Continue hangs after background or detached child-process tool calls

Continue hangs after background or detached child-process tool calls

Before submitting your bug report

• I checked for duplicate local reproduction drafts.
• I reduced the reproducer to deterministic local-provider cases.

Relevant environment info

• OS: Linux x86_64 Docker
• Continue version: @continuedev/cli 1.5.45
• IDE version: CLI (cn), no IDE
• Model: OpenAI-compatible local mock provider, gpt-4

Description

Severity / performance impact: High. When Continue receives a tool call that starts a background or detached child process, the CLI does not complete the session cleanly. The public reproducer observes that the tool request reaches the client, the child-spawning command writes its PID side effect, and the CLI then fails to finish before the timeout.

Modelock repeat-3 replay also reproduced the stronger process-lifecycle oracle for both representative variants: no_hang fired three times and no_orphan_process fired three times for the background-shell sample, and the same pair fired three times for the detached-child sample. Those replay artifacts are in findings/open/2026-06-16-continue-child-process-cleanup/.

To reproduce

Prerequisites: Docker, Python 3, and the GitHub CLI (gh) for the clone command below. The linked reproducer is self-contained and uses only Python standard-library modules plus Docker. It builds the affected CLI version from the public package registry, starts a local mock provider, and runs the CLI in an isolated workspace. The Docker run is limited to 2 CPUs and 4 GiB RAM by default.

Complete self-contained reproducer: https://gist.github.com/trackoor/58792d4d031da6f42e9ce770ddf35de6

The Gist contains continue-child-process-cleanup.reproduce.py. Download and run:

gh gist clone 58792d4d031da6f42e9ce770ddf35de6 continue-child-process-cleanup-reproducer
cd continue-child-process-cleanup-reproducer
python3 continue-child-process-cleanup.reproduce.py

To reuse an already-built local image:

python3 continue-child-process-cleanup.reproduce.py --skip-build

Key output from a local run against the affected version:

provider_requests_at_least: observed=1 expected>=1
side_effect_file_exists background-shell.pid: bytes=... expected>=1
timeout_or_signal: returncode=... timed_out=True
REPRODUCED

The embedded case runs both background-shell and detached-child variants.

Log output

Expected successful reproduction output includes REPRODUCED.

Observed behavior: Continue accepts a child-spawning tool call from the local provider, the command records a PID side effect, and the CLI does not return normally within the bounded run.

Expected behavior

Continue should track and clean up tool-started background children, then finish or cancel the session without hanging or leaving unreaped process state.

Public-upstream status

This reproduces against stock public @continuedev/cli@1.5.45 built from npm.

Start with continue-child-process-cleanup.reproduce.py from the linked Gist and run it against @continuedev/cli@1.5.45, including the background-shell and detached-child cases. Trace the CLI's tool-call and child-process cleanup paths from the reproducer's observed timeout. Done means both cases finish or cancel without hanging, orphaning processes, or leaving unreaped process state.