsheet` renders it when typst is on PATH. A config.zig
test reads every .typ in docs/typ and fails when one names a builtin,
key binding or default tag pardes no longer has.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
| |
drain, and lspprobe is lspbench's probe subcommand
B4: test/bench_util.zig holds what pdf_bench, pdf_pointer_bench, pdf_scroll_bench and pdf_sections_bench each copied (nowNs x4, fatal x3 named per bench, mix x3, parsePositive/parseCount x3, drainEffects x2). They stay four entry points: pdf_bench links its own mupdf module (with its C source and c_heap) where the others link the core's, which one executable cannot import under one name, and their reports and identities are each bench's own. pdf_sections_bench had rotted since the pane's cursor and selections moved to its body and the render revision became next_raster_revision; it builds and runs again. C6: tools/lspprobe.zig is lspbench's `probe` subcommand (zig build lspbench -- probe <kind> <file> <line>:<col>), and the lspprobe build step goes; docs/lsp.md names the new command.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
| |
other language does: the in-process ZLS backend goes
lsp_zls.zig (1,522 lines), the zls dependency, build.zig's zls wiring and the import graph it baked into the options only for it, and main.zig's reference go. lsp_client's table gains a zls row (PARDES_LSP_ZIG overrides it); Hover, Rename, Diagnostics, WsSymbols, Lspinfo and gd answer through it (checked with lspprobe and the snapshots against zls 0.16.1-dev). What changes for a user: zls must be on PATH; a completion's rows are zls's candidates at the cursor, where the linked analyser gave each one's declaration; document symbols carry no signature. The Zig snapshots (lsp, lspcomplete, lspcompletemoved, lspdebug, lsprelpath) run the real zls now: lspdebug reads the client's explain and report, lsprelpath spells gr's rows where it spelled a completion's, and their goldens are recorded again (stable over three runs). Lspinfo still opens with its backend line. The ReleaseFast tty binary is 105.35 MB to 96.28 MB, the Debug one 296.5 MB to 245.9 MB. docs/lsp.md, web.md, design.typ and the README say so.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
| |
copies, identical to their sources, go, and so does the label saying which was used
shaders/prebuilt/*.glsl were byte-identical to shaders/X.glsl or to prefix.glsl + post/X.glsl (checked with cmp, all 14), kept only so a -Dprebuilt-shaders build could name them; EffectCode and Config then labelled the GUI's shader source as live or prebuilt. The GUI embeds shaderSource (a concatenation, no glslc) either way now; the prebuilt SPIR-V stays, and zig build shaders refreshes only it. Config's report and EffectCode lose their GUI shader source line.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
| |
the post chain's builds) takes a type's name
The naming the split agreed on: a file with fields is a type, TitleCase, as Messages.zig and Layer.zig are. Its importers bind it as ShaderBuild; comments and docs/render-pipeline.md name the new file. No behaviour changes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |\ \
| | |
| | |
| | | |
+ tty ssh acked frames
|
| | | |
| | |
| | |
| | | |
test read equal as changed
|
| | | |
| | |
| | |
| | | |
weighed up per cell in the tty
|
| |/ /
| |
| |
| |
| |
| | |
It was a scratch tool (.scratch/render/tools/gui_monkey.py) that the gates rely on and that imported fs.py from a workspace since removed. It is test/gui_monkey.py now, importing from its own directory, and zig build monkey-gui -Dplatform=gui runs it on the built window, the same hidden test windows; it exits 1 when a combination fails, and its outdir is optional.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
documented rules
test/monkey9p.py starts pardes --detach with its own HOME, XDG dirs and
runtime dir, speaks raw 9P (test/ninep.py) with a timeout on every reply
and Tflush for held reads, and after each seeded operation checks that
pardes is alive, the one-failure rule over a marker-delimited log window
(with and without a follower), /index against /pane, and chunked against
whole reads. Every run writes a replayable ops log; a failure is saved
with its seed and step, shrunk by delta debugging over replays, and the
run goes on in a fresh session. Teardown makes the fuzzer a subreaper and
kills every process carrying the session's token.
zig build monkey-9p -- --seed N --steps M | --replay F | --shrink F;
a 150-step fixed-seed smoke run is part of fs-test.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |\ \ |
|
| | |/ |
|
| |/
|
|
|
|
| |
The in-process ZLS's report listed the module map of the build pardes came from (vaxis, ghostty and the rest) for any .zig file, a project that has none of them included. It now lists them only for a file in that tree (project_root, a build option), and says there are none for any other.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
let the GUI rest; a perf gate
somtrmsz's contrast floors made ChromeTheme.fromTheme run its focus-tint
and separator searches (pow calls each), and Output's RowDecoration.styleAt
called it for every grapheme of every highlighted row: the 50k-line file's
render went from 0.11 ms to 13 ms. Pardes.bodyChrome now keeps the theme's
chrome, worked out again only when the theme differs; recolorSyntax asks
for it once a decorated row, and a plain row (nearly every row of a file)
never asks. ReleaseFast, medium fixture, median us, before -> after (main):
render 13150 -> 76 (111), key-down 14658 -> 79 (110), wheel 13589 -> 76
(109), open 14891 -> 2095 (2055), edit-char 16703 -> 3006 (1251; the rest of
that gap is editing and tree-sitter, not this).
Post.animating asked for frames whenever the window had focus and any pass
was ready, so a still pass kept the GUI drawing at the display's rate
(Bloom: 49% of a core idle, in a hidden test window). A pass now says
whether it moves on its own: the CRT (its hum and dither) and a Shadertoy
file whose source reads iTime, iFrame or iDate (shader_build.readsTime; the
flag rides the wire's post message); Bloom, Vignette and Grain are still.
Bloom idle: 49% -> 1.2%, as with no pass.
zig build perf-gate: the 50k-line file's gestures, each's fastest sample
within 3x of the recorded baseline's (test/perf-baseline-<platform>-
<optimize>.json, recorded from this build), run with every unit-test. On
the regressed code it fails at 180x for render.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
chain
shader_build.zig takes the Shadertoy compile out of the GUI's Post: the
process that holds the core compiles each chain file behind the prefix on a
thread, watches its directory (file_watch's new shader slots), and on a
change reads and hashes each file, compiling only one whose bytes moved. A
failed compile keeps the last good SPIR-V and is said once; the same bytes
are never compiled or reported twice. A file is read with std, not fs.zig's
readFile, whose turn hand-off a worker thread does not hold.
A detached session compiles for its frontends and sends them the chain
(wire post: each pass's scene and level or the file's SPIR-V, and
ShaderAnimation) on attach and on every change, so an attached GUI runs the
same passes, levels and animation as a local one while still reading no disk
and running no program. The attached GUI describes its frames to the chain
from the session's chrome and redraws an animating chain on its own.
|
| |\ |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Three bundled Shadertoy passes beside Crt, each `Bloom|Vignette|Grain
0..3` (off by default, nothing drawn while off), in shaders/post/:
- Bloom: what is brighter than the page (a channel's linear peak over
0.75/0.7/0.6 and over the page's own + 0.15, so a light page does not
bloom) glows, 2/3.5/5% (§8.2's budget). The glow is a dual Kawase blur
in RGBA16F textures at half size and down, 3/4/5 steps by level
(shaders/bloom-down, bloom-up; Post.bloomChain), composited in linear
light and dithered.
- Vignette: round, from halfway to a corner, 10/18/28% at the far corners,
in linear light, dithered.
- Grain: a still, triangular-spread grain of 1.5/3/5 of 255 on
page-coloured pixels only, a logical pixel big; still, so it asks for no
frames.
Focus rule: the shell passes every bundled pass the rectangles of the tags
with their grips, the column and workspace tags, the notices and the
cursor (pardesSpare, at most 32, neighbours on one row merged), and a
bundled pass leaves them as they are; and a per-theme ceiling (pardesCap,
Post.capsOf) keeps the page's text and the selection's text at min(their
contrast, 4.5) under each pass at its strongest. User files get no
ceiling and no spared rectangles.
Fix found on the way: SDL GPU binds at most 4 KiB of a uniform block, so
every Shadertoy colour after iPalette (iBackgroundColor, iForegroundColor,
the cursor and selection colours, iPalette[>=4]) read zero. The prefix's
block is cut in three (Globals, Palette, Colors) with ghostty's names and
order, each pushed as a slice of the one struct; pardes's own block moves
to binding 3.
Tests: the caps hold over every native theme; each uniform slice fits
4 KiB; the prefix's members in ghostty's order across its blocks.
Gates: tty/gui unit-test and core-test pass; snap fails only
nested-optout; GUI goldens 01-08 byte-identical (09-18 wait on
wkzlsqvy, as in somtrmsz); web builds. Feel review material:
.scratch/render/{bloom,vignette,grain}/ (levels on forge, dusk, acme,
crops, a levels mp4).
|
| |/
|
|
|
|
|
|
|
|
| |
mvzr refuses $ anywhere but at the very end, so /h$|zz/ and Edit ,x/r$|pi/
were bad patterns while /zz|h$/ worked. Its matcher already ends an
alternative at .end, so a build-time patch, anchor-checked like the group
one, lets $ stand before | and ). The upstream report is in
docs/mvzr-group-backtrack.md.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |\ \ |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
esp32p4
uzksuumxruzr turned -Dembed-sources on for esp32p4 (before it, the board
was the one build without them). The sources are ~1.8 MB against a 1.5 MiB
app partition, so the object built but the firmware did not link:
.flash.rodata overflowed by 786392 bytes and .flash.text by 1732312.
Default off everywhere; -Dembed-sources=true still embeds them.
Without them the image is 1459760 B, 113104 B under 1.5 MiB.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
config.zig imports pardes.zig and builtins.zig, so it cannot be compiled
"without the editor" any more; the step failed with 40 missing-module
errors (zls, uucode, cloud9, mupdf, ...). Its two tests (Space-k unbound,
wheel drift guard) already run in core-test, checked with -Dtest-filter.
Wiring every core import into it would only rebuild core-test for two tests.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| | |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The git commit sat in the options module every binary imports, so each
HEAD move (jj new, jj commit) recompiled every test and snapshot binary:
measured 24.4 s and +0.6 GB of .zig-cache for test-build -Dplatform=tty
after one no-op HEAD move. Now -Dstamp-commit decides, defaulting to on for
release builds and the bare zig build that installs into ~/.local; every
other Debug build prints `pardes <version>` alone. After: 0.21 s, +0 bytes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
| |
On aab, /a+ab/ matched but /(a+)ab/ and /(a*)ab/ missed, and so did
(.+)_area and ([a-z_]+)_area, in addr, Edit and look alike. mvzr's hasAlt
counts a group's own ) as an alternative, so every group with pattern after
it takes matchGroup's alternatives branch, which never tries a repeat's
shorter matches. No newer mvzr fixes it (trunk is the pinned commit), so
the build patches the fetched source as it does the step budget,
anchor-checked: when the rest fails after a group, the group's shorter
matches are tried, longest first. docs/mvzr-group-backtrack.md is a report
for upstream.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |\ \ |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The user tried them live and wants them removed entirely: the builtins,
their levels, config keys and leader paths (tR, tg), their slots in the
built-in chain, their shaders (source and prebuilt), EffectCode paths, and
on macOS their scene flags (pardes.h), the Metal kernel's coordinate warps
and the Swift pointer mapping that mirrored them. Docs say so (decision 8
in docs/render-pipeline.md); config.md documents Crt's levels, Shader and
ShaderAnimation instead.
Shared files touched: macos.zig (flags). Not touched: pardes.zig,
Messages.zig, mouse.zig, gui.zig, detached/*.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
rewritten as Shadertoy files
Stage 9 of docs/render-pipeline.md. src/gui/Post.zig runs the chain over
the finished frame: each pass reads the one before, ping-pong textures
between, the last writes the window; an empty chain is the direct path.
The prefix (shaders/post/prefix.glsl) is ghostty's uniform block, name
for name and offset for offset (tested against a copy of its Uniforms, the
selection colours by their GLSL names), at SDL GPU's bindings, y down.
A user's file (Shader <path>) is compiled by glslc with compileGlsl's own
flags on a thread of its own, taken in at the next loop step; errors name
the file's own lines; a failed compile keeps the last good pipeline;
glslc missing is said once. The bundled passes compile with the build,
through the same prefix.
Crt, Ripple and Glitch are Shadertoy files, rewritten: no barrel (input is
identity, tested), everything in linear light and dithered, keyed to iTime.
scene_effects, crt.zig and crt.frag.glsl are gone; the core keeps no clock
for the chain. ShaderAnimation off|on|always drives redraw level A: the
chain alone over the retained frame (34 us CPU a redraw, measured).
Shared files touched: pardes.zig (nextWake loses the scene line,
disableSceneEffects empties the chain, two tests), builtins.zig (one switch
arm), ninep/ctl.zig (two switch arms), macos.zig (flags from the chain),
gui.zig. Not touched: Messages.zig, mouse.zig, detached/*, host_io, tty.
|
| | |/
|/|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Stage 8 of docs/render-pipeline.md. The frame is drawn in groups: tier 0,
one per track in paint order (tiers 2 and 3), tier 4 (notices, then guides
and the debug box), tier 5 (bar cursors), each cells, images, decor. Decor
(rules, rails, thumbs, grip marks, spines, the workspace and column rules,
the bottom band, notice rules, bar cursors) is whole-pixel rects through
the new decor.frag over ui.vert, so it carries its track's transition and
clip. A closing pane's chrome comes from Surface.previous_regions.
Deleted: taglineBaseRgb, topbarPaneBorderHeight, bottomTaglinePresent,
frameChromeBg, cellBackgroundIs and the rail inference, paneGripCell's
scan, transient_on, PaintPlan. One cover map (coverFrame) is marked from
the layers and regions once a frame. New regions column, guide and debug;
Surface.chrome is the palette, carried in wire v8 (not shipped yet), so an
attached GUI draws the same chrome (test). A per-instance clip replaces the
vertical transition's scissor.
Goldens: 01-12 byte-identical. 13-17 differ only past the grid: the image
pass left its scissor at the grid's size, cutting rule ends, rail feet and
the bottom band in the leftover pixels whenever a picture was on screen.
16-debug now shows the debug box, which a context-row layer had hidden.
18-mid-transition is new: virtual clock (PARDES_TEST_CLOCK in the GUI),
PanelSlide Newcol with the picture, frame 6 of 12.
Also: the GUI sleeps when idle instead of polling every 16 ms, and a
minimized or occluded window sleeps through animation. Tracy 'gui frame
build' at 200x60: 992/1015 us median before, 989/987 us after.
Shared files touched: pardes.zig (one export), detached/client.zig,
detached/server.zig, detached/wire.zig (all additive), gui.zig. Not touched:
Messages.zig, mouse.zig, tagline.zig, colors.zig, tty.zig, host_io.zig,
dump.zig, exec.zig, panes.zig.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
helix's textobjects.scm for pardes's grammars are vendored in vendor/queries
under MPL-2.0, each with a header naming its source and what it inherits,
inlined; a README says what is there and what is not. syntax.objectAt and
objectNext run them over the file's kept parse, with their #eq? and #match?
predicates, for ]f [f ]t [t ]a [a ]c [c ]T [T ]e [e ]x [x and mi/ma with
f t a c T e x. erlang's and zig's queries do not compile against the grammar
versions pardes builds and are left out; fortran, markdown and powershell
have none in helix. A test replays sixteen Python and JSON cases whose
results were taken from the installed hx, and another holds that every
vendored query compiles.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| | |
| |
| |
| |
| |
| |
| |
| | |
reverse_golf_example from ""N on needs *, N and n to walk the search
register; in pardes those keys are the acme look ring, by decision. The seven
steps are pinned in golf-waivers.jsonl, so a new mismatch stands out.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |/
|
|
|
|
|
|
|
|
|
| |
The ten examples on helix-golf are imported as 118 cases: each numbered step
of an example's walkthrough is the case whose keys are the command up to and
including that step, so the first mismatch names the step where pardes and
helix part. Six more restart at a step that begins with %, from helix's own
text there, so a later step is tested even when an earlier one misses.
zig build hxgolf compares all 124 with goldens from hx-harness.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
not windows and a repeat cap
The windows returned wrong matches: a candidate reaching a window's edge was
left to the next window, half a window on, which could answer a match starting
mid-token rather than the leftmost, and addr then pointed data's next write at
the wrong bytes. The repeat cap missed mvzr's own worst case, a chain of a?,
and alternation under a repeat, each exponential inside one mvzr call the
deadline could not interrupt; and it refused ordinary s/S patterns. build.zig
now patches the fetched mvzr at build time with a step counter on its
backtracking recursion (matchPattern), so a fresh fetch keeps it and a moved
anchor stops the build; regexp.zig gives each compiled pattern a budget, about
300 ms here, and a search that spends it fails as taking too long. Windows,
the cap and their special cases are gone, and matches are exact again.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
runtime
syntax.zig points the tree-sitter runtime at a Zig allocator, but the
grammars' external scanners did not follow it. A grammar's
tree_sitter/alloc.h maps ts_malloc to the runtime's current allocator only
when TREE_SITTER_REUSE_ALLOCATOR is defined, and plain malloc otherwise, so
every scanner's state and its arrays came from libc. Five scanners (bash,
markdown, markdown_inline, python, typst) also call malloc, calloc, realloc
and free by name.
build.zig now compiles every grammar with TREE_SITTER_REUSE_ALLOCATOR and
forces in src/tree_sitter_heap.h, which includes stdlib.h and then defines
the four names as calls through ts_current_malloc and friends. A header
forced in from the build rather than a patch to the grammars keeps zig-pkg
pristine. No grammar archive refers to libc's allocator any more; only the
runtime does, for its defaults. A scanner frees only what it allocated
itself, and the syntax tests that create and destroy a parser for every
grammar under a leak-checking allocator pass.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
MuPDF compiled its bundled zlib 1.3.1 (ZLIB_SRC) into libmupdf.a, and the
freetype package links the zlib package, 1.3.2, for its gzip module: 68
symbols defined twice, with the linker keeping one of each. MuPDF now links
that same zlib artifact and stops compiling its own. build.zig asks
FreeType's build for zlib with the options FreeType asks with, which the
build's dependency cache answers with the very artifact FreeType links, so
the tty and the gui each carry exactly one zlib.
MuPDF hands zlib its own allocation functions on every stream, so nothing
about where the memory comes from changes. Every page of the three PDFs in
docs/ still renders bit for bit as before, at 144 and at 300 dpi.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
MuPDF compiled its own FreeType 2.13.3 (FREETYPE_SRC, trimmed by
scripts/freetype's slimftmodules.h and slimftoptions.h) into libmupdf.a, and
the gui also links the freetype package, 2.14.3 in its full configuration.
The two define the same 312 symbols, and a linker handed both keeps one of
each: the gui's MuPDF ran the package's FreeType compiled against the slim
headers, while the tty ran the slim one. Now there is one. mupdf.zig takes
the build's freetype artifact, links it (its headers come with it) and no
longer compiles FREETYPE_SRC or puts the slim headers on the include path.
The shared artifact is built at c_optimize, as MuPDF is, so a Debug gui's
glyph atlas now uses a ReleaseFast FreeType too; one artifact cannot be both.
The tty gains the full FreeType's extra modules, which FT_Add_Default_Modules
registers whether or not a PDF needs them. MuPDF asks nothing of FreeType the
full configuration lacks: every page of docs/design.pdf, docs/9p.pdf and
docs/registry.pdf renders bit for bit as before, at 144 and at 300 dpi.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
C's malloc, calloc, realloc and free over a Zig allocator were written three
times: syntax.zig's for tree-sitter, pdf.zig's for MuPDF, and gui.zig's for
FreeType and HarfBuzz. They are now one, src/c_heap.zig's Heap. Each hook
names the allocator variable its heap reads; gui.zig exports its heap under
the ui_malloc names font.c and HarfBuzz call. It is a module of its own
because pdf.zig is the MuPDF module, and a file cannot belong to that module
and the core's at once.
A block starts with a 16-byte header holding its size and the allocator that
made it, so it goes back where it came from even after the heap is repointed,
as the fonts' copy did. The fonts' copy carried the 16-byte Allocator itself
in a 32-byte header; here the header holds the allocator's index in a table of
every allocator a heap has used (entries are published once and never change,
and a heap remembers where its current one was last found). The difference is
measurable: with the 32-byte header, pardes-pdf-bench's filtered page render
at 96 dpi took 1.1% longer than before this change, slower in 15 of 16 pinned
rounds in two separate runs, and highlighting all of src/pardes.zig (305k
tree-sitter blocks) up to 0.6% longer, against an A/A pair within 0.7%. With
the 16-byte header both are back within the A/A pair's noise.
Two other things change. For tree-sitter and MuPDF, a block made before the
allocator was repointed now goes back to the one that made it rather than
through the current one. For the fonts, realloc(block, 0) now frees and
returns null, as glibc's does and as the other two copies already did;
neither FreeType nor HarfBuzz asks for it. The heap's test replaces the
tests of the old copies.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
FreeType and HarfBuzz allocated with libc malloc, out of sight of the Zig
allocator the rest of the gui uses, so its Debug leak report and a test's
std.testing.allocator never saw them. Now every allocation the font code
makes goes through ui_malloc, ui_calloc, ui_realloc and ui_free, exported
from gui.zig over `font_allocator`:
- HarfBuzz is compiled with hb_malloc_impl and friends pointing at them.
- Each UIFont gives its FreeType library its own memory manager
(FT_New_Library with an FT_MemoryRec_ over the same functions, then
FT_Add_Default_Modules and FT_Set_Default_Properties: exactly what
FT_Init_FreeType does over its malloc manager), torn down with
FT_Done_Library.
- The shim's own UIFont and scratch arrays use them too.
C's free and realloc pass no size, so a block starts with a 32-byte header
(the caller's pointer stays 16-byte, max_align_t, aligned) holding its size
and the allocator that made it, so a block goes back where it came from
even after font_allocator is repointed. runNative points font_allocator at
the gui's gpa. HarfBuzz makes a few process-wide objects on first use and
never frees them (the default Unicode functions, the locale's language,
hb-ft's font functions); ui_font_prime makes them from the default heap
before that, so a Debug build's leak report stays about fonts. Tests do the
same: the shaping tests, and a new test that creates, rasterizes and shapes
fonts under std.testing.allocator, report any leak. Only the render thread
calls into fonts, and every allocator used is thread-safe.
It costs nothing measurable: against the previous change, 10 interleaved
rounds with an A/A pair show every scenario as fast or faster with Adwaita
Mono and Maple Mono alike, and first paint over 30 launches each matches
main.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
|