summaryrefslogtreecommitdiff
path: root/src/9p_io.zig
Commit message (Collapse)AuthorAge
* 9p_io: the parked-write test rests again until the client's release has ↵Gabriel Schneider2026-10-30
| | | | | | landed before it reads the body: a write through an open is held (pane.zig batchFor) until the open is let go, and the client lets it go without waiting, so under load the test read the body truncated and the bytes not yet in (it failed on the first run of a loop beside builds, then 30 of 30) Co-Authored-By: Claude Opus 5.5 <[email protected]>
* 9p: a mount keeps its connection and sends each operation's walk, stat, open ↵Gabriel Schneider2026-10-30
| | | | | | and first read or write in one flight, a new connection's version and attach before them: an operation through it is one round trip where it was six or seven (a directory, or a file of no stated length, two: 9P's only end is an empty read); a file whose stat gave its length is read ahead, its Treads 14 at a time; reads are timed by their progress, so a large remote file opens at 80 and 207 ms (1.3 MB timed out); a write refused unopened behind a truncating open that parked is sent again after it; Unmount lets the connection go; the CLI, --wait and Handover dial afresh, pipelined; reference.typ says how a session elsewhere mounts through ssh -L Co-Authored-By: Claude Opus 5.5 <[email protected]>
* upgrade: the Upgrade word (SPC s U, a tag, 9P) moves a running session onto ↵Gabriel Schneider2026-10-29
| | | | | | the pardes installed where it started, in place, by exec at a step boundary: its dump in the DumpDir, its pty masters, terminals, stubs and daemon link kept open and named on the new command line, every view and running shell its own again; the daemon re-execs too, its registry on its command line; a build that cannot load the dump falls back to the old binary, or starts afresh with the shells adopted and the dump kept (test/daemon.py --upgrade, --upgrade --gui) Co-Authored-By: Claude Opus 5.5 <[email protected]>
* 9p_io.zig: the /os table is not swept while a request is out in a syscall, ↵Gabriel Schneider2026-10-29
| | | | | | | | | | | | | | which made a walk of a directory just made answer ENOENT The tty frame's tick sweeps the table with the turn, but the turn comes back to the editor after a step of its own with a 9P request still out in a stat: the entry that request's walk had just made, which no fid named yet, was freed under it, and the next element looked up nothing. That was fs.py's one-in-a-few `/os` listing failure; walks of a fresh directory against file churn in the session's directory failed 144 in 2799, and none in 4440 after. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* 9p_io.zig: a session's liveness is a flock on its socket's .lock, not a ↵Gabriel Schneider2026-10-29
| | | | | | | | | | | | | | | | | | | connect, which FreeBSD refuses to a full listen queue as to a dead socket The owner holds `<socket>.lock` flocked from before it binds until after it unlinks (claim, release; the fatal-signal path unlinks it too). The startup sweep, the registry sweep, a registry entry's replacement and `taken` ask that lock: held is live, free is dead. They unlink only while holding it, so no new owner comes between the judging and the unlink, and two starters of one name are told apart by who took it. A socket with no lock, a pardes from before the locks, is still judged by a connect. The daemon's own start was already decided by its flock. The sweep test gains a live session whose socket refuses connects on every system (bound, not listening, its lock held) and a dead one's free lock, which goes with it; it passes on the FreeBSD box now. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* FreeBSD: the tty pardes builds for it and runs there, terminals, 9P and the ↵Gabriel Schneider2026-10-29
| | | | | | | | | | | | | | | | | | | | | | | | 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]>
* sessions register with the daemon: a terminal's session forked off its stub, ↵Gabriel Schneider2026-10-29
| | | | | | | | | | | | | | | | Detach and --attach in place for terminals and windows, pty masters kept, Adopt pardes in a terminal reaches (or starts) the daemon, opens /dev/tty and forks: the child is the session, in a session of its own (setsid), drawing on that terminal (tty.Term.fd, a raw vaxis Tty made from the fd); the parent stays as its stub, the job the shell waits on. Not the daemon spawning it from a passed fd, env and argv: the user's terminals each run in a 9ns user and mount namespace, which a session spawned by a per-user daemon would not share, and a fork costs one small process's page tables where the spawn cost an IPC and an exec. GUI windows and --detach sessions just register. tty: Detach (and a terminal's end: EOF, EIO, a hangup the stub passes on, taken on the signal thread) puts the terminal back, drops it and the stub's stdio, tells the stub (which ends, exit 0), and leaves the session running with nothing drawn (Host.watched), its messages on the clock and `size` its own. pardes --attach[=name] hands its terminal through the daemon; the session enters it as the first one, sized, queried and drawn whole, and that process is its stub until it leaves. A hangup never ends a kept session; TERM and INT do, telling the daemon so the shells go with it (one async-signal-safe send). gui: Detach closes the window (its swapchain or renderer) and keeps the device, atlas and pipelines; the loop sleeps on SDL's queue; pardes-gui --attach[=name] opens a window again. Closing a window still quits. --detach is a session started detached: the tty host with no terminal yet, the GUI with a hidden window closed at once. forkShell sends the daemon a dup of every master and the hosts tell it what they close, so a session killed outright leaves its shells running; the new Adopt word takes them in as terminal panes (the Tty word's pane, its spawn taking the pty, a 1x1 size then the pane's to make it redraw). The Attach word, which switched a session's own terminal to a detached session through the frames, goes, with its effect and request in the core. PARDES_DAEMON=off (the e2e harness's) runs a session as before. test/daemon.py: 50 detach/attach cycles at a level fd count with no stale socket, kill -9 and Adopt, and nothing left; --gui the same for a window. test/fs.py: a forked session's socket is found by its own pid, and a hangup leaves a kept session running. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Audit (misc): std mkdir -p and JSON escaping, a 9P late failure's errno from ↵Gabriel Schneider2026-10-23
| | | | | | ctl.failureErrno, dead switches and dumpHex out Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Audit (core, modal): goto/insert/select tables, edit through sam.Op and ↵Gabriel Schneider2026-10-23
| | | | | | sam.apply, Turn.bump and awaitPast, dead helpers out Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Callstack lists every caller and pages a tree too big for one query; memory ↵Gabriel Schneider2026-10-23
| | | | | | | | | | | | | | | | bounds and resident-size fixes Callstack walks while the server answers, counts on the message row past a second, puts test callers last at each level, and ends a tree past one query's memory (memory.limits.arena.lsp_query, a 256 MiB region taken and given back per query) in a row whose Callstack shows the next page. Grep reads whole files. New hard limits: undo text per file, image decode pixels, the LSP snapshot of other open files, one paste. Safe builds no longer fill worst-case buffers with 0xAA (memory.unfilled), big regions are kept off transparent huge pages, and malloc_trim runs after big results. Also: Look's git show pages into a +Diff, and dropped marks in File. 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]>
* Edit's w to a file changed on disk since read is refused once, as Save is (w ↵Gabriel Schneider2026-10-01
| | | | | | again to overwrite), the rest of an X/'/w written and the refused named; Edit filters get a pane command's environment ($PARDES_MOUNT, $PARDES_9P, $PARDES_PID, $winid); an Edit with no commands runs beside one whose commands run unless it touches a file that one changes, refused busy (EBUSY) as is a second Edit with commands Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An Edit's held write: < | > commands no longer block their 9P connection ↵Gabriel Schneider2026-10-01
| | | | | | (answered like a held read, so a status read or a filter reading the session's own mount runs beside it), a Tflush or hang-up of the write kills the commands' process groups and changes nothing; e loads by Get's way (asked once, clean after, undo puts the name back); ~ in e r w f B; B checks every name first; an Edit that runs commands is a write of its own, refused up front with other lines; Edit's +Errors output keeps the keyboard where it was; X goes in pane order; a refused write open says its errno; the reference's Coming from acme rows say what Get file's undo, failure and directory refusal, Putall's answer, Zerox across Dump and Incl's directories now do Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A refused Putall is answered once the other saves have landed; a bad Tab ↵Gabriel Schneider2026-10-01
| | | | | | width is invalid (EINVAL), from a pane's exec too; Incl refuses a relative or missing directory and takes - only alone Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Edit runs acme's file and shell commands: < | > through the session's Shell ↵Gabriel Schneider2026-10-01
| | | | | | off the loop (one undo step, a failure changes nothing, its stderr in +Errors, a 9P write answered when done), X and Y over the open text panes, b B D e r f w and the "file" address; checked against sam -d, and the reference's Edit section and Coming from acme row say so Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A close's held last line runs before the close is answered when no editor ↵Gabriel Schneider2026-10-01
| | | | | | step is out, so its err is in the log by then; with a step out it still waits for that step (follows rzxqzosq; monkey9p's one_failure_rule) Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Every pane pardes spawns gets $PARDES_MOUNT, the session's own directory ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | under 9ns --mntgen ($NINE_MOUNT/pardes/<name>), unset when it is not known 9ns says --mntgen with NINE_MNTGEN=1 (cloud9). Under `9ns --unix` the mount is the server it dialled, never this session (its socket does not exist until it starts), so it stays unset there; scripts fall back to `9p -a "unix!$PARDES_9P"`. Nothing is statted: the name comes from the session's socket and the environment. selfmount checks both modes. 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
* `answer z` quotes `z`, a bare `name` says it names nothing, and `Repl ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | klingon` is invalid (EINVAL), its message citing no doc path A refused answer quoted the word `answer`, not the choice refused. A `name` with nothing after it was an unknown control message. A Repl of no such language failed EIO with a pointer into the reference; it now says `invalid language` (EINVAL, here and through a mount) and how many more there are. A builtin's failure that says invalid anywhere is EINVAL, as 9ns reads it. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Mount dials: a relative socket path is a bad dial address, a socket path not ↵Gabriel Schneider2026-10-01
| | | | | | | | | | there says no such socket (ENOENT), and the hint says tcp! takes a numeric address `Mount peer rel/x.sock` dialled relative to wherever the editor ran; `Mount peer unix!/nonexistent` said `dial failed: no answer` (EIO); and the hint offered `tcp!host!port` where only a numeric address is taken. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A Mount dial with a network pardes does not dial (`nowhere!x`) is a bad dial ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | address, EINVAL, not a session name that does not answer A dial with a `!` and no `/` that was neither tcp!, quic! nor unix! was taken for a session's name, failing as `dial failed: no answer`, EIO. It is now refused as the malformed address it is, before any dial. A path with a `!` in it still dials. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A Rename that previews other files answers its preview pane to the 9P write ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | that asked it, and says on the message row that it previewed and applied nothing A multi-file Rename filled a +Search preview but the waiting exec write answered the pane it was asked from, and nothing was said. lspResponse now notes the pane it filled (fs.lsp_result); the write that waited for the answer reads that pane back, as a look or an exec answers the pane it went to, and the message row says `Rename: N edit(s) across files, previewed, not applied`. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A name under a directory that may not be searched or written is refused, ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | permission denied, and a Save there says so, not no such directory `pardes noperm/a/b/c` and `pardes /proc/1/root/x` took the missing directory for one Save would make, opened a pane and exited 0; its Save then said "no such directory". fs.deniedAbove finds the nearest directory there and asks whether it may be searched and written: forwarding refuses such a name, exit 1, "permission denied", and a failed Save says "permission denied", which the writer's 9P error carries as EPERM (9ns: EACCES). 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 log follower hears every record queued before a Restore: the read it holds ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | takes all that fit, before the hangup cuts it A follower reads one record per read, and a Restore hangs every client up, so records queued just before it -- the ones a script most wants, what led up to the Restore -- were lost: the read after the first was cut. Before the cut, each held read of /log is answered with every whole record queued that fits it. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* pardes - writes the whole text through one open of the body, so a page over ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | 8 MiB arrives whole, not as its last chunk Each 8 MiB chunk went in by its own Client.write, which opens a body with OTRUNC, so every chunk replaced the one before and a long page kept only its tail, with exit 0. The text now goes in through one open, truncated once and written at rising offsets; a long write is timed by its progress, each 9P write answered within 30 s rather than the whole transfer within 2. fs.py pages 11 MiB of numbered lines and compares the body byte for byte, and pages past the 256 MiB cap and checks the first 256 MiB and the cut note. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Building gathers the platforms, options, tests, release gates and how the ↵Gabriel Schneider2026-10-01
| | | | | | core, threads, detached sessions and 9P fit, from the old design notes checked against the code Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An editor's private pardes-osc133 prompt files go at Exit and at a fatal ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | signal, and a start sweeps the ones a dead editor left /tmp had gathered thousands of private prompt files: Exit removed its own, but an editor killed by a signal, or a crashed one, left them for ever. They now carry their editor's pid in the name, a SIGTERM or SIGHUP removes them before the signal's death, and each start removes those whose pid kill(pid, 0) answers ESRCH for. Files of the older shape carry no pid and are left alone, as a running older editor may still read them; the shared ones in the runtime directory are kept, being every session's. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An editor killed with SIGTERM or SIGHUP removes its 9P socket, and a start ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | sweeps the pardes-9p-<pid>.sock files whose pid is gone The runtime directory had gathered a hundred sockets of dead editors: only a clean exit removed its own, and a signal or a crash left it. The terminal's kill handler and, where nothing else handles the signals, a handler of the listener's own now unlink the socket before the signal's death. At listen, each pardes-9p-<pid>.sock whose pid kill(pid, 0) answers ESRCH for, that is a socket of this user's and that nothing answers on, is removed; a live pid's, a named session's and anything else are never touched. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The 9P socket appears already listening: it is listened on under a name of ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | its own and renamed into place A client that waits for the socket file and then connects, as fs.py's session helper and scripts do, was sometimes refused under load. The listener bound the socket at its final name, so the file existed a moment before listen(2), and a connect in that window got ECONNREFUSED. Now it listens under `<name>.<pid>`, is chmodded, then renamed over the final name, so that name only ever names a listening socket. A final name held by a live listener is still refused, and a name with no room for the suffix is bound in place as before. The registry test checks the socket is listening where it appears and that the temporary name is gone. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A connection holds up to 128 reads, the next refused "too many reads ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | | | | waiting: 128", and a mount beside 40 held event reads still answers Two limits met. pardes parked 32 reads per connection and refused the next with EAGAIN's C string, "Resource temporarily unavailable". 9ns kept 32 tags and 31 workers, each pinned by a held read. So 31 followers through one mount took every worker, and `cat layout` then queued for ever: the whole mount deadlocked. cloud9 (705be665, pushed to sr.ht and pinned here) now refuses in words past the cap. Its 9ns window is 256, so a mount always has workers past what pardes holds. pardes raises its cap to 128. fs.md documents the limit beside held reads. selfmount.py holds 40 event reads through the mount and reads layout beside them. It times out with the 9ns installed before this change and passes with the new one. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* pardes FILE in a pane whose session answers never falls back to a nested ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | | editor: the refusal is printed, exit 1, and no stray pane is left When the environment named a live session that answered, several failures broke out of forwarding and started a whole editor inside the pane that asked: a refused name, a pane the session has not, a failed look write. Its screen was drawn over the shell. Now, once the session answers (Client.probe), every refusal is printed as `pardes: <file>: <why>`, in the session's own words (the Rerror, now kept by the client), and the launch exits 1, as acme's B does. A new name the session refuses deletes the pane made for it, so no empty +New is left. A missing environment or a session that does not answer still starts a separate editor. fs.py checks a refused name and a stale PARDES_PANE. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A socket path too long for sun_path leaves no listener and the editor's turn ↵Gabriel Schneider2026-10-01
| | | | | | | | untouched: the GUI no longer calls into a freed one at startup listen started the editor's turn and pointed its answer_held and wake_parked hooks at the new Listener, then, when socketPath found the path too long for sun_path (108 bytes), freed the Listener with gpa.destroy while the hooks stayed set; the GUI's next rest (waitInput) called answerHeld on freed memory and died with SIGSEGV. The lapis bench's runs crashed for 12-character names in an 82-byte runtime directory (a 110-byte path) and not for shorter ones. The path is checked before anything is made; a unit test asks listen for a name too long and finds the turn's hooks clear and the turn not started. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* pardes --wait (-w) in a pane returns only once the pane its file landed in ↵Gabriel Schneider2026-10-01
| | | | | | | | is deleted, so it can be $EDITOR A launch inside a pane forwards its file to the session's look and returned at once, so fish's Ctrl-O (edit_command_buffer), git commit and crontab -e read their still-unedited file back and went on. With --wait, as acme's E against B and read the way plan9port's E reads acme's log, it finds the pane /index names the file by (the one already showing it, if open), follows /log on one connection with 'follow new', reads /index once more to catch a Del that came first, and then blocks with no deadline until that pane's del record: exit 0, or 1 when the connection ends with the session. The 9P client's one-shot requests keep their 2 s deadline; its new follow takes it for the setup only. Outside pardes nothing changes. --help, docs/fs.md, the README and the 9P skill say to set EDITOR='pardes --wait', which GIT_EDITOR follows. Tested in fs.py (it returns within 50 ms of the Del, a second -w waits on the same pane, one whose session is killed exits 1) and end to end: fish 4.8 in a detached session, echo hi, Ctrl-O, the waiting launch idle in poll, the line edited over 9P, Save, Del, and fish ran echo edited. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A failure said while a 9P write waits on it posts no msg, only the write's errGabriel Schneider2026-10-01
| | | | | | A failure the editor said as it performed a write's work (a Save's disk write, a shell, a language server) was posted as a msg and taken back out when the write's err was logged, which it cannot be once a log follower has read it: a follower saw both. Such a failure is now posted without a msg in the first place (fs.write_waits), and nothing is taken back. 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 path with a part over 255 bytes is refused, never a panicGabriel Schneider2026-10-01
| | | | | | std's statFile takes the kernel's ENAMETOOLONG for a bug (errnoBug), so a name, look, DumpDir or Save path with a part over 255 bytes panicked the editor (exec.kindOf via writeName, recentKeeps, dumpFailed). Every non-test statFile now goes through fs.statPath, which refuses such a name as NameTooLong first; fs.py drives long, looping, not-a-directory and not-ours paths through name, look, DumpDir, Dump and Save. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A language server question that finds nothing, cannot be answered or has no ↵Gabriel Schneider2026-10-01
| | | | | | | | room fails, and so does a write to an image's or PDF's body Callers or Callees with nothing found, Hover or Lspwhy with no room for its pane, and a question asked of no file's pane said nothing, the 9P write succeeding; a body write to an image or PDF pane went nowhere. A write that asks a language server now waits for its answer (turn.lsp_answers), and what fails there -- nothing found, no answer, a backend's @none why, a placement with no room (ENOSPC) -- fails the write and logs its err. No diagnostics is said, not failed. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A shell that cannot start fails Tty and pty/ctl exec with why, before either ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | | answers A shell whose exec failed (a script's missing interpreter) was reported started: Tty and pty/ctl exec succeeded, then the pane died. The child now reports a failed chdir or exec through a close-on-exec pipe the parent reads before acknowledging the shell; the host's spawn fails with the reason (`shell not found`, ENOENT), which fails the waiting write. A Tty whose first shell never ran leaves no pane; a terminal restarted by pty/ctl exec keeps its pane. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A Save into a directory that is not there says no such directory, ENOENTGabriel Schneider2026-10-01
| | | | | | | | It failed EIO with `file not found`, which named no file that was missing. It says `Save <path>: no such directory`, and a late failure saying `no such` answers ENOENT, as a builtin's does. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Mount dials its peer at once and fails if none answers; a peer out of reach ↵Gabriel Schneider2026-10-01
| | | | | | | | | | | | | is said as `dial failed: <reason>` Mount took any well-formed dial and nothing was found out until the first look through it, which then said `look: look: dial`. Mount now probes the peer (version, attach, its root) with the turn out, and fails its write, mounting nothing, when none answers; the dial errors read as words, `no answer`, `timed out`, `hung up`, and a look through a peer gone since names its path: `look: /n/peer/f: dial failed: no answer`. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A Save that fails logs its err, not a save recordGabriel Schneider2026-10-01
| | | | | | | | | The save record was logged as soon as the host was asked to write, so a Save into a read-only directory logged save and then failed. It is now logged only once the write is done; a failure's err record (the write's, naming the path) stands alone, the msg it was also said as dropped. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A command line cut across writes runs once whole, and look/exec answer per openGabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | A mount cuts a write at its message size, and each piece ran on its own: a line cut at the boundary ran as shell commands and a 50 KB Edit block broke. An open of look, exec, tagexec, any ctl or a column's exec now keeps an unfinished last line, or an Edit block still open, until its next write or its release, and never runs a fragment. The same open record holds what its last write touched, which a read on it answers (as /net/tcp/clone does); an open that never wrote reads the session's last as of its open. A release that runs a held line waits for quiet and counts as a change; a clicked line with a control character is still refused at its write. Tested in unit tests, fs.py (python client, 8 KiB) and selfmount.py (the kernel mount, a 1000-line seq burst and a 50 KB Edit block). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The QUIC listener tests are deterministic, and fs-test --quic passesGabriel Schneider2026-10-01
| | | | | | | | | | | | | | | | | | "Unix TCP and QUIC share one listener" failed 17 of 360 runs under load (12 parallel copies), always at p.fs.opens: the client had hung up but its connection's task had not yet seen it, reset() then cut the connection, and the releases the hangup owes were refused as the old editor's. The test's replacement editor is the same editor, so those opens stayed. It now waits (resting, with its own deadline) for the runner to empty before reset: 0 of 360 after. fs-test -Dquic=true failed too, at two points. A session Looking at its own tree through a QUIC mount is features.txt's documented "left as they are" case (QUIC is answered on the editor's thread); that check is dropped with a note. The /screen check asserted body text that the view may have scrolled past (2 of 7 runs); it now asserts the pane's tag. 10 of 10 after. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Shell prompt files are one per content under $XDG_RUNTIME_DIR, so a killed ↵Gabriel Schneider2026-10-01
| | | | | | | | pardes leaves none behind; tests remove their /tmp dirs Every pardes host wrote its own pardes-osc133-bash-* and -fish-* files to /tmp and removed them only at a clean teardown, so each killed session, test and crash left two: 72K of them had piled up. They are now written once per content, named by its hash, in the user's private runtime directory, renamed into place whole and shared by every pardes; without that directory the old private /tmp files remain. fish's -C source is quoted. The 9p_io tests remove their runtime dirs with what the listener left in them, and the snapshot runner removes its retry captures when every retry passed. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Sixteen 9P connections, and the next one is told there is no room instead of ↵Gabriel Schneider2026-10-01
| | | | | | | | being reset Four slots ran out under scripts plus a mount, and the fifth client was closed without a word, which through a mount looks exactly like a dead session. pardes now serves sixteen, and builds against cloud9 09b77cf, whose runner answers the Tversion of a client it cannot seat with an Rerror "too many connections" and tells pardes, which logs `err - 9p: too many connections`. QUIC keeps its own sixteen and still just closes. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A Save the host cannot do fails the 9P write that asked for it, and the log ↵Gabriel Schneider2026-10-01
| | | | | | | | names the path and why The write answered ok and the pane stayed dirty, with `save: AccessDenied` on its message row: a script saw success. The failure now says `Save <path>: <why>`, and a 9P write that waited on the save fails with EIO and that text, logged as an err record. Co-Authored-By: Claude Opus 5.5 <[email protected]>