Allow editor/terminal dictation to take over an active Chat dictation session

[Copilot]

Description

Chat, editor, and terminal dictation share a single on-device speech engine, so only one can run at a time. While Chat dictation was active, "Start Dictation in Editor" and "Start Dictation in Terminal" silently did nothing: the editor path early-returned from startDictation (the shared _active session was owned by Chat), and the terminal path aborted when the engine was busy.

The fix lets a new surface take over: each surface already tears down its own UI when it observes the engine go Idle, so a new start just cancels the current session before attaching its own listeners.

• dictationSession.ts (startDictation) — instead of no-op returning when a session is active, cancel the current session (Chat, another editor, or terminal) and take over; the previous surface cleans up via its own Idle listener and keeps whatever transcript it had inserted. Same-editor re-entry still returns early (toggle handled by callers).
• editorDictation.ts (startBuiltin) — register the "self-teardown on Idle" listener after startDictation so the takeover's cancel of the previous surface can't tear down the just-started editor session.
• terminalVoice.ts (_startBuiltin) — replace the "abort when busy" guard with a service.cancel() takeover (before subscribing), keeping a defensive bail-out if the engine somehow stays busy.
• dictationSession.test.ts — added coverage for cross-surface takeover (first editor keeps its text, second becomes active); updated the mock cancel() to emit the real Idle transition.

Takeover is now symmetric across the reported directions (Chat → editor, Chat → terminal) as well as editor ↔ terminal, with no cross-contribution imports.

• Fixes Can't start in editor or terminal once Chat dictation has started #328809

[vs-code-engineering]

📬 CODENOTIFY

The following users are being notified based on files changed in this PR:

Anthony Kim (@anthonykim1)

Matched files:

• src/vs/workbench/contrib/terminalContrib/voice/browser/terminalVoice.ts
• src/vs/workbench/contrib/terminalContrib/voice/test/browser/terminalVoice.test.ts

[Copilot]

Pull request overview

Enables editor and terminal dictation to take over the shared speech engine from another surface.

Changes:

• Cancels active dictation before starting on another surface.
• Adjusts editor listener ordering to avoid premature teardown.
• Adds editor-to-editor takeover coverage.

Show a summary per file

File	Description
terminalVoice.ts	Adds terminal takeover behavior.
editorDictation.ts	Delays editor idle-listener registration.
dictationSession.ts	Adds shared-session takeover logic.
dictationSession.test.ts	Tests editor takeover and realistic cancellation.

Review details

• Files reviewed: 4/4 changed files
• Comments generated: 2
• Review effort level: Balanced

[Megan Rogge (meganrogge)]

@copilot resolve the merge conflicts in this pull request

[Copilot]

Copilot resolve the merge conflicts in this pull request

Merge conflicts resolved in c4b13ce. Conflicting hunks in the dictation files were merged (combined DictationSettingId + isDictationActiveOnSurface import in newChatInput.ts, kept the surface-aware test helpers in dictationSession.test.ts); all other conflicted files were taken from main. Also updated terminalVoice.test.ts's mock cancel() to be async to match the new Promise<void> signature. Remaining typecheck-client errors are pre-existing on main (agentHost).