summaryrefslogtreecommitdiff
path: root/src/ninep/tree.zig
Commit message (Collapse)AuthorAge
* gui: the window a replica client (pardes-gui --attach=<socket>:<name> and ↵Gabriel Schneider2026-10-30
| | | | | | <host>:<name>, as the tty's, far through ssh as amort's: its blobs by hash and its stream in zstd) and a window session serving replicas (a terminal's client or a window's): the client feeds its replica what its window does, read over its own view (views.within) and applied after, draws its view between steps, takes the stream on its own reader and rejoins as the tty's does (replica.Link, the client's half shared with it, the far path's Have with it); the session takes clients at its step boundary, one person's input a wait, a client's view no window of its own, its font and opacity answers taped; a client says its frontend first (Event.frontend, in wire 16 beside the far word) and the tty and window builds step alike (one desktop semantics: each has the other's words, which a view takes by its own frontend, builtins.registry.here); a check asked 300 ms after the last input and check, not per wake; test/daemon.py --window-clients Co-Authored-By: Claude Opus 5.5 <[email protected]>
* tape: every input the core consumes recorded in turn order, replayed into ↵Gabriel Schneider2026-10-29
| | | | | | | | | | | | | | | | | | replicas that must reach the same digests; the core's leaks of render, clock and host closed The recorder (src/tape.zig) is null unless PARDES_RECORD=<prefix> names a tape (<prefix><pid>.tape); the local path pays a null check a call. It records a tree at the host boundary, both ways: each call the host makes into the core (`Input`: update, serveFs, pump, perform, the shell's acknowledgements and failures, a Restore, ...) with its arguments, and each call the core makes out -- a Host function (the vtable wrapped generically while recording) or a syscall of its own (`tape.site` for every old `turn.yield()`, `tape.value` for a clock or environment read) -- with its result; `done`/`ret` close each, a 9P request served while a step is out lands inside that step's syscall, and fields a host writes between calls (native_images, present_skipped, ...) are found by a diff at every crossing. After each call the host made at the top, the session's digest (src/walk.zig), one per field of the core, its stream and each pane, file and places. `PARDES_REPLAY=<tape> pardes` (main.zig) replays it with the same build into three replicas: two at different addresses (the first's memory held while the second runs) and one that never draws. A host call answers what it answered, a site returns what it returned, nothing is performed and no syscall is made; `turn.yield` panics in a replica, so a syscall no site covers is a stack trace; a call the recording did not make, or a digest that differs, is a divergence that names the fields. The headless replica steps over the presents it does not make, applying what the host told the core inside them. What the replicas found, fixed: - Render wrote core state. The view's geometry (wrap map, context rows, body layer rows) is File.geometry, a lazy cache keyed by a hash of what it is made of (revision, scroll, size, tags, tables read, context declarations, the session's view: wrap, layered context, row metrics, rect); the render makes it afresh, a step reading it makes it on demand (wrapAt, wrapRow, ensureCursorVisible). Context declarations, notice bands and the header's reveal are the step's (sync), the tables' reading the tick's (a first look of 4096 lines, a count, wherever it is asked; the rest a frame's clock budget a tick), and a view's paint-ahead is asked of the host as the frame is drawn rather than queued as an effect. - The clock: advance's wakes are the recording session's (a picture decoding or a fade adds no tick to a replica), as are the budgets of Places, Names and Tables (`tape.over`), a message's time of day (the stamp; msg_said_ms and the unattended expiry use the session clock), File.mtime, and the pump reads its clock whether or not it waits. - The host: the session directory (getcwd, asked once at init), $SHELL and its existence, $HOME for `~`, `Shell` naming one that is not there, a missing interpreter, a directory refused above, a deleted file, a pointer seen through a sliding pane (Pardes.pointerCell). - Worker handles (futures, the Io), the 9P server's state, the store file every window shares (a replica keeps its own in memory) and Recent read from it are authority-only; frames, animations and the render's caches local. Query rows' paints count apart from a view's. Terminals replay from their taped pty bytes and are digested as their active screen, cursor and scroll; kitty images are left out until phase 5's screen-state terminals. tagnav.golden: the header's caret reveal now runs every step, as a pane's does, not once per frame drawn, so after `g l B B` typed in one burst the bar stays scrolled as a human typing it sees it. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Store 2/3: one bounded store every window shares (Store.zig): Recent as ↵Gabriel Schneider2026-10-24
| | | | | | | | | | | | | records read on use, the panes a session closes kept as one-pane dumps; Panes lists them (closed), Enter or a look at @pN brings one back beside the pane that took its rows An append-only log under $XDG_STATE_HOME/pardes/store, framed records with a CRC, flock on store.lock (never blocking a step), a stat on use, garbage passed over and compacted away by the next writer, compaction by rename, 100 MiB cap. dump.dumpPane/restorePane lifted out of dumpState/initDump; layout.giveBack; Options.recent_path is store_path. Host calls yield one at a time; /recent refreshes only when no step is out. Tests: two windows and a busy lock, the cap, garbage/torn/partial records, a closed pane's reopen with the layout invariant; fs.py: two live sessions over 9P.
* Merge: Def/Refs, the diff viewer, Helix n/N, Callstack, Zig 0.17, Window ↵Gabriel Schneider2026-10-22
|\ | | | | | | | | | | | | | | | | | | | | mode and the terminal cwd fix onto 0.24 The docs keep both sides: 0.24's Ctrl-i fallbacks and typed ls with this line's Window mode, raw Esc and Helix n/N wording. The CHANGELOG's unreleased entries go above 0.24, Def/Refs and the context rows among them. cheatsheet-a4.pdf rendered again; .scratch/ ignored. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| * zig fmt for Zig 0.17Gabriel Schneider2026-10-22
| |
| * Build with Zig 0.17Gabriel Schneider2026-10-22
| | | | | | | | | | | | | | | | | | | | | | 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]>
| * Def and Refs in a language-served file's tag, a diff viewer's tinted word ↵Gabriel Schneider2026-10-22
| | | | | | | | | | | | | | | | | | | | edits, context rows that hold still under a scroll, and the pardes skill Def and Refs (gd, gr, Ctrl-click; chorded 2-1 at the selection) for a file a language server speaks for; hunks.zig stacks a unified diff as a diff viewer does, its changed words on a tinted ground as delta's do; the tree-sitter context rows stay put while the body glides under a scroll; .agents/skills gains the pardes skill beside pardes-9p.
* | An Edit filter that can never be reaped no longer holds the Edit's answer: ↵Gabriel Schneider2026-10-02
|/ | | | | | at the limit the command's group is killed and reaped off the answer's path, the stdin writer owning its input, so a filter writing its own ctl with > or >> answers EIO at 10 s instead of hanging; a running Edit keeps its own copies of the names it reports, and the panes it changes refuse Del, delete, rmdir, Undo, Redo, Get and Zerox (busy, EBUSY); e and r say unreadable files in Get's words; the reference says both sides of a filter's own ctl write need <> Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Round 39's small ones: a name write past its one line is refused by that ↵Gabriel Schneider2026-10-01
| | | | | | write; Get and Edit's e load over a scratch under 100 bytes unasked, as Del closes it; Putall saves a Zerox pair once; bare Tab on the root ctl says the width as from a pane, and a ctl check of Tab no longer sets it; a directory read is "is a directory" (EISDIR), an unreadable file "permission denied" (EACCES), Incl of a file "not a directory" (ENOTDIR), and a device "not a regular file", never read to the stream limit Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A name held for its newline that no newline could make good (a terminal's, a ↵Gabriel Schneider2026-10-01
| | | | | | component over 255 bytes, a path too long) is refused by the write that held it, not at the close (monkey9p's close_runs_only_line_files; follows sswkwwvv) Co-Authored-By: Claude Opus 5.5 <[email protected]>
* `> exec` through a mount no longer hangs the editor, and a close holding a ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | | | last line is answered at once: the line runs at the editor's next step A shell's truncating open of exec (or look, ctl, log, pager) sends a setattr that changes nothing; it waited for the editor to be quiet, and through a 9ns mount that quiet could never come: 9ns answers nothing else on the mount while the setattr is out, and the editor's step may be out reading a pane's file through it. This was the "truncating open of exec hangs, only in selfmount" mystery. Only a truncate of a pane's file or a tag waits now. A close holding a last line with no newline hands it to the editor's step (tree.runClosedLines, beside fillClosedPagers), as a /pager close does, and is answered at once; a refusal is the log's err. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A +Pager keeps its program's colours: the session parses what /pager is ↵Gabriel Schneider2026-10-01
| | | | written after its directory with ghostty-vt, SGR becomes spans drawn over the plain text, every other escape is dropped, and PagerColor off pages it plain
* A command or terminal pane refused for room is refused before anything is ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | logged or answered of it, and a serial is never handed out twice A command pane with no room logged `new N` and `run N ls`, was answered to its exec, then went with `del N` and no `exit N`; and sync gave N back (`next_serial -= 1`) to the next pane made, so a script holding N found another pane under it. Tty did the same. Both now take a refused placement back at once, before the run is logged or the serial answered (the pane was never announced, so it leaves no record), and fail the write with no room. The decrement is gone: gaps are fine, and a serial once said names one pane for good. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A command line runs when its newline comes, or at its open's close, never ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | because a write was short: a long line cut by a mount is one line A write shorter than its message, and not a multiple of 4096, was taken as whole, its tail a line. 9ns cuts a FUSE write of up to 512 KiB into 65512-byte Twrites, the last one short, so a 600 KiB `Edit ,c/.../` arrived cut and its tail ran as an unknown control message; the 1 MiB refusal failed the same way. Each open now holds a partial line until its newline, or its release, which ends the last line; an Edit block is whole when it closes, as before. A name write with a NUL still fails its write at once. selfmount writes a 600 KiB Edit line and a 1.1 MiB line through a real 9ns mount; building.typ says how lines are cut. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A ctl, exec or look line over 1 MiB is refused once, EINVAL in words a mount ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | maps, and the rest of it is dropped through its newline, not run as a second line Past the 1 MiB a line may hold, the write that crossed it was refused and the pending text dropped, but what followed of the same line was taken as a new line and ran when its newline came: a second refusal, `unknown control message "zzz…"`, and a second err. The open now drops the rest through the newline, and the refusal says `invalid write: a line or Edit block over 1 MiB`. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A look of `./x` that is not there is a miss, as `x` and `/abs/x` are: the ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | write answered, the err logged, not ENOENT `./x` and `../x` failed their write with ENOENT while every other name that opens nothing was a miss the write was still answered with. They are misses now too, said and logged `look: ./x: no such file`. A pane's own directory gone stays a failure. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A command pane an exec open was answered is leased to it while that open ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | stays open: another client's command in the same directory gets a pane of its own A finished command pane was the next command's for its directory, from any client, so with four clients scripting one session, one read another's output as its own last run. An open of a look or exec file whose write was answered a pane's serial now holds that pane: reuse passes over it and a new command pane runs the line. The open's own next command may run there again, and once it is closed the directory's one command pane is reused as before. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* pardes - asks /pager for its +Pager and reads the serial on the same open: ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | another client's exec meanwhile can never take the text into its pane pardes - wrote `pager <dir>` to /ctl, then read /exec on a fresh open, which answers the session's last answer: a client that exec'd in between had its pane overwritten by the paged text. /pager takes the directory and answers the +Pager's serial on the open that wrote it, as /pane/new answers its own open; the ctl verb is gone. Client.ask writes and reads on one open. The README's look recipe reads its answer on its own open too. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A write to sel leaves the text it wrote selected, as acme's wrsel does, so ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | cat sel reads it back; one open's writes run on from the last The selection collapsed to a point after the write, so sel read back nothing. What is written is now dot, and the next write of the same open goes at its end and grows it (acme's wrselrange), so a shell's several writes make one selection; another open's write replaces it. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An event reader whose pane closes reads EOF, not ENOENTGabriel Schneider2026-10-01
| | | | | | | | | | A script following a pane's event in a loop ended in an error when the pane went, though nothing had gone wrong: the pane was simply done. A read on an event open whose pane has closed now answers no bytes, the end a pipe's reader sees when its writer goes, so cat and while-read end cleanly. pty/data keeps acme's answer. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* log stats 0, and pardes re-pins cloud9 whose 9ns opens such a file direct ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | and nonseekable: fs.md's follow recipe with cat works through a mount The log stated the length an open would freeze, a length a mount's kernel could take as its end, and a splicing cat read it from the kernel's offset. log now stats 0, as the other streams and generated views do, and the re-pinned cloud9's 9ns opens any file stating 0 FOPEN_DIRECT_IO | FOPEN_NONSEEKABLE. A look at such a file (the mounted index) reads a regular file that will not seek as a stream rather than refusing it as a pipe. selfmount runs the recipe, exec 3<>log; echo follow >&3; cat <&3, through the mount and hears a record made after it started. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* name takes any byte a file name can hold, controls, edge blanks and bytes ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | not UTF-8 included, so what name reads writes back as it was A pane opened on a file whose name held a tab, a control byte or bytes not UTF-8 read that name from its name file, and the same bytes written back were refused. Only what no file name holds is refused now: a second line and a NUL. The ctl word name still refuses a second blank after the word, as a line's words are split at blanks. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Views generated by each read stat 0, as acme's do: data, xdata, index, ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | layout, recent and listeners; fs.md says which files have real lengths A stat of index, layout, recent or listeners built the whole view to count it, and data and xdata reported the body's length though a read answers what addr selects. Those now stat 0, as screen and pane/new already did, and fs.md lists the files whose length is real and says to read the rest to the end. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A refused write to dot after a shell's truncation changes nothing: the ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | truncation waits for the write, as limit's does echo /nomatch/ > dot truncated dot to an empty range at the top before the write was refused, so a failed address still moved the selection and scrolled. The truncation is now held until the write that follows it: that write sets dot, or, refused, leaves it as it was. A truncation with no write after it (: > dot) empties it when its open goes. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A rename (mv) of a pane's directory or any file of the tree is refused as ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | not permitted, EPERM, not an EIO cloud9's engine passes a wstat that changes a name to the backend, and the tree's setattr ignored the name: the rename answered as done, the name did not change, and a mount then gave `mv` EIO. The tree names its own files, so a setattr with a name is now refused, "rename not permitted", which 9ns reads as EPERM. A pane's buffer is still renamed through its name file. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A pane refused for want of room spends no pane serial, as a refused Newcol ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | spends no column serial A pane/new, look or exec refused ENOSPC made its pane before placing it. The refusal closed the pane unannounced, but its serial was gone, so the next pane skipped one. When sync removes a refused pane that holds the newest serial, it now puts the counter back. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A write of a multiple of 4096 bytes holds the partial line it ends with, as ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | | | | | | | | | | | | | one that fills its Twrite does: a burst through a mount no longer runs its cut lines as two commands selfmount's 1000-line burst reported 943 lines, on main too. The log ring was not losing records: a follower opened before the burst heard exactly 943 and no `lost`, and the burst itself exited 1 with `err 1 exec: wrong #args in control message "Msg"`. `seq ... > exec` through the mount arrives as stdio's 4096-byte writes, each shorter than a Twrite. The rule that such a write is whole ran each one's cut last line at once, so both halves of the line ran as commands (hence `exit 127` records). At 12288 bytes the cut left a bare `Msg`, which was refused and failed the write and the rest of the burst. A write of a multiple of 4096 bytes is where a writer's buffer (stdio, a page cache) filled, so it may go on: its unended tail now waits for the next write or the close, like one that fills its Twrite. `printf Save > exec` stays whole at once. fs.md says so, the 9P fuzzer's model of served lines follows it, and a unit test cuts a `Msg` at 4096 bytes. A follower does not lose records silently: a ring that outruns one already reads it `lost N` first (documented, tested). selfmount now follows the log from before the burst, so it checks every line heard once, in order, or a loss said, not a fresh open of a 64 KiB ring. It passed 3 in 3 and joins the gates with the new 9ns. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A column's and the workspace's tag read with no trailing newline, as a ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | pane's does A pane tag read back its text alone, but a column tag and /tag appended a newline and gave a stat size one byte longer. So `$(cat tag)` agreed across the three kinds while `cat tag | wc -c` and a byte compare did not. All three now read the text alone, and fs.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Nothing about a pane is logged before its new, and the boot terminal's new ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | carries the session directory, not / and a rename A pane named in the update that made it (Config's init file, a look renamed on open) logged `rename N <name>` before `new N`, so a follower met a serial it had never seen. The boot shell, started with no directory, was announced as `new 1 /` and renamed once its shell spoke. Now noteLog drops the rename of a pane not yet announced, since its new carries the name, and announces a pane first for any other record about it. A terminal whose shell has not said where it is is named by the session's directory, where it was started, and the shell saying so is no rename. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A control character in a write to any ctl file refuses the whole write, and ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | a refused line is quoted with its control bytes shown as blanks fs.md already promised that the whole of a write to /ctl, a pane's ctl and a column's ctl is checked first, as it is for exec and look. But a ctl write ran its lines one by one, so a line holding a control byte gave that line's own, misleading reason ("unknown command"). Worse, the reason quoted the raw byte back into the err record. The check now happens in tree.write before anything runs, with the same EINVAL and reason exec gives. A refusal's quoted line shows control bytes as blanks, so an err record never carries a raw escape. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Pane, column and workspace tags take one set of write checks, and a refused ↵Gabriel Schneider2026-10-01
| | | | | | | | `>` write leaves the tag as it was A column's or the workspace's tag took any length, so a write past 4096 bytes went in and a later Dump failed on it (bad dump tag), and a truncating open wiped any tag before the write after it could be refused. The column and workspace tags now refuse what a pane tag refuses: the 4096-byte limit (ENOSPC, tag: no space: over 4096 bytes) as well as the control characters they already did. A truncation of any of the three is held until the write after it is known to fit, and done with it; a refused write drops it; a close or read with no write after it does it then (so `: > tag` still clears). A test runs a truncating write too long, a control character and a bare truncation at each kind, and Dumps after. docs/fs.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
*---. Merge: loose end + msize 64K + docs slimming + docs-audit fixes + monkey-gui ↵Gabriel Schneider2026-10-01
|\ \ \ | | | | | | | | | | | | + tty ssh acked frames
| * | | Newcol shares a column's rows out again when its narrower tags would leave a ↵Gabriel Schneider2026-10-01
| |/ / | | | | | | | | | | | | | | | | | | | | | pane under its tag and two rows, or is refused A seeded sweep of placements (New, pane/new, Edit =, Tty, look, Delcol, deletes, refusals included) over screens 6 to 40 rows high found no refused placement that leaves a pane short since lyltlqnz and lrpzwyom; the pane a row short came from Newcol, which halves a column's width so long tags wrap onto more rows, raising those panes' minimums with no rows moved. Newcol now shares each such column's rows out again, and where a column has not the rows for every minimum there is no new column (Newcol: no space for a column: the panes' tags would not fit), as a size too small is refused. The sweep, 300 seeds, is the test; a user's drag or a tag grown by editing may still take a pane under, as before. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* / / A 9P error past 128 bytes keeps its reason: the path in it gives up its ↵Gabriel Schneider2026-10-01
|/ / | | | | | | | | | | | | | | middle to … An Rerror carries 128 bytes (Plan 9's ERRMAX) and the engine cut longer text there, and a builtin's failure was first clipped between words to 256 with ...: either way the end, which says why, was what went. tree.handle now fits a long reason by cutting the middle of its longest path, else of the whole text, on a character boundary, and reportFailure fills its buffer the same way. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | Only a write logs an err: a refused open or truncation, create or remove, or ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | a write to pane/new, is its error alone, and every refusal is said in words The one-failure rule is a write's. A refused remove, pane/new's open refused for want of a slot, and an OTRUNC open's refused truncation (data's after a failed addr) also logged an err, and a reply carrying only an errno reached the client as the C library's text (Operation not permitted). tree.handle now gives such a reply Plan 9's words, chosen so 9ns maps each back to its errno (EPERM's to EACCES, as Plan 9's does), and serveFs logs errs for writes only, not a write to pane/new, which is only read. docs/fs.md and the 9P skill say so, and that a look miss quotes what was written. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A name written whole with no newline is checked with its write, as the docs ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | now say, not at the close Round 24's no-newline rule (uwzuqzxp) already covers name, a lines file; the 9P monkey's close_runs_only_line_files repros were from a binary before it. This pins it with a test and corrects docs/fs.md, which still said such a name waited for the close. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | An open's writes in a row to one place go in as one edit: a 10 MB body write ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | is linear Each body or data write copied and hashed the whole buffer, so a write the mount cut in 8 KB pieces was quadratic: 10 MB took 50 s and a 1 MB insert into 10 MB 11 s, holding the editor's turn. An open's appends to body, or inserts going on at data's address, are now held and put in as one splice (one copy, one undo step, one line-starts pass) before any other request, the close, or the editor's step once the writes pause 20 ms. fs.py-driven: 2 MB 2.19 -> 0.16 s, 10 MB 50.17 -> 0.78 s, 1 MB data into 10 MB 11.49 -> 0.26 s. A body-2m case (2 MB in 256 KB writes on one open) joins the perf gate: 84023 -> 15785 us; baselines re-recorded. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A ctl or exec line with no newline runs with its write and fails it, unless ↵Gabriel Schneider2026-10-01
|/ | | | | | | | it needs more A last line with no newline was held to the close, so printf bogus > ctl answered 0 and failed only in the log, Edit ,s/zzz/y/ too. A write shorter than its Twrite is the whole of what was written (acme takes each write whole): its tail runs then and a failure is the write's. It is held only when the parser says it needs more -- an open { block, an a/c/i text -- or when the write filled its Twrite and may go on (the listener tells the core the room, msize less its header). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Every catch unreachable, orelse unreachable and syscall assert outside tests ↵Gabriel Schneider2026-10-01
| | | | | | | | is a real refusal or says why it cannot fire A sweep for round 23's crash: a run's answer (pty/run) was bufPrint'd into 48 bytes with catch unreachable, so a foreground program's long name (macOS gives up to 32 bytes) panicked; it now cuts at the room, keeping its newline, in 96 bytes. The rest were numbers into buffers sized for them, a braille codepoint, pthread calls on the queue's own mutex, and pdf_view's resolved outline entries: each now carries a one-line comment saying why it cannot fire. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A second line written to name is refused EINVAL, in one write or on the same ↵Gabriel Schneider2026-10-01
| | | | | | | | open bash writes printf 'a\nb\n' a line at a time, so each line arrived as a whole name on the held open and renamed the pane twice. An open that has named its pane now refuses a further line, as a two-line write already was. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A listing freezes at its open, so a read in several chunks never splices two ↵Gabriel Schneider2026-10-01
| | | | | | | | moments /index and the other listings were rendered afresh for each read, so a client reading one in chunks while panes changed got rows from two moments spliced together (the fs-test flake). /index, /layout, /recent, /commands, /status, /listeners and a ctl opened only to read now freeze at the open, as /screen does. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Two opens writing one body are separate undo steps, however they interleaveGabriel Schneider2026-10-01
| | | | | | The join mark a write left was the pane's, so another open's write joined the first open's step. Writing opens of body, data and xdata now get an open record, the mark names the open that set it, and another open's write starts its own step (the pane's one history: each turn the opens take is a step). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A name cut across writes on one open is one name, applied onceGabriel Schneider2026-10-01
| | | | | | Each write to name was taken as a whole name, so a name a client wrote in two pieces renamed the pane twice, the first time to half a name. name's writes are now held per open, as exec's and ctl's lines are, until the newline or the close. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The writes of one open of data, body or xdata are one undo stepGabriel Schneider2026-10-01
| | | | | | bash line-buffers its stdout, so printf 'x\ny\nz\n' > data arrives as three writes and took three Undos. A write now leaves the join mark a truncation already left, so the next write on the file joins its step until the open is released; a truncation and any other edit start a step of their own. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Recent lists the files opened lately, closed ones too, and a jump back to a ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | closed file reopens it A file closed by accident, its path forgotten, could not be found again: the jumplist dropped closed panes. Every file opened (a look, a Save of a scratch, a Restore, a rename) goes first in a list of 200, kept once, its dot noted when it closes, kept across sessions in $XDG_STATE_HOME/pardes/ recent. `Recent` (SPC f r) shows it in a reused +Recent, each row `path:line:col open|closed` a look reopens; /recent reads `open|closed <path>`. The jumplist keeps a closed file's entries, +Jumps marks them `(closed)`, and Back to one opens the file at its place. acme has no such thing; its dump and Load are the nearest. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An err record's reason that names its file already is not prefixed twiceGabriel Schneider2026-10-01
| | | | | | | A refused look logged `err N look: look: no space for a pane...`: the reason began with the file's name, and the record put it before it again. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* /index shows names as the log does, one line of UTF-8, and a newline in a ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | name as \n /index printed a name as it was, so a directory with a newline in it split its row in two, and bytes not UTF-8 went through; the log turned the newline into a space, silently another name. Both now show a name one way: a newline `\n`, other controls, DEL and C1 a space, bytes not UTF-8 `\xNN`. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Every EINVAL a write gets says why; DEL in an exec or look line is a control ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | character An empty name, a word written to log other than follow, a data, sel, dot or dirty write to a pane with no text, a truncate other than to zero, and a control character in a look line written through a held open were refused with no words, logged as `Invalid argument`. Each says its reason now; and DEL is refused in exec and look lines as in names. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A builtin's Verbose announcement is logged only when it succeedsGabriel Schneider2026-10-01
| | | | | | | | | | | A failure said while it ran (Exit's refusal, a Del refused) or later by the host (a Save's write, a Dump's) left `msg N Save` in the log as if it had worked. The announcement's record is remembered and taken back out on either, unless a follower read it already. A refused pane/new logs its err alone, and a failed Dump says its reason in words (`file not found`, not `FileNotFound`) so its msg goes and the err is the one record. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A pane from pane/new is the session's directory's scratch, not the last ↵Gabriel Schneider2026-10-01
| | | | | | | | | | active pane's No pane asked for it, so borrowing the directory of whichever pane last had the keyboard named it after an accident; acme's new window has acme's directory. Co-Authored-By: Claude Opus 5.5 <[email protected]>