https://gitlab.synchro.net/main/sbbs/-/commit/cf1981eeec2eb92ddfe3172e
Modified Files:
src/syncterm/scripts/load/wrentest.wren src/syncterm/scripts/syncterm.wren src/syncterm/wren_host.c wren_host_internal.h
Log Message:
SyncTERM: route Wren hook fires through Hook.dispatch_

Every hook fire (onKey/onInput/onOutput/onMouse/onStatus/every) now
goes through Hook.dispatch_(fn[, arg]) instead of fn.call() directly.
The wrapper runs the handler in a child fiber via Fiber.try; if the
handler yields directly (e.g. naively calls Input.nextEvent or a
parking SFTP op at its top level), the wrapper detects f.isDone ==
false, logs a clear error naming the right escape hatch, and returns
null so the dispatcher's bool/string slot read falls through to passthrough/default. Errors are caught the same way and pushed
through REPL.printTrace_ for visibility.

Hook drops its (misleading) `foreign` modifier — it has no instance
state and no allocator was ever registered. Plain `class` lets the
new dispatch_ / finishDispatch_ methods sit alongside the foreign
statics, matching the Color / ScreenSupports pattern.

C dispatchers cache state.hook_class + dispatch0/dispatch1 method
handles after the syncterm module loads, then push slot 0 = Hook,
slot 1 = fn, slot 2+ = arg(s). build_match_list now takes a
list_slot parameter so the regex match list can land at slot 2.

Six tests in wrentest.wren cover: passthrough (1-arg, 0-arg, bool
true/false), direct-yield rejection (1-arg, 0-arg), child-fiber
yield permitted (the supported `Fiber.new { ... }.call()` pattern
for parking work from inside a hook), and Fiber.abort caught.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---
￭ Synchronet ￭ Vertrauen ￭ Home of Synchronet ￭ [vert/cvs/bbs].synchro.net