| Commit message (Collapse) | Author | Age |
|---|
| ... | |
| | | |/
| | |
| | |
| | | |
whole column. Buggy paths fixed: Alt-n inserted a weight-1 shell (whole column reflowed) — now splitBelow the active pane; stackDocLeft (file/image/tutor open with no doc yet) gave the doc a fair share (every pane shrank) — now it halves only the old top pane's slot; ttyForDir (execute from a file pane) appended a weight-1 shell to the rightmost column — now splitBelow its bottom pane; Alt-c moved a pane out without absorbVWeight (all old siblings grew) — now the sibling above absorbs first. Underneath: splitBelow/absorbVWeight/stackDocLeft snap the whole column's weights to on-screen row counts, because computeGeom's per-pane round(avail*w/vsum) let fractional weights jiggle bystanders by a row; evictLonePristineTty recomputes geometry first (a just-stacked doc has no rect). New snap tests layout-open + layout-close cover Alt-n, doc stacking, ttyForDir, Alt-c, tag Del, pty EOF, and column-width handoff. Six goldens regenerated (fileedit, images, look-file, syntax, tutor, windowops): they had frozen the fair-share/append reflow this change removes; every diff is only the doc/split pane sizing, with untouched panes' boundaries restored bit-identical (e.g. bottom pane tag back at its boot row 17 instead of pushed to 22).
|
| | |/
| |
| |
| | |
update emitted a big selection exec (emitWrite 64-byte chunks + padOutputBelowEdits CRs) and the emit assert panicked. Effects queue is now a growable ArrayList FIFO (payloads stay inline values); guard the two other Buf(256) asserts reachable from look (spawn cwd >256 -> error.PathTooLong in newShell, >256-char url skipped). New snap exec-bigdrag: giant middle-drag exec on a 60x200 grid, panicked before, alive-marker golden after.
|
| |/
|
|
| |
vaxis.Tty.init makes the tty raw but run() never called tty.deinit (termios stayed raw -> no echo, staircased output in the spawner shell); the pty reader tasks were never joined (Kill-quit hung the exit in the runtime thread-join with readers blocked in read(2), and finished readers dumped DebugAllocator leak traces onto the tty). Fix: defer tty.deinit() (runs last, after vx.deinit flushes its resets); cancel each pty reader at teardown (cancel interrupts the blocked read) then close masters and drain the loop queue (queued pty_read/paste bytes are gpa-owned); await the reader on pty_eof before closing; readPty posts its eof via tryPostEvent so a post-cancel full queue can't re-block; root panic handler wraps vaxis.recover() so a panic restores cooked mode/main screen/mouse before the trace prints (vaxis.Panic itself is stale: references std.debug.FormattedPanic which 0.16 dropped). Verified with a scripted pty harness (bash spawner, DSR replies): both quit paths (exit-EOF of last pane, middle-click Kill) end with rmcup+cursor+sgr+mouse-off and identical stty -a before/after, no leak spew, prompt back in 0.5s; snap suite 18/18.
|
| |
|
|
| |
test/ (snapshot parity harness + 18 frozen goldens). One sans-IO core, vaxis tty + SDL3 GPU native + wasm web shells, 18/18 parity with the purged prototype, 7.6k lines vs 12.1k. Fix: gui shell pre-sized the core at init so the greet-releasing resize never fired (blank panes until first interaction); live sessions now init at defaults and get the real grid as a resize event (the shell contract, documented on Options).
|
| | |
|
| |
|
|
| |
4x larger, fade trails, SDL GPU + WebGL overlay shaders, touch-action:none in HTML, --debug-overlay flag, pinch event dispatch for all SDL backends
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
