Since #34 (imgui 1.92.6 bump, merged 2026-06-26) every macOS integration run fails with the worker-startup signature — the spawned daslang-live instance never opens its live-API port, and the playwright side spams:
ERROR connect 127.0.0.1:909X failed! [HttpClient.cpp:217:http_client_connect]
Evidence it's the bump and not the runner flake (the dasImgui#208 slow-startup family): master run 28236488085 (the #34 merge itself) fails with this signature, and every macOS run since — including two attempts on the CMake-only #35 — reproduces it deterministically, while ubuntu integration stays green on the same commits.
Likely the worker app dies at startup on darwin under 1.92 (texture-model or vendored node-editor patch interplay) before the HTTP server binds; needs a mac (or lane-side capture of the worker's own stderr, which would also make the next one of these a one-log diagnosis) to root-cause.
Found while landing #35 (both-worlds CMakeLists), which merges over this documented pre-existing red.
Since #34 (imgui 1.92.6 bump, merged 2026-06-26) every macOS integration run fails with the worker-startup signature — the spawned daslang-live instance never opens its live-API port, and the playwright side spams:
Evidence it's the bump and not the runner flake (the dasImgui#208 slow-startup family): master run 28236488085 (the #34 merge itself) fails with this signature, and every macOS run since — including two attempts on the CMake-only #35 — reproduces it deterministically, while ubuntu integration stays green on the same commits.
Likely the worker app dies at startup on darwin under 1.92 (texture-model or vendored node-editor patch interplay) before the HTTP server binds; needs a mac (or lane-side capture of the worker's own stderr, which would also make the next one of these a one-log diagnosis) to root-cause.
Found while landing #35 (both-worlds CMakeLists), which merges over this documented pre-existing red.