an opinionated terminal multiplexer in C with programmable pane chrome https://git.sl0p.foo/slosh.git/atom/vendor/libghostty-vt.vendor.json?h=master 2026-09-12T14:05:54Z 2026-09-12T14:05:54Z user user@clank 2026-09-12T14:05:54Z urn:sha1:7617cf20d908386ab3052d53e0c909b4159b7280 The browser demo came up with `slosh: cannot start session main` and nothing else. The session server was segfaulting before its socket existed, on the first calloc() of the config parse -- because in that binary calloc() returned NULL, always. pkg/wuffs/src/main.zig exports weak `calloc`/`free` stubs whenever `!builtin.link_libc`, and callocStub returns null unconditionally. They are there because wuffs' generated `..._alloc()` helpers are the only code referencing libc's allocator, we never call them, and linker GC usually strips them; the stubs cover the case where it does not. The comment reasons that weak linkage means "any real definition wins". That holds for a shared libc and not for a static archive. Once the linker pulls this member in for another symbol, the weak calloc here *satisfies* the reference, so it never searches libc.a for musl's strong one -- and every allocation in the consuming program returns NULL while free becomes a no-op. Only the demo could hit it. It is the one build that must pass -Dsimd=false (highway emits RVV intrinsics that do not compile for rv64gc), GhosttyZig.zig sets `link_libc = if (cfg.simd) true else null`, and the guest links the result statically against musl. The host build links libc, never exported the stubs, and passed everything -- which is why the whole test suite was green over a demo that could not start a session. Patch 0002 adds `and builtin.output_mode == .Exe`. The stubs still do their job in a final link with no libc; a library stops interposing on the allocator of the binary that owns that choice. Worth sending upstream: the assumption is upstream's, not ours. 2026-09-02T13:45:32Z user user@clank 2026-09-02T13:45:32Z urn:sha1:311979f10132697b990a670c6c99b19c1c33d623 Three weeks of upstream: 504 commits, 120 touching the terminal core -- OSC 5522, kitty relative placements, C0 fixes, the wuffs pixel path. Rebuilt from ghostty's own dist step (zig build dist -Demit-lib-vt=true), zig floor unchanged at 0.16.0, and slosh's C usage compiled unmodified. Patch 0001 is re-implemented, not retired: upstream rewrote the area into ImageStorage.clearScreen but kept the behaviour we patch out -- a terminal clear still reclaims transmitted-but-unplaced images, which breaks the upload-once place-per-frame pattern the protocol's two verbs exist for. The patch now removes just that sweep; the placement clearing and orphan reaping stay upstream's. Their recent 'preserve data on unmatched delete' fixes the same class of bug, so sending ours upstream remains the way to actually retire it. 2026-08-14T21:13:56Z user user@clank 2026-08-14T21:13:56Z urn:sha1:d1442bc9550b00080c2f5671fa93eaad67fb0299 libghostty-vt's eraseDisplay called kitty_images.delete(.{ .all = true }) for ED 2 and ED 3, where the `true` also frees the image *data* of anything no longer placed. A full-screen program that clears and redraws therefore loses every image it uploaded, and every later placement of that id draws nothing. That breaks the half of the protocol worth having. Kitty separates transmitting an image (`a=t`, stored under an id) from placing it (`a=p`, put on screen) precisely so a program can upload once and place many times -- which is also why the delete command distinguishes `d=a` (placements) from `d=A` (placements and data). A screen clear should be the former. The patch removes only the freeing: placements on the cleared screen still go, and the image store still bounds itself with its 320MB limit and LRU eviction on insert -- the mechanism that makes keeping them safe, and that would be pointless if a clear dropped everything anyway. Found with a bouncing-DVD screensaver that uploads ten frames up front and places one per frame. It works in kitty; under us it drew nothing. D12 said "with a patches file if we ever need one". vendor/patches is that file, with the reasoning and what a re-vendor has to re-apply; the vendor json lists it. Worth sending upstream. 2026-08-13T20:00:03Z sl0p sl0p@sl0p.foo 2026-08-13T20:00:03Z urn:sha1:2c1b855c3cf8cf8841bf25cd939e478a7f0bfac8 Design contract in DESIGN.md. libghostty-vt vendored fresh from ghostty main@bedb6f2 (zig 0.16), built to a 14MB static lib. M0: pty spawn (no libutil), a lib-vt terminal per pane, a screen cell buffer, a diff emitter that repaints only what changed, raw mode, SIGWINCH, and a ~120Hz frame cap that coalesces pty reads. Also the seed of the headless harness: --headless dumps the composited screen without a tty, so composition is testable in milliseconds. tests/live_m0.py covers what headless cannot: raw mode, the diff emitter, resize propagation, and the quit path. 6/6 green.