summaryrefslogtreecommitdiff
path: root/anim/test/asan.script
Commit message (Collapse)AuthorAge
* anim: scripts drive the player's deckGabriel Schneider4 days
| | | | | | | | | | | sketchbook's stock player now keeps its playhead in an anim.Deck, so the scripts set deck.t and deck.playing where they set t and playing. anim.Frame has no clock field any more: the two cursor tests stop setting it. (A paused playhead set past the end holds the end, as it did: the player that looped it back to 0 was fixed in sketchbook.) Verified zig build test in anim: 203/203 steps; scene durations and play counts unchanged against the old library's screenshots.
* canim: relations as one data-oriented subsystem; aim; the ASan pair; the ↵Gabriel Schneider10 days
introduction's chapters From the user. Relations between panes are one subsystem: per-frame tables of places and values in struct-of-arrays, owned by the frame; the kinds one table of data; a resolver of pure passes (join, filter, priority, cut) unit-tested on hand-written tables. Ported kinds draw byte-identical frames; a synthetic map of 4000 places and 6000 values resolves in 0.56 ms against the old table's 5.7. aim, a word holding a code address tied to its row, is the first new kind, off unless a scene turns it on. The ASan pair: one overflow built plain and with ASan, recorded by instruction; the frame as bytes eight a row; the plain run goes on alone in the column, its return address aimed at main's row until the letters reach it; the fault at ret. The introduction's three chapters are built and tested, not yet exported. Annotations, rows and scrolls that changed in one frame now fade; captions stand twelve seconds at most. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_016Y9MNCKgYCGiANnEwcEQxa