summaryrefslogtreecommitdiff
path: root/src
Commit message (Collapse)AuthorAge
* Save path logs a save record naming the path it wroteGabriel Schneider4 days
| | | | | | | | | | | Save of a pane's own file logged `save <serial> <name>`, and so did a scratch's `Save path` (the pane takes that name). A `Save path` that writes a copy of a file pane, or a terminal's scrollback, elsewhere logged nothing, so a follower could not tell the file was written. It now logs `save <serial> <path>` with the path written, once the host has written it (events.noteLogAs). fs.md's log table says so. 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 Schneider4 days
| | | | | | | | | | | | | 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]>
* name refuses a directory, EISDIR, when written, not at the Save that could ↵Gabriel Schneider4 days
| | | | | | | | | | | | only fail `name ~`, `name /` or `name foo/` renamed the pane to a directory, and only its Save failed later. A name ending in `/`, the root, or one naming a directory that is there is now refused at once, with words a mount reads as EISDIR (`name: <path> is a directory, not a file`). The ctl word name and a forwarded `pardes ~` go through the same check. fs.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A terminal's word looks refill the directory's one +Search, as +Lsp is ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | shared, rather than stacking one per word Each word looked for in a terminal (clicked, or written to look), where no file of that name was, opened another +Search in the same directory, one per word, until the column filled. Such a search now refills the directory's +Search, as the file searches and the +Lsp answers already share theirs. A `/` search keeps a +Search per pattern, which n and N walk across (psearch.snap), so the sharing is marked on the word look alone (Pardes.word_look). 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 Schneider4 days
| | | | | | | | | | | 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]>
* fs.md says that in a > to a tag cut into several writes, those taken before ↵Gabriel Schneider4 days
| | | | | | | | | | | | | a refused one stand A `>` to a tag that a client sends in several writes is checked a write at a time. When a later one is refused, the earlier accepted ones stay in the tag. Buffering the whole open to apply it or none would cost more than it saves, so fs.md now says what happens: each write stands on its own, and only the refused one changes nothing. A test covers a truncation, an accepted write, then a refused one. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An addr, dot or limit pair out of order or past the text is refused as the ↵Gabriel Schneider4 days
| | | | | | | | | | | | address form is, not clamped `#5,#2` and `#2,#99` were refused (addresses out of order, address out of range), but the pair form `5 2` became 5..5 and `2 99` became 2..end, silently a different range from the one asked for. A pair is now checked the same way: past the text is out of range, and an end before its start is out of order. fs.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A Restore logs and names the whole path of the dump it reads, not the name ↵Gabriel Schneider4 days
| | | | | | | | | | | | | as typed `Restore x.dump.zon` read the DumpDir's x.dump.zon but logged `restore x.dump.zon`, and `~/d.zon` logged the tilde. A follower could not tell which file was restored. The builtin now resolves the name once (fs.restorePath: `~` expanded, a relative name the DumpDir's when one is there, else the current directory's) and hands that path to the host, which reads it and logs `restore <path>`. Its messages name it too. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A tag rewrite counts its text without the newline it drops: echo of exactly ↵Gabriel Schneider4 days
| | | | | | | | | | | | 4096 bytes fits `echo <4096 bytes> > tag` wrote 4097 bytes and was refused "over 4096 bytes", although a rewrite drops that one trailing newline and would have held exactly 4096. The limit now counts a truncating write's text without the newline it drops, for a pane's, a column's and the workspace's tag. An append still counts its newline, which it keeps. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Every line of a diff section that deletes its file looks for a file "deleted ↵Gabriel Schneider4 days
| | | | | | | | | | | | | by this diff", the diff --git and --- lines too In a section that deletes its file, a look at a removed or hunk line said "<file>: deleted by this diff". The `diff --git` and `---` lines said "no such file here", because they come before the `+++ /dev/null` (or `deleted file mode`) that says so, and the walk had not reached it yet. They now look ahead within the section, so every line gives the one message. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An empty range written back to event acts on the word a click there expands ↵Gabriel Schneider4 days
| | | | | | | | | | | | to, as acme's expand does `ML960 960` or `MX1 1` written back with no text acted on the empty range between the offsets: a look of nothing, an exec of nothing. A click at a point expands to the word or file name under it, and acme's written-back event does the same. It now expands with look.expandedWord, as a no-drag click does, on its line of the body or tag. fs.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A focus moved over 9P tells the programs that asked for focus reports, as a ↵Gabriel Schneider4 days
| | | | | | | | | | | | click does A program that set DECSET 1004 heard `CSI O` and `CSI I` when the keyboard moved by a click, a key or the window's focus, because update() checks after each event. A /focus write went through serveFs, which never did, so the program kept believing it had focus until the next input event. serveFs now runs the same check once a request has settled. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* diff -u in a command pane names its files though the terminal turned the tab ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | before each timestamp into blanks A `---`/`+++` path was cut at the tab before diff -u's timestamp. Run in a command pane, the terminal had turned that tab into blanks, so the name took the timestamp with it and a look at a changed line opened nothing. The name is now also cut where blanks run up to a timestamp, either GNU diff's `2026-09-30 10:00:00...` or ctime's `Wed Sep 30 10:00:00 2026`. One blank is enough, since a tab can expand to a single blank, and the stamp is strict, so a name with blanks and no stamp after them stays whole. Tested on the parsed text and, in fs.py, on real `diff -u` output in a command pane, where a look at `+TWO` opens the new file. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A 9P look keeps its leading blanks: an indented line, or blanks alone, is ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | looked for as written Every line written to look or exec went through Messages.next, which trimmed blanks from both ends. A look of ` indented` then found the first bare `indented`, and a diff's blank context line ` ` looked at nothing. A look's line is now its text: only its `\r` and newline go. lookAt keeps the blanks for the word search too; it expands `~` only for a look that starts with one, and the path, address and command readings trim for themselves as before. runSearch keeps a pattern's blanks, since a builtin's argument comes trimmed already. An exec line is trimmed as before. fs.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A restored pane whose file is gone from disk comes back dirty, and Del asks ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | before throwing its text away A dump keeps a pane's text. Restored after its file was deleted, the pane came back clean: the watch's baseline read of the missing file failed quietly, and Del closed it without a word. Its text was then nowhere at all. Restore now checks the file: a local path with nothing on disk comes back as one deleted on disk, as the watch would have said. It says so once, logs `changed N deleted`, and is dirty, so Del, Exit and Restore ask first. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A 9P write scrolls only the view: dot moves as acme's textinsert moves it, ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | never to the end of the write writeData, flushBatch and writeBody showed what they wrote with showOffset, which put the cursor at the write's end. The cursor is dot's head, so a data write or a body append moved dot there, and a `sel` write after it replaced the wrong text. showOffset now scrolls the view to the offset and leaves the cursor, pinned, where it was. Dot shifts only by the rule it already followed, acme's: an insert before dot shifts it, one at its start or inside it grows it, one at its end or after it leaves it. The test covers each case, then a body append followed by a sel write. Co-Authored-By: Claude Opus 5.5 <[email protected]>
*-. Release Pardes 0.9Gabriel Schneider4 days
|\ \ | | | | | | | | | Co-Authored-By: Claude Opus 5.5 <[email protected]>
| | * The window's F1-F24, Insert, Print Screen, Scroll Lock, Pause and Menu keys ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | reach a terminal's program, which they never did; every key a program gets follows its cursor keys' mode and kitty keyboard flags, pinned for both shells The keys went through ghostty's encoder already (DECCKM, the kitty flags); the GUI's native key map simply dropped the function and editing keys it did not name. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| | * A paste into a terminal is xterm's paste: bracketed when the program set ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 2004, and the control bytes a paste could smuggle a command in by, ESC first, become spaces, so no ESC[201~ in the text ends the brackets early Every paste path goes through edit.typeToTty (the GUI's and the tty's paste, p from a register, the acme chord's paste, a REPL's send), which now takes ghostty's paste encoder. A 9P write to a terminal's body or pty/data stays typed input, not a paste. The GUI test shell's stdin now takes a bracketed paste as one paste, as the tty does, not as keys. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| | * A program that asked for focus reports (1004) hears CSI I when its pane ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | takes the keyboard and CSI O when it gives it up, to another pane, a tag or another window The core has a window-focus event now, from SDL's window focus and from the tty's own focus reports (vaxis turns 1004 on outside); the detached wire carries it: version 10. The test shell's ESC]777;mouse;focus;<1|0> fakes the window's. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| | * A program that tracks the mouse gets its clicks, drags and moves over its ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | grid, in the format it chose; Shift keeps pardes's select, execute and look, and chrome is never the program's Presses and releases under 1000, drags with a button down under 1002, every move once per cell under 1003, X10 presses under 9, in raw terminal input only. A press the program took keeps its drag and release, reported clamped to the grid's edge wherever the pointer goes. The press also makes the pane the active one. docs/tags.md says so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| | * A mouse event carries Shift and Alt as well as Ctrl, from SDL and from the ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | tty's input; a program's wheel report carries them as xterm counts them, and Shift keeps the wheel pardes's own scroll The reports are ghostty's encoder now (shift+4, meta+8, ctrl+16), with the mode and format ghostty tracks. The detached wire carries the two new bits: version 9. The test shell's ESC]777;mouse takes an optional fifth field of modifiers. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| * | A right click on a diff line opens the place it names: its file, the hunk's ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | new line, the line itself In a diff pane (a .diff/.patch file, or a command pane's diff output), the look's own expansion, which its hover shows, takes a whole line for a `diff --git`, `---`, `+++` or `@@` line wherever the pointer is on it, markers included, and for a hunk line when the pointer is on its `+`, `-` or space. Then the look opens the file (a/ b/ dropped, timestamps cut) at the line from the hunk header and the lines above it: `@@` the hunk's first new line, context and added lines their own, a removed line the new line now where it was. `diff --git` and `+++` open the new file; `---` the old one, unless the `+++` under it names another. On a hunk line's code the expansion is its words, the prefix never one of them, and looks as ever. Other panes are untouched. Paths resolve in the repository (walking up to `.git` or `.jj` from the diff's directory, or a command pane's), then that directory. A file not here opens nothing and says `Look: b/x: no such file here`, or `deleted by this diff`. A 9P look writing a whole line of the diff does the same. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| * | A diff reads as the code it changes: hunks in their file's language, added ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | and removed rows tinted A .diff or .patch pane, and a command pane whose output is a diff (`git diff`, `git show`, `diff -u`; told by a `diff --git` line, or `---`/`+++` with a `@@` under them, in its first rows), colour each hunk's code in the language of the file its section names (`+++ b/<path>`, `--- a/<path>` for a deleted file), with tree-sitter, as highlightLocations does for results. src/diff.zig walks a unified diff by its `@@` counts, so a removed line that reads `--- x` is no header. Each side of a hunk is parsed as one text (context and removed lines the old side, context and added lines the new), so a string or comment across lines colours as it does in the file; a side no line takes its colours from is not parsed. A section's hunks share a parse, in pieces of about 40 lines (a hunk past 80 is cut), only those in view, and each piece's colours are kept by its bytes, so scrolling back and a terminal's every repaint parse nothing again. Added and removed lines carry a flag in their style byte; the painter tints their rows to the pane's edge, a little way from the page toward the theme's ANSI green or red, and draws the prefix in that hue pushed to read on it. A file in no known language keeps the old line colours. A command pane is read once its command has finished: its rows are copied once (File.DiffOutput, dropped when the pane runs again), so no frame dumps the scrollback, and painted over what git printed. A running command is shown as it prints. 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 Schneider4 days
|/ / | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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]>
* | Release Pardes 0.8Gabriel Schneider4 days
|\ \ | | | | | | | | | Co-Authored-By: Claude Opus 5.5 <[email protected]>
| * | A background job killed while the tty probe walks it is not the tty's owner, ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | however far into exiting it is host_io's "a background job is not the tty's owner" failed now and then, 2 runs in 20 under load, always just after `kill %1`. The probe walks the shell's children and reads each one's exe. A killed background sleep, in the moment it is exiting (its exe link gone, not yet a zombie), has no exe to read and is not a zombie to skip, so the probe called the tty taken. A process outside the foreground process group holds nothing whatever state it is in, so one whose exe cannot be read is now skipped when its group is not the foreground one. Under the same load the test passed 50 runs in a row. The unit-test step's perf gate trips under that load, which is not this test. Co-Authored-By: Claude Opus 5.5 <[email protected]>
| * | The 9P socket appears already listening: it is listened on under a name of ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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]>
| * | pardes --wait for a new name waits on the pane it made, by serial, and a ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | pane closed before it looks is done, not a failure fs.py's `pardes --wait 'wait new.txt'` failed two runs in three on the 0.0.8 merge. The stress case repeats the test's steps: 4 of 50 failed under load, each with "pardes: --wait: no pane shows that file". A new name's launch makes its pane, names it, then found it again by path in /index. The script had already seen the name, saved and removed the pane by then, so the lookup missed and --wait exited 1. Now the launch hands --wait the serial pane/new gave it. After a look, finding no pane on the file means that pane is already closed, which is the end --wait waits for: exit 0. The follow with its /index recheck covers a close in between, as before. Under the same load the stress case went 0 of 100. The same load showed a test race too, at fs.py's second Restore. The exec write is answered before the hang-up, but the client's clunk after it can meet the connection already cut. The test now requires the answer and tolerates the reset. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | | A path typed with a leading ~ is the home directory, at every entry point ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | that takes one, through one helper Only DumpDir (dump.zig) and ShaderBuild expanded `~/`, each its own way. So a look at `~/notes.txt`, `name ~/x`, `Save ~/x`, ThemeFile, Restore and a quoted `pardes '~/x'` all took `~` as a directory of that name under the pane's. There is now one rule, filesystem.expandHome. A leading `~`, alone or before `/` or `:`, is $HOME, else the passwd entry's home; `~user` is that user's (getpwnam). As in a shell, it applies even beside a file named `~`, and `./~` names that file. Every entry point runs its typed path through it: - look, B3 and 9P, before any parsing, so `~/x:12`, `~/x:12:3` and `~/x:/re/` all work; - name, and a tag's rename; - Save <path>; - ThemeFile, DumpDir, Restore and ShaderBuild's shader files; - the forwarding launch. Recent rows are not shown with `~` and need nothing. How names are shown is unchanged. Tests: expandHome's cases, and HOME unset (passwd), with HOME passed in rather than set, since setenv moves the environ a spawned child reads. fs.py covers Save into missing directories under `~`, `name ~/x`, a look at `~/x:3`, and a forwarded `pardes '~/x'`. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | | Merge: round 26 + codex wheel (the 0.8 candidate)Gabriel Schneider4 days
|\ \ \ | |/ / |/| / | |/ Co-Authored-By: Claude Opus 5.5 <[email protected]>
| * A wheel over a terminal whose program owns it goes to the program: its mouse ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | report when it tracks the mouse, arrow keys on the alternate screen with alternate scroll, so codex, htop and less scroll codex 0.158 moved its chat onto the alternate screen with any-event SGR mouse tracking (1049, 1000/1002/1003/1006, 1007 off); its sign-in screen keeps 1007 on with tracking off. pardes never forwarded a wheel to a program, only scrolled ghostty's scrollback, which the alternate screen does not have, so the wheel did nothing there. A shell on the main screen still scrolls its scrollback as before. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A detached session no client watches draws no frames: yes | head -c 50M ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | through pty/run goes from over 30 s to 11 s Profiling the 50 MB run showed 70% of its time in draw.render, recolorAnsi alone 43%, walking cells through PageList.pin. The detached session drew a frame after every pty read, with no client attached to see it. A host now says whether anyone watches (Host.VTable.watched). The detached server answers whether a client is attached, and while none is the core keeps the frame owed instead of drawing it. A /screen read draws its own, and a client attaching gets one (hello sets needs_frame). Debug, 80x24: 10 MB went from 7.36 s to 2.10 s, and 50 MB from over the 30 s client timeout (about 37 s by rate) to 11.08 s. What is left is the terminal's own parsing and scrollback growth. A session a frontend shows still draws after every read. Pacing those frames is the next step if that case matters. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | pane/new takes rows from a pane in another column before refusing ENOSPC, as ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | acme's placement does A pane/new whose column had no room was refused "no space for a pane in that column", even with another column holding a tall pane that could give rows. acme takes rows wherever a window has them before saying there are none. pane/new now tries the other columns as +Errors already did: an empty one, else its tallest pane's bottom half, last column first. It is refused only when no pane anywhere can give the rows. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | Kill with a name and nothing running says "Kill: nothing running" and ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | succeeds, as bare Kill does Bare `Kill` with nothing running is a message, not a failure: a script's Kill has nothing left to do. `Kill make` in the same state failed the write with "no running command has that first word", as though make were the mistake. With nothing running at all, named or not, Kill is now the message. It still fails when commands run but none has the word. The two tests that used `Kill zzz` as a builtin that always fails now use `Unmount zzz`. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | An event record written back without its trailing newline is taken, as every ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | other line file's last line is `printf 'MX0 10' > event` was refused as a malformed record, while exec, look and every ctl take a last line with no newline. The event file now does too: a write not ending in a newline is read as though it did, and the write still counts the bytes that came. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A tag write over the limit is refused "no space: over 4096 bytes", so its ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | err record says tag: once The refusal's text began with "tag: ", and the log's err record puts the file's name before the reason. So the record read `err - col/1/tag: tag: no space: over 4096 bytes`. The text is now the reason alone, as the control-character refusal's is, and the record names the file once. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | Save makes the directories its file goes in, always: for the pane's own name ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | and Save path alike Whether a Save made missing directories depended on how the name had come about. Config's pane did, a name written into a missing directory did (since kkrqxtsp), and `Save path` and every other name did not, failing ENOENT instead. There is now one rule: a Save makes the parent directories of the file it writes, a pane's own or a terminal's scrollback included, and the make_dirs flag is gone. A directory it cannot make still fails the write with why, as fs.py's /nonexistent-pardes-dir case shows. fs.py now also saves into missing directories by Save path and by a written name. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A look at file:3:0 is refused as addr 3:0 is: a column counts from 1Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | addr refuses a column 0 with EINVAL, "address out of range: a column counts from 1". A look took `file:3:0` as line 3 with no column, since the parser holds no column as 0, and succeeded. Now a look whose column, or end column, is written as 0 is refused with addr's words and errno. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | DumpConfig's lines that set nothing start with #, and a line starting with # ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | runs as nothing at an exec, a ctl or the init file DumpConfig's information lines ("Platform: tty", "Startup config: ...") began with words. Written back, as the report means to be, or clicked, each ran as a shell command or failed as an unknown control message. Each now starts with `# `. A line starting with `#` is a comment, as in a shell: it runs as nothing, silently, at an exec, at the root's, a pane's or a column's ctl, and in the init file. Documented in fs.md and config.md. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | winsize refuses fewer than 2 rows, saying so, and its range reads 1x2 to ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | 4096x4096 as the docs say `winsize 20 1` was taken and quietly raised to 2 rows, so the write succeeded with a size other than the one asked for. The range refusal also said "1x1 to 4096x4096", although fs.md says at least 2 rows. One row is now refused with "invalid winsize: at least 2 rows", and the range text starts at 1x2. size already refused anything under 20x6 with a text that names it. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A run whose output a redraw moved above its start answers cut, never an ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | empty whole pty/run reads a command's output between its C and D marks. When the end landed at or before the start, it answered `exit 0` with no output, as if nothing was printed. That happens when the shell redraws its prompt over the output, as it can in a window two rows high (winsize 10 2). Now only equal marks mean nothing was printed. An end at the top-left cell or above the start says `cut`, since how much was printed is not known. bash, fish, long lines and 3000-line outputs at winsize 10 2 all came back whole on this build, so the dogfood's loss could not be reproduced directly. The test drives the marks into the inverted case the loss needs. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | With no pane at all, the root look opens its file and the root exec runs its ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | command in the session's directory /look and /exec run at the active pane, and with every pane closed they failed ENOENT, "file does not exist", about a file the writer never named. Now an exec runs as the workspace tag's would, in the session's directory: a builtin at the stand-in, a command in a command pane. An empty window's click now runs a command line too, not only session words. A look runs from a scratch made as New makes one, which also makes the column. That scratch is dropped, unannounced, when the look went to another pane, so opening a file leaves only the file. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | Lspinfo, Lspwhy and the other +Lsp answers share one +Lsp pane per directoryGabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | Each query kind found its own output pane by its origin. So Lspinfo and Lspwhy, which both write `<dir>/+Lsp`, opened two panes of the same name in one directory, and so could a code action list or a format diff. The language server's prose now shares one +Lsp per directory, the way the file searches share one +Search, and each answer replaces the last. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | Lspwhy explains the recorded last query in its own file and place, wherever ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | it is run Lspwhy kept the last query's kind and argument, but ran from another pane it asked in that pane's file at its cursor. So it narrated a different lookup from the one that failed. When the pane the query was asked in is still open, Lspwhy now asks in it, at the query's offset, whichever pane it is run from. With that pane gone, it falls back to the pane it is run in. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A Newcol refused because the panes' tags would not fit logs its err alone ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | and spends no column serial Newcol tries its column before keeping it: the new column is inserted, the panes' rows are shared out again, and where a pane would be left under its tag and two rows the column is taken back. The trial showed. insertColumn logged `newcol N` and the take-back `delcol N`, so a follower saw a column come and go around the err. Laying the column out also gave it a serial, so the next column made skipped one. The trial is now unannounced, since insert and drop have unlogged forms. A refusal puts the serial counter back where it was, after first giving every existing column its serial so the trial's is the only one spent. Only a column that stays is logged, and it gets the next serial. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | A connection holds up to 128 reads, the next refused "too many reads ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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]>
* | A tty editor killed with SIGTERM or SIGHUP puts its terminal back before it diesGabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | A nested pardes killed in a pane died with the default action. That left the pane's terminal on the alternate screen, in raw mode, with mouse and paste reporting on, so the shell under it was unusable until a reset. The tty frontend now handles both signals: it writes the resets a clean exit writes (keyboard protocol, mouse modes, focus, bracketed paste, SGR, cursor, main screen) and restores the cooked termios, all async-signal-safe. Then it dies of the same signal, since SA_RESETHAND put the default back. fs.py kills a tty session with each signal and checks for the main-screen switch and cooked mode; it fails without the handler. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* | pardes FILE in a pane whose session answers never falls back to a nested ↵Gabriel Schneider4 days
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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 backward search along one long line takes time in the line's length, not ↵Gabriel Schneider4 days
|/ | | | | | | | | | | | | | | its square: ?a? over 117 KB goes from 6.4 s to 0.03 s The 9P fuzzer (seed 124) hung a session for over 5 s on `?a?` in a 117 KB file of one line. A backward search is every forward match up to the limit, so Regex.find ran once per match. Each call looked back from its start to the line's start, and forward to the line's end, over the whole line each time. That is quadratic, and the step budget, which counts only mvzr's steps, never saw it. find now keeps where it last learned a line starts and ends in the same text, and looks back only to there. The literal prefix's look back stops at the known part of the line. Co-Authored-By: Claude Opus 5.5 <[email protected]>
generated by cgit v1.3 (git 2.53.0) at 2026-10-04 19:00:17 +0000