The clear-only metadata test used a 1 ms ttl, so on loaded Windows runners set_agent_metadata could see the entry as expired, drop it, and recreate it with a None ttl before next_agent_metadata_expiry() ran. Expiry is forced explicitly later in the test, so use a long ttl to keep the no-extension assertion independent of wall-clock scheduling.
Live handoff refused any session with more than 64 panes. The pane count
was checked twice against MAX_FDS_PER_HANDOFF, and the transfer itself put
every pane's pty master into one SCM_RIGHTS control message, so the guard
was the only thing keeping the send inside the kernel's per-message limit.
A session past the limit could only be updated by closing panes or by a
normal restart, which ends every pane process.
Send the descriptors in batches of 64 instead and drop both guards. The
receiving side accumulates across recvmsg calls until the expected count
arrives, bounds every SCM_RIGHTS payload it reads by the control bytes the
kernel returned, rejects a batch that carries more descriptors than it
asked for, and closes the descriptors it already holds on any failure. A
session of 64 panes or fewer still produces one batch, so the bytes on the
wire are unchanged and HANDOFF_VERSION stays at 1.
refs #3393
Co-authored-by: kataokatsuki <kataokatsuki@users.noreply.github.com>
Co-authored-by: JJ Liebig <jonathan.liebig@gmail.com>
Windows CI runs these documentation contract fixtures under variable load. A degraded runner pushed the preview and versions snapshot integration tests past their 30s ceiling, while a healthy run finishes in seconds. Keep 30s elsewhere and give Windows a 120s ceiling.
The polling peer drains at production's 2ms Windows read cadence, so a 1 MiB
frame took ~5.5s locally and exceeded the test's 10s flush deadline under CI
load. Use a 256 KiB frame, still well above the 64 KiB batch limit, and a 30s
deadline so the assertion no longer depends on CI scheduling.
* fix: support opencode v2 lifecycle reporting
* fix: repair Japanese docs
* fix: ignore payload-less OpenCode events
* fix: harden opencode v2 integration install and reporting
Follow-ups on top of the V2 lifecycle support:
- create `cli.json` when OpenCode has no V1 TUI preferences (`tui.json` or
`kv.json`) to migrate, instead of only registering into an existing file
- settle stalled socket attempts with a plain connect timer, and resend the
latest lifecycle state after a failed delivery so the pane cannot get stuck
- resolve the OpenCode state directory from `XDG_STATE_HOME`
- remove the managed `herdr-opencode` directory on uninstall
- document failed executions reporting `blocked`, keep the integration test
environment independent of an inherited `XDG_STATE_HOME`, and simplify
`reconcileBlockers` to reassign its map
---------
Co-authored-by: Jonathan Liebig <jonathan.liebig@gmail.com>
* ci: retain zig build caches across validation runs
* ci: reduce debug information in validation builds
* ci: build only the ghostty test target for scalar checks
* ci: measure debug profiles and remove duplicate macos caching
* ci: compare scalar build scopes on one runner
* ci: keep the zig compiler path stable across runs
* ci: run macos and windows lint alongside tests
* ci: inspect remaining zig cache misses
* ci: remove completed measurement probes
* ci: compare windows linkers without changing release builds
* ci: test linux incremental reuse and correct the linker probe
* ci: measure windows incremental state reuse
* test: verify incremental reuse after a source change
* ci: finalize measured compiler cache optimizations
Native Windows lint and check only clippy-checked the herdr binary, while macOS and Linux run clippy across all targets. Enable --all-targets on Windows and gate the Unix-only test helpers that this exposed so Windows check matches the Unix one.
Codex's Windows input reader does not surface bracketed paste. It buffers
the prompt as a paste burst and rewrites a following Enter into a newline
until the burst idles out, so the previous size-based delay only reduced
the chance of losing the submit. Append a non-character key after the
paste so Codex flushes the burst synchronously, giving every prompt a
deterministic submit boundary.
refs #3187
JJ Liebig