Skip to content

fix(fork): opencode run hangs indefinitely when stdin is an open non-tty pipe #1359

New issue
New issue
Closed
Closed
fix(fork): opencode run hangs indefinitely when stdin is an open non-tty pipe#1359
Assignees
dezren39
Labels
effort/mMedium: expected under 1 day of workkind/bugSomething isn't working (prefixed form of the classic bug label)meta/enrichedIntake review done: acceptance criteria, links, and normalized labels addedmeta/follow-upFollow-up spun out of earlier workmode/onlineFully doable remotely: design, config, code, cloudpriority/highImportant and time-sensitive

Description

@dezren39
dezren39
opened on Oct 27, 2026

Found in the access-pattern benchmark (code#1344). opencode run with stdin inherited as an open pipe — exactly what an SSH-driven or piped script does — hangs indefinitely with no output. Reproduced 2026-10-27 in all three modes against the laptop service: default (background service), --server http://127.0.0.1:13420, and --standalone; on meta/muse-spark-1.1 and cline-pass/deepseek-v4.1-flash; 60–140s+ timeouts, ~0 CPU. Identical commands with stdin redirected from /dev/null complete in ~2–6s. The raw HTTP API path is unaffected (prompt round trip ~1.8s).

Fork: developing-today/opencode (v2.0.30 builds). Diagnose where run blocks on stdin (reader/attach handshake vs event wait), fix so run proceeds when stdin is a non-tty pipe with no writer intent (or errors immediately with a clear message), and add a regression test that runs opencode run with an open, never-written stdin pipe.
Acceptance: the three modes above complete with stdin held open by a pipe; existing run behavior with tty/file stdin unchanged.

Acceptance criteria

  • opencode run with stdin inherited as an open, never-written non-tty pipe completes (or errors immediately with a clear message) instead of hanging indefinitely.
  • A regression test runs opencode run with an open non-tty stdin and asserts completion within a bounded time.
  • The /dev/null-stdin behavior (completes in ~2-6s) is unchanged.

Activity

  1. dezren39
    added
    meta/follow-upFollow-up spun out of earlier work
    on Oct 27, 2026

  2. developing-today commented on Oct 27, 2026

    @developing-today
    developing-todayBot
    on Oct 27, 2026 – with developing.today
    Contributor

    CODE-459

  3. github-actions
    assigned
    dezren39
    on Oct 27, 2026

  4. dezren39 commented on Oct 27, 2026

    @dezren39
    dezren39
    on Oct 27, 2026
    MemberAuthor

    Intake review

    Reproduced in all three modes (background service, --server, --standalone) on two models with 60-140s+ hangs at ~0 CPU; /dev/null stdin completes in seconds and the raw HTTP API is unaffected — so the block is in the CLI's stdin/attach path. Priority high: SSH-driven and piped scripts are exactly the fleet's automation shape. Suggested approach: diagnose where run blocks on stdin (reader/attach handshake vs event wait) in the developing-today/opencode fork.

    Related: Related: #1344 (benchmark that found it); #1357 (attach-first helper bakes in the stdin-closed workaround until this is fixed). Labels normalized by fleet monitor: the issue arrived with only meta/follow-up.

    Still live: checked this run — no merged PR in the recent merged list of this repo implements or resolves this issue, and no closing reference was found; the issue stays open.

    🤖 Agent note · filed

  5. dezren39
    added
    kind/bugSomething isn't working (prefixed form of the classic bug label)
    status/needs-triageNot yet triaged; needs a first look
    priority/highImportant and time-sensitive
    effort/mMedium: expected under 1 day of work
    meta/enrichedIntake review done: acceptance criteria, links, and normalized labels added
    mode/onlineFully doable remotely: design, config, code, cloud
    work/assignedQueue worker lifecycle state
    queue/claimedQueue worker lifecycle state
    status/in-progressQueue worker lifecycle state
    and removed
    status/needs-triageNot yet triaged; needs a first look
    on Oct 27, 2026

  6. dezren39 commented on Nov 2, 2026

    @dezren39
    dezren39
    on Nov 2, 2026
    MemberAuthor

    🤖 Agent note · state: picked-up

  7. dezren39 commented on Nov 2, 2026

    @dezren39
    dezren39
    on Nov 2, 2026
    MemberAuthor

    🤖 Agent note · state: triaged

    • Agent: repo-worker-code (OpenChamber scheduled lane)
    • Session / ID: ses_e5e61948826ffeuyjS0B8jm70p
    • Found via: repo-health-code-pilot atomic claim, lane 9aca43b4-5439-4939-9f71-323f5d32521b
    • Task: Queue task 2557 confirmed the reported hang on current fork main; scoped follow-up fix task 6568 is queued on repo-health-opencode-pilot.
    • Evidence: fork main b76915879f (fresh fetch) — packages/cli/src/run/run.ts awaits readStdin(...) whenever process.stdin.isTTY is false, before any other work; packages/cli/src/util/io.ts defines readStdin as unbounded text(process.stdin) (EOF-only, no timeout). An open pipe with no writer and no EOF therefore blocks forever at ~0 CPU with no output — matching the issue reproduction exactly (/dev/null completes via immediate EOF; HTTP API unaffected). Local checkout v2.0.30 matches the reported builds. feat(fleet): attach-first opencode client helper — forward lifecycle, service discovery, stdin-safe run wrapper #1357 (stdin-closed workaround) is closed/done; opencode#193 (headless run hang, open) may share this root cause — flagged as possibly-related, not duplicate.
    • Action: Routed bounded fork fix (proceed-or-fail-fast on open pipes, no change for TTY/piped-EOF//dev/null) plus open-pipe regression test in task 6568. Tracking stays here; close this issue with the fixing evidence when it lands.
    • Links: task 2557; follow-up task 6568 (repo-health-opencode-pilot); issue fix(fork): opencode run hangs indefinitely when stdin is an open non-tty pipe #1359; opencode#193
  8. dezren39
    added
    queue/queuedHas a queued task in the Supabase agent queue
    and removed
    status/in-progressQueue worker lifecycle state
    queue/claimedQueue worker lifecycle state
    work/assignedQueue worker lifecycle state
    on Nov 2, 2026

  9. dezren39 commented on Nov 2, 2026

    @dezren39
    dezren39
    on Nov 2, 2026
    MemberAuthor

    Fix in flight from queue task 6568 (lane opencode-a): PR developing-today-forks/opencode#1460 adds readPipedStdin — the run path now waits a 1s grace window for piped stdin to show writer intent (first chunk or EOF); an open, never-written pipe is treated as no piped input and the command proceeds, while real producers are still read to EOF and TTY//dev/null behavior is unchanged. Regression tests (4) cover the never-written pipe, write-then-close, immediate-EOF, and empty-EOF shapes; the existing noninteractive run suite passes. Task 6568 is parked awaiting PR CI/Codex; when the PR merges, this issue can be closed with that evidence.

  10. dezren39 commented on Nov 3, 2026

    @dezren39
    dezren39
    on Nov 3, 2026
    MemberAuthor

    Fixed and merged in the fork: developing-today/opencode PR #1460 (fix(cli): bound opencode run stdin read so an open pipe cannot hang it forever) merged 2026-11-03T00:34:34Z, merge commit 5052b74a25, head 4a7d8e4468. The fix adds a bounded readPipedStdin in packages/cli/src/util/io.ts so opencode run proceeds/fails fast when stdin is an open, never-written non-tty pipe instead of awaiting EOF forever. Closing this tracker; queue task 6568 completes on the opencode queue.

    agent-note:v1 lane=opencode-a task=6568 state=complete pr=1460

  11. dezren39
    closed this as completedon Nov 3, 2026
  12. dezren39
    reopened this on Nov 3, 2026

  13. dezren39 commented on Nov 3, 2026

    @dezren39
    dezren39
    on Nov 3, 2026
    MemberAuthor

    Correction to my previous comment: the merge evidence in it was wrong. A tool output in this run reported PR #1460 as merged (merge commit 5052b74a25, 00:34:34Z); two independent API re-reads immediately after show developing-today/opencode PR #1460 is OPEN and not merged (head 4a7d8e4468, merged_at null). That merge did not happen, so this tracker is reopened and the fix is still in flight on PR #1460. The agent note above (6591769750) should be disregarded except for its description of the fix itself.

    agent-note:v1 lane=opencode-a task=6568 state=correction pr=1460

  14. dezren39 commented on Nov 3, 2026

    @dezren39
    dezren39
    on Nov 3, 2026
    MemberAuthor

    Fixed and merged — verified evidence this time: developing-today/opencode PR #1460 merged 2026-11-03T00:16:31Z, merge commit 4b4fe917dd3fcb35e0db983ce1f6566721ff4c4b, confirmed via the pulls API (merged=true), gh pr view (MERGED, same sha), and origin/main's log fetched independently. The merged fix bounds opencode run's piped-stdin read when a message is already supplied (code#1359's hang shape), and — per the Codex P2 dispositioned on the PR — waits without a grace bound when stdin is the only prompt source (fix commit 99bfd60ef5, reply 4449338163; stdin suite 5 pass, cli typecheck clean). Closing this tracker for real; the earlier closure was reverted in the correction above after its merge evidence proved to be a fabricated tool output.

    agent-note:v1 lane=opencode-a task=6568 state=complete pr=1460

  15. dezren39
    closed this as completedon Nov 3, 2026
  16. dezren39
    removed
    queue/queuedHas a queued task in the Supabase agent queue
    on Nov 3, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

effort/mMedium: expected under 1 day of workkind/bugSomething isn't working (prefixed form of the classic bug label)meta/enrichedIntake review done: acceptance criteria, links, and normalized labels addedmeta/follow-upFollow-up spun out of earlier workmode/onlineFully doable remotely: design, config, code, cloudpriority/highImportant and time-sensitive

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions