[tom-rigelblu]

Takes upstream ghostty-org's fix, 297507f ("termio: send the config
change color scheme report directly", Fixes ghostty-org#14833, 2026-10-24), with
its exact text and position, so the sync that brings it keeps one copy.
manaflow-ai/ghostty main doesn't carry it yet (checked 2026-11-09).

The IO thread is the only thread that drains a surface's termio mailbox.
While it handled .change_config, StreamHandler.changeConfig queued a
mode 2031 color_scheme_report into that same mailbox with a blocking
send. When the mailbox was full, the IO thread waited on itself forever,
and the main thread's next push for that surface waited behind it: a
light/dark switch froze the app. A cmux hang sample from 2026-11-05 shows
exactly that: main and one of 79 io threads both in the queue-full wait
of the same BlockingQueue push, the other 78 idle.

• src/termio/stream_handler.zig: changeConfig no longer queues the report
  (upstream's text).
• src/termio/Termio.zig: changeConfig writes it last, with
  colorSchemeReportLocked(td, false), inside the renderer lock it already
  holds (upstream's text). A failed write is logged and the config change
  still completes. Before, a failed queued report ended the IO thread's
  drain pass (the try in Thread.zig), so later messages waited for the
  next wakeup.
• src/global.zig: tmpDirPath returns "/tmp" in tests with no global
  state, like the file's other builtin.is_test fallbacks, so a test can
  build a real Termio. Ours, like both tests; upstream's commit has none.

Tests, each seen failing first:

• "changeConfig returns on a full termio mailbox": red on ed5c48f,
  "changeConfig was still waiting after 2 s on a full termio mailbox".
• "changeConfig writes one mode 2031 report with the new theme and queues
  none": the terminal starts dark and switches to light. With the
  queued send put back it finds nothing; with the report written before
  the new config is swapped in it finds the old theme, "\x1b[?997;1n",
  where it expects "\x1b[?997;2n".
  Mutations: dropping the new call turns the second test red alone;
  putting the queued send back turns both red.

🤖 Generated with Claude Code

https://claude.ai/code/session_019faQDfp9Q3KPyaDHXEosfG

[changeset-bot]

⚠️ No Changeset found

Latest commit: 1cb1aaa

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR