smolvm v1.12.0
What's Changed
- Add an opt-in virtio-gpu scanout with a host-side VNC server and virtio-input so machines can run an interactive Linux desktop by @BinSquare in #1027
- Ship virtio-input-capable guest kernels and an input-enabled macOS libkrun by @BinSquare in #1043
- Harden forked workloads and add transactional embedded batch forking by @BinSquare in #1044
- Make large pack pushes reliable with upload-scoped timeouts and a chunked fallback by @BinSquare in #1045
- Rewrite a loopback proxy host to the host's outbound address so guests can reach it by @BinSquare in #1046
- Cut v1.12.0 with the display-capable libkrun and rebuilt platform libraries by @BinSquare in #1048
- Harden live fork execution for rollout sandboxes by @BinSquare in #1049
- Pack a machine created from a local image archive or rootfs directory by @BinSquare in #1047
- Pin the Vulkan driver on exec only when the container actually has the bundle by @BinSquare in #1052
- Bump the workspace crates to 1.12.0 so the library crates ship at the engine's version by @BinSquare in #1051
- Mount the bundled Vulkan driver into pod and overlay containers too by @BinSquare in #1053
- Boot a pod sandbox with a virtio-gpu device when the pod asks for one by @BinSquare in #1054
Full Changelog: v1.11.1...v1.12.0