Repository navigation
INT-03 S5: native wasi:filesystem (open/read/write/close) — unblocks INT-06 #485
Description
Activity
- added a commit that references this issue
on May 31, 2026 Preview1 S5 lowering is on the branch:
file_open/file_read/file_fd_write/file_close→path_open/fd_read/fd_write/fd_close, EffectFile, dirfd 3, combo canonical-order lock, gatedtests/componentize/filesystem_smoke.sh(wasmtime run --dir host::.).Landed on
wasi-485-closeout(main...feat/wasi-485-closeout). Index #175 closed.PR opened: #757 (
feat/wasi-485-closeout→main).Landed in #757 (merged to
main,ff7ccd7). Preview1file_open/file_read/file_fd_write/file_close→path_open/fd_read/fd_write/fd_close, EffectFile, dirfd 3, combo lock, gatedtests/componentize/filesystem_smoke.sh.Closing this follow-up. Native-preview2 filesystem is #486 remainder, not this ticket.
ADR-015 canonical S5 (per
.machine_readable/6a2/META.a2ml): bring upwasi:filesystemprimitives —path_open/fd_read/fd_write(non-stdout) /fd_close. INT-03 substrate (#180) is otherwise shipped; this slice is what unblocks INT-06 (server-side runtime profile, ledger-only today, blocked-by INT-03).Pattern is settled from S4a/S4b/S6b: declare imports in
lib/wasi_runtime.ml, on-demand canonical-order indexing viaoptional_wasi+ctx.wasi_func_indices, Effect_sites pre-scan, surface builtins on theFilecapability withEffect File(reserved). Component path bridges towasi:filesystem/typesvia the preview1→preview2 command/reactor adapter (already provisioned by S2 / S3 / S6a). Path encoding goes through the byte-IR (I32Load8U/I32Store8, shipped #339) the same wayenv_at/arg_atdo.Gates: hermetic test under
tests/componentize/that round-trips a small file throughwasmtime runwith a--dirpreopen; combo regression that filesystem + clock + env + sock_shutdown imports stay canonically ordered.Ledger: INT-03 (S1, substrate) follow-up. Refs #180.