summaryrefslogtreecommitdiff
path: root/src
Commit message (Collapse)AuthorAge
* Every refusal over unsaved text names a scratch by its pane tooGabriel Schneider2 days
| | | | | | | | | | | Exit, Restore, Del and Delcol share one list (warnModifiedIn), but it was built in 256 bytes, so with a few long paths a +New scratch near the end fell off Exit's and Restore's, which list every pane; and several +New panes were indistinguishable. The list is built in 1024 bytes (the write's error is still cut to fit, between words), and a scratch is named with its serial, /dir/+New (pane 12), the same everywhere. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A chorded exec an event reader holds is told its argument, and runs with itGabriel Schneider2 days
| | | | | | | | | | | | With event open, a 2-1 chord's exec went out as one record with flag 8 and nothing after it: the argument was freed unsaid. As acme does (exec.c:150-190), two records follow it, the argument's text and where it came from (file:#q0,#q1, empty when not known), and the pane keeps the argument, so the record written back runs with it (acme's own write-back, xfid.c:842, passes none; the reader was told, so it runs as the click would have). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A looked word selects its next place in its own pane, as acme's look3Gabriel Schneider2 days
| | | | | | | | | | | | | | A plain word looked at (a look write, a right click, an ML record written back) opened another identically named +Search pane each time, until the panes ran out, and never wrapped. As acme's look3 searches the text for it (look.c:210), it now selects the word's next place in that pane after the dot, wrapping at the end, and opens nothing. LookWord list, a new root setting, keeps the +Search listing (search is the default); the rsearch snapshot and an output test that walk the listing say LookWord list. (The leader-path check over every builtin gets a comptime branch quota, as one more builtin took the GUI build past the default.) Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The docs: sam gotchas, Restore's log order and paths, stream reads, layout's ↵Gabriel Schneider2 days
| | | | | | | | | | | | | | | | | current and active Round 13's doc findings: a sam gotchas box in the skill ($-1 with and without a final newline, N+1 at the end, ^/$ at the end, 2,1, y's leading empty piece, c's literal &, columns from 1); the skill's Restore log order made the true one (new, restore, restored, restoredcol); a relative Restore path is looked for in DumpDir then the session's directory, and a bare one takes the last dump; a follower sees each (xN) as a new line; reading data or xdata moves addr; a second read of look/exec on one fd is EOF (re-open or seek 0); /layout's current (the keyboard's column) and active (acme's activecol) defined; no room, not no rows left; why a scratch's dirty blocks nothing under 100 bytes; the changed record listed. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A bare Shell names the shell it goes back to, and 12:0 says columns count from 1Gabriel Schneider2 days
| | | | | | | | | A bare Shell reset to the default without saying which; it now says Shell: the default again, <shell> ($SHELL, else /bin/sh). A line:col of column 0 said only address out of range; it now adds that a column counts from 1. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A column's ctl refuses as the other ctls do, and a bare answer says what it ↵Gabriel Schneider2 days
| | | | | | | | | | | | takes A column's ctl refused an unknown word in words 9ns maps to EIO, and took New extra as New. It now refuses the word as unknown control message (EINVAL, as the root and pane ctls) and any word but Tty with an argument as wrong #args. A bare answer on a pane's ctl says answer takes a choice or -, not unknown control message. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* winsize of no size says out of range, bounded as size isGabriel Schneider2 days
| | | | | | | | | winsize 0 0 was refused as ill-formed, as if the words were wrong, and any u16 was taken. Numbers that are no size (0, or past 4096 a side, the root ctl size's bound) now say invalid winsize: out of range, 1x1 to 4096x4096 (EINVAL through a mount). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A file changed on disk never reloads over unsaved editsGabriel Schneider2 days
| | | | | | | | | | | | | | | A watched file rewritten on disk was loaded over a buffer's unsaved edits and its dirty flag cleared, so Exit and Del no longer asked about them. A clean buffer still reloads; one with edits keeps its text and stays dirty, says <file> changed on disk (get reloads it, Save overwrites it), logs changed <serial>, and its Save warns once before overwriting the newer file, as acme's Put refuses a file modified since read (exec.c:577); get asks as ever (winclean, exec.c:513). The restored-watch test's dirty case now keeps the snapshot, and the filewatch snapshot script (re-recorded by name) checks both halves: the edited pane keeps its edit and says so, the clean one reloads. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A multi-line paste into Python's REPL runs whole, back at its promptGabriel Schneider2 days
| | | | | | | | | | | | The Enter went on Python's first output after the paste, before it had taken in the whole paste, and one Enter leaves a multi-line input at ... anyway. Python's output now only restarts a 100 ms quiet wait (the 250 ms fallback stays), and a paste of more than one line gets the second Enter a block gets. fs.py runs print(11)\nprint(22), the same with a trailing newline, and a for block followed by a statement in a real python3, each back at >>>. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An Edit text block is found in linear time: only a . line is parsed againGabriel Schneider2 days
| | | | | | | | | | | Finding where an Edit block ends parsed the whole block again for every line, copying it into the scratch arena each time: a 10k-line $a block took minutes and tens of GB. sam_edit.waitsFor says what the block waits for, and while that is a text block's . line, only such a line is parsed again (editEnd, shared by Messages and completeEnd), its copy freed. A 10k-line block now takes well under a second and 50 MB of scratch. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The docs say what Restore renumbers and re-runs, what Tty takes, and how a ↵Gabriel Schneider2 days
| | | | | | | | | | | | | | | range reaches a REPL Round 12's doc findings: the README's ctl line gains Restore, Placement and size (still 45 lines, detail in fs.md); Tty's argument is a shell and Tty on a pane's ctl opens a new pane; a .py pane's range goes to its REPL as an MX<q0> <q1> event write-back; after a Restore panes are renumbered (columns count from 1 again, so often keep theirs), a command pane shows exit ? and does not re-run, and REPL bindings are gone. acme's p adds no newline to what it prints (ecmd.c:744, pdisplay), so pardes's p does not either. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A command run from a column's tag gets its pane in that columnGabriel Schneider2 days
| | | | | | | | | A command from a column's tag ran in a command pane at the foot of the last column, however far from the tag, and could reuse a finished one anywhere. It now goes to the tag's column, under both placements, and reuses only a finished command pane in that column. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* pty/run's busy names the program holding the terminalGabriel Schneider2 days
| | | | | | | | | busy said only that something ran. Where the host can tell (Linux, the tty's foreground group's /proc comm, a new fg_name host call), it says busy: <program> is running; elsewhere, busy as before. The GUI's host gains the one small callback (render agent's file: gui.zig, fgName). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* unlock's refusal says pane, not windowGabriel Schneider2 days
| | | | | | | "window not locked by this open" was acme's word; pardes has panes. Its test compares the constant, so it follows. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Repl's unknown-language refusal names a few whole and points to the restGabriel Schneider2 days
| | | | | | | | | The refusal listed every language and was clipped to fit, ending mid-word. It now names six, those that start as the word did first (pithon: python), says how many more, and points at docs/tags.md, which lists them all. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An event write-back past the text says range past end of bodyGabriel Schneider2 days
| | | | | | | | | A well-formed record whose range ran past the body (or tag) was refused as "bad event syntax", which sent a client looking for a typo. It now says "range past end of body" (or tag); a malformed one keeps the old words. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* size says its limits, keeps every pane its rows, and ptys follow their panesGabriel Schneider2 days
| | | | | | | | | | | | size refused an out-of-range line without saying the range, took a size too small for the panes (leaving bodies of no rows), and a terminal squeezed to its tag kept a pty of its old size. The refusal now names 20x6 to 4096x4096 (a bare size is EINVAL like any malformed line), a size that would leave any pane less than its tag and 2 rows is refused and undone, and a terminal's pty is resized with its pane down to one row (only a folded or unplaced one keeps its grid). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Tty and Shell refuse a directory: not a shellGabriel Schneider2 days
| | | | | | | | | | A directory passes access(X_OK), so Tty /etc made a pane whose shell exited 127 and Shell /etc was taken. The shell lookup now refuses a directory, and both say "not a shell: /etc is a directory" (Tty and Shell share one refusal, host_io.Shell.refusal). shellset's golden takes the shared wording (re-recorded by name). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A REPL paste's Enter waits on the clock, not on stale core timeGabriel Schneider2 days
| | | | | | | | | | | | sendToRepl timed the Enter's fallback from stepped_ns, which in an idle detached session is as old as the last frame: the wait was over before it began and the \r went out with the paste, which 3.13's REPL then reads as part of it (4 of 5 sends lost their Enter). The wait now starts from the host's clock and is checked against now_ns; the REPL's echo of the paste still sends it sooner, and the wake is scheduled for the deadline. fs.py sends five times after two idle seconds each in a real pty. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A pane restored with unsaved text is asked about before it is thrown awayGabriel Schneider2 days
| | | | | | | | | Exit, Restore, Del and get warn only about a pane edited since it was opened, which its saved hash says (set at its first edit). A pane a Dump kept dirty came back with no hash, so a restored session quit at once over it. Restoring one dirty now gives it a hash its text does not match. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* dump.zig's tests run: nothing had included themGabriel Schneider2 days
| | | | | | | | dump.zig was imported by the code but by no test block, so its tests, the dump settings and restored-dot ones among them, never ran. The pardes test block imports it; they all pass. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The docs give the full-column refusal, an Edit block left open at close, and ↵Gabriel Schneider2 days
| | | | | | | | | | | | byte columns Round 11's doc findings: the skill names the full-column refusal beside the 64-pane cap; fs.md and the skill say an Edit block still open when its open closes fails with unmatched { and an err; and line:col's byte columns get the tip for character columns from tools (the same on ASCII, else a search or #n). A unit test pins the unmatched { at close. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A detached session with no frontend takes its size from the root ctlGabriel Schneider2 days
| | | | | | | | | | A --detach session nobody is attached to stayed 80x24, too small for an agent's panes, and nothing could change it. The root ctl takes size <cols> <rows> while no frontend is attached (refused while one owns the size), and a client that attaches later still sets its own (the server compares with the core's size). Documented with the 80x24 default. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Words from the workspace and column tags run in the session's directoryGabriel Schneider2 days
| | | | | | | | | | A command or Tty run from the workspace tag or a column tag started in the focused pane's directory, though those tags have none of their own. As acme's row and column tags do, they now run in the session's directory, where pardes started (Pardes.sessionDir); a pane's tag and text keep the pane's. The same for /tagexec, a column's exec and ctl. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* +Errors with no room anywhere logs its text rather than fail the EditGabriel Schneider2 days
| | | | | | | | | Edit's p and = put their output in +Errors, and when its column had no room the placement refusal failed the whole Edit. +Errors now tries every column with room, the last first, and with none its text is logged as msg records, a line each; the Edit succeeds. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An exec line over 1024 bytes fails the write, naming the limitGabriel Schneider2 days
| | | | | | | | | | A command line longer than the 1024 bytes a command pane takes was said on the message row after the write had already succeeded. An exec write (the root's, a pane's, tagexec, a column's) now refuses it before anything runs, EINVAL with "a command line is at most 1024 bytes" and an err record; a builtin's line, a long Msg or an Edit block, is not limited. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A failed builtin's err replaces its msg, and (xN) is a running totalGabriel Schneider2 days
| | | | | | | | | | | A builtin failing a ctl write logged its Verbose announcement, its words as a msg, and the write's err, so the same failure again never repeated the newest record and never collapsed. Its err alone is logged now (the msg is still shown and kept in +Messages). A repeat of a record a follower has read is a new line with the running total, (xN) being N in all. fs.md states both rules. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Shell refuses a name that is no executable, and bare restores the defaultGabriel Schneider2 days
| | | | | | | | | | | Shell took any word and the host quietly ran another shell at the next spawn. It now refuses a path or name that is not executable ("Shell: x: no executable by that name", failing a ctl write, logged err), a bare Shell goes back to the default, and a default $SHELL that is not executable falls back to /bin/sh. shellset's golden takes the refusal on the message row (re-recorded by name). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Edit's x, y and s match whole runes, as addresses doGabriel Schneider2 days
| | | | | | | | | | | mvzr matches bytes, and only the addr path widened a match to runes, so ,x/./a/|/ on é split it into invalid UTF-8 and ,s/./X/g gave a combining sequence four X's. Edit's matches now go through the same rune alignment (and an empty match steps a rune, not a byte). runeStart no longer walks a stray continuation byte back onto the ASCII before it: an invalid byte is a place of its own. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* No placement leaves a pane shorter than its tag and two rowsGabriel Schneider2 days
| | | | | | | | | | | | | | | | Placement could split a pane down to a tag and one row, or give a new one less. Every placement now checks the pane it made and each pane it shrank against its tag (wrapped at its width) plus two rows: where they lack the room, the column's tallest pane is halved instead, and where none can give it the new pane is refused with "no space for a pane in that column" (ENOSPC over 9P) and closed at the step's end, once its maker is done with it. splitBelow and splitParent leave two body rows (a boot layout, split before any row exists, is as before). Re-recorded by name: lookpanes (panes one or two rows lower) and splitparent (the 6-row squeeze now refuses); fs.py closes the scratches and long-named panes it no longer needs, since its 24-row session filled. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The default Shell is $SHELL, else /bin/shGabriel Schneider2 days
| | | | | | | | | | With no Shell set, pardes ran fish, which a box may not have and a user may not use. The default is now the user's login shell ($SHELL), else /bin/sh; an explicit Shell still overrides it. The tests pin their own shell (snapshots bash, the python suites /bin/sh or bash) and pass with SHELL unset; the Config report and docs say where the default comes from. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A read on a closed pane's event ends ENOENT, and the pane cap is ENOSPCGabriel Schneider2 days
| | | | | | | | | | 9ns reads an error's words, not its errno: "window shut down" came out as EIO where every other file of a gone pane gives ENOENT, and "no pane slots" as EIO. The event and pty/data reads a pane closed under now say "no such pane: its window shut" (ENOENT), and the cap says "no space for a pane: 64 max" (ENOSPC; 9ns has no word for ENFILE). Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A pane closed with its question standing logs answer <serial> -Gabriel Schneider2 days
| | | | | | | | A follower that saw ask never heard the question end when its pane closed first. Retiring a pane with a question still up now logs answer <serial> - before its del. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A msg said again word for word is that line counted in the logGabriel Schneider2 days
| | | | | | | | | | 1000 Undos with nothing left made 2000 log lines: the Verbose announcement and the answer each time. A msg repeating the newest record is now that record counted, (xN), as err's already were (shared as pushCounting), unless a follower has read it; and a builtin's own words that start with its name take the place of its announcement just before them. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Undo and Redo with nothing left say so, and the write succeedsGabriel Schneider2 days
| | | | | | | | | A pane's Undo or Redo with no step left did nothing and said nothing. It now says "Undo: nothing to undo" (or Redo's) as a message, not a failure, so a ctl write of it still succeeds as acme's silent Undo does. The docs give the 256-step history. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* addr reads empty after a failed address instead of failingGabriel Schneider2 days
| | | | | | | | With no address after a failed write, a read of addr itself failed (EIO through a mount), which reads as a broken file rather than as no address. It now reads empty, as an unset limit does; data and xdata still refuse. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A limit write that fails keeps the limit it hadGabriel Schneider2 days
| | | | | | | | | A shell's > truncates limit, which lifted it, before the write; a write that then failed left no limit at all. The truncation now keeps the old limit until that open's write, and a failed write puts it back; a bare truncation still lifts it once its open is released. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Addresses snap to runes, not grapheme clustersGabriel Schneider2 days
| | | | | | | | | | sam and acme address runes; pardes snapped #n, line:col, dot and data's reads to grapheme clusters, so a lone combining mark or a CRLF's \r could not be addressed. #n, line:col, a search's match, dot (both ways) and data now land on rune boundaries (modal.runeStart and friends), and the tty's cell drawing is untouched. The README, fs.md and the skill say so. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A line address past the last newline is sam's empty line thereGabriel Schneider2 days
| | | | | | | | | Forward line addresses followed acme's number(), which refuses the empty line after a text's last newline, so 1 of an empty buffer and 2 of "a\n" were out of range and Edit 1i/header/ failed on an empty file. They now follow sam's lineaddr (editors/sam/address.c:180-215): #0,#0 and #2,#2. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A command line cut across writes runs once whole, and look/exec answer per openGabriel Schneider2 days
| | | | | | | | | | | | | | | | 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 workspace tag has /tagexec, every exec file reads back what it touched, ↵Gabriel Schneider2 days
| | | | | | | | | | | | | | and a tag's words log as the session's /tagexec runs a word as a click in the workspace tag, as col/<serial>/exec does in a column's. A column's exec was write-only; it and /tagexec now read back the serials the last write touched, as /exec does. A click in a tag no longer goes to the focused pane's event reader, and what a word run from a column's ctl or exec or /tagexec says is logged msg -, not against the focused pane. The README, fs.md and the skill say so, and document the answer record. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Every answer to a pane's question is logged, answer - includedGabriel Schneider2 days
| | | | | | | | | The log said ask but never what came of it, so a follower could not tell a question taken back from one still standing. answerQuestion now records answer <serial> <choice|-> whichever route answered, key or ctl; a key the question does not offer is -. (Docs are in the tagexec change.) Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A look's miss repeats the path whole, up to 256 bytesGabriel Schneider2 days
| | | | | | | | | The miss message clipped what was looked at to 48 bytes, mid-word, and a path past 160 bytes turned the has-no-line message into a bare no match. Both now carry up to 256 bytes, cut on a character boundary with an ellipsis past that. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* An error on a column's ctl, exec or rmdir names that file in the logGabriel Schneider2 days
| | | | | | | | noteError named every column node col/<serial>/tag, so a refused ctl write or rmdir pointed at the tag. It now names col/<serial>/ctl, /exec, /tag or the directory itself. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* Kill stops a command pane's whole command, its & jobs includedGabriel Schneider2 days
| | | | | | | | | | With job control on, a command's & jobs get process groups of their own, so Kill's signal to the foreground group and the shell's group left them running, orphaned. A stop of a command pane now signals every process group in its shell's session (Linux; a /proc scan). Jobs still outlive a command that exits on its own. Docs and the skill say it per route. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* $ may end any alternative in a pattern, not only the lastGabriel Schneider2 days
| | | | | | | | | | mvzr refuses $ anywhere but at the very end, so /h$|zz/ and Edit ,x/r$|pi/ were bad patterns while /zz|h$/ worked. Its matcher already ends an alternative at .end, so a build-time patch, anchor-checked like the group one, lets $ stand before | and ). The upstream report is in docs/mvzr-group-backtrack.md. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A look at file:/re/ on an open pane goes on from its dot, and a miss leaves itGabriel Schneider2 days
| | | | | | | | | Look's clean start cleared the active pane's selection before the address read the dot, so the root look's file:/re/ always found the same match, a miss collapsed the dot to a caret, and dumps kept the collapsed dot. A pane already open on the path is now addressed where it stands. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The look walk goes on from a restored dot both waysGabriel Schneider2 days
| | | | | | | | | | | After a Restore the dot came back on the match it was on, but N stayed on that match instead of wrapping to the last as it does live: the walk remembers the spot it landed on (look_at), and a restored pane has none, so it stood at the cursor, the selection's end, and found the match it was in. A one-row selection now counts as where the walk stands, from its start and past it either way, as a landed spot does. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* A pane renamed but never edited closes on the first DelGabriel Schneider2 days
| | | | | | | | | | | The GUI golden's wide-glyph pane, renamed and never edited, was refused by Del as modified, and every frame after it kept the pane. A rename marks a pane dirty, but its text is what it loaded: the check now asks whether the text was ever edited (its first edit keeps a hash of what it had), not only whether it is dirty, as get's refusal does. Tests that faked an edit by the dirty mark alone now fake the hash too. Co-Authored-By: Claude Opus 5.5 <[email protected]>
* The tree's README fits its 45 lines again, and fs-discovery-test knows the ↵Gabriel Schneider2 days
| | | | | | | | | | | | new entries /README had grown to 51 lines with the layout files and the round 8 notes, past the 45 the discovery test holds it to. It is condensed (the log's record names on one line, the column files on one, two examples fs.md covers), and fs-discovery-test lists layout, tag and col, reads /index's new last field, and reads /commands' setting values. Co-Authored-By: Claude Opus 5.5 <[email protected]>
generated by cgit v1.3 (git 2.53.0) at 2026-10-03 04:21:47 +0000