Fix terminal PTY interruptions during remounts#138
Merged
Conversation
This was referenced May 25, 2026
Owner
|
Thank you! |
brooksc
added a commit
to brooksc/parallel-code
that referenced
this pull request
May 27, 2026
Users can now type into any coordinator or sub-task session without first taking control. Coordinator and system prompts queue and deliver automatically when the target session is safe: agent idle, no user draft, no pending terminal input, no recent user activity. Backend (electron/mcp): - Backend-owned initial prompt delivery with scheduling/retry so sub-tasks no longer depend on a mounted renderer panel. - Improved terminal prompt detection: Working(...), background terminal running, q-corrupted variants, trust dialogs. - New 'landed' PendingNotification state so successful land_self queues a "self-landed successfully" notification. - Suppress coordinator notifications while children are still active. Renderer: - Per-tick autofire arbiter with explicit wait reasons (paused, waiting-for-user-draft, waiting-for-terminal-input, waiting-for-user-activity, no-prompt, fire). - Prompt draft and terminal pending-input tracking so queued automation can't overwrite half-typed user input. - Per-task user activity leases — typing pauses automation without taking durable ownership. - Removed Take/Release control UI from TaskPanel. Tooling: - scripts/check-coordinator-run.mjs — log checker for regression triage from captured app logs. - scripts/fake-agent.mjs — PTY mock for integration tests. - Real-PTY and real-CLI integration suites for Codex/Claude/Gemini /Copilot. Dead-code cleanup: - Removed src/lib/terminalDisableStdin.ts; both helpers had become constant-wrapping shims after the take/release removal. OpenSpec: new capability spec session-prompt-delivery. Stacked on johannesjo#138 (terminal PTY remount lifecycle) and johannesjo#145 (sub-task self-landing). Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
johannesjo
pushed a commit
that referenced
this pull request
May 28, 2026
Users can now type into any coordinator or sub-task session without first taking control. Coordinator and system prompts queue and deliver automatically when the target session is safe: agent idle, no user draft, no pending terminal input, no recent user activity. Backend (electron/mcp): - Backend-owned initial prompt delivery with scheduling/retry so sub-tasks no longer depend on a mounted renderer panel. - Improved terminal prompt detection: Working(...), background terminal running, q-corrupted variants, trust dialogs. - New 'landed' PendingNotification state so successful land_self queues a "self-landed successfully" notification. - Suppress coordinator notifications while children are still active. Renderer: - Per-tick autofire arbiter with explicit wait reasons (paused, waiting-for-user-draft, waiting-for-terminal-input, waiting-for-user-activity, no-prompt, fire). - Prompt draft and terminal pending-input tracking so queued automation can't overwrite half-typed user input. - Per-task user activity leases — typing pauses automation without taking durable ownership. - Removed Take/Release control UI from TaskPanel. Tooling: - scripts/check-coordinator-run.mjs — log checker for regression triage from captured app logs. - scripts/fake-agent.mjs — PTY mock for integration tests. - Real-PTY and real-CLI integration suites for Codex/Claude/Gemini /Copilot. Dead-code cleanup: - Removed src/lib/terminalDisableStdin.ts; both helpers had become constant-wrapping shims after the take/release removal. OpenSpec: new capability spec session-prompt-delivery. Stacked on #138 (terminal PTY remount lifecycle) and #145 (sub-task self-landing). Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Task AI terminals currently terminate their backend PTY when the TerminalView component cleans up. In dev/HMR or other renderer remounts, that kills an active Codex session, then the renderer respawns it with resume args. Codex sees that as an interrupted conversation and waits for manual continuation.
Fix
Task AI terminals now preserve their PTY across component cleanup and reattach to the existing process on remount. Intentional replacement remains explicit: restart and agent switch set attachExisting=false, so those actions still replace the running process. Standalone close/cleanup paths continue to own termination.
Verification