Repository navigation
Fix ghost terminal surface rebind after close - #808
Merged
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
📝 WalkthroughWalkthroughAdds debug instrumentation and guarded portal-binding lifecycle checks across terminal surface/view, portal registry, panel/tab/workspace split flows, and exposes portal stats via v2; also adds a regression test verifying no portal-orphan rebinds after Ctrl‑D closes. Changes
Sequence Diagram(s)sequenceDiagram
participant TabMgr as TabManager
participant Panel as TerminalPanel / Window
participant PortalReg as TerminalWindowPortalRegistry
participant Surface as TerminalSurface / GhosttyView
participant Test as Test harness / Client
TabMgr->>Panel: request create split (direction)
Panel->>PortalReg: ensure surface/host wiring for new panel
Note over PortalReg,Surface: Binding attempt may include expectedSurfaceId & expectedGeneration
PortalReg->>Surface: bind(hostedView, anchorView, ..., expectedSurfaceId, expectedGeneration)
alt expected matches and surface live
PortalReg-->>Surface: bind success (map entry)
Surface-->>PortalReg: debugStats / generation confirmed
else mismatch or closing/closed
PortalReg--xSurface: bind blocked (increment blockedBindCount, log reason)
PortalReg->>PortalReg: cleanup stale mapping if needed
end
Test->>PortalReg: query debug.portal.stats()
PortalReg-->>Test: aggregated portal + blocked-bind stats
Note right of Surface: close lifecycle: beginPortalCloseLifecycle -> markPortalLifecycleClosed -> generation++
PortalReg->>Surface: further bind attempts with old generation will be blocked
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
Comment |
Contributor
Greptile SummaryFixes the ghost terminal bug where Ctrl-D closed terminals could briefly reappear in splits due to stale portal bind paths reattaching surfaces during close. Key Changes:
Validation: Confidence Score: 5/5
Important Files Changed
Flowchart%%{init: {'theme': 'neutral'}}%%
flowchart TD
A[TerminalSurface Created] -->|generation=1| B[State: live]
B -->|Panel.close called| C[beginPortalCloseLifecycle]
C -->|generation=2| D[State: closing]
D -->|Surface deinit| E[markPortalLifecycleClosed]
E -->|generation=3| F[State: closed]
G[Portal Bind Request] --> H{canAcceptPortalBinding?}
H -->|state != live| I[REJECT - Block stale bind]
H -->|expectedSurfaceId mismatch| I
H -->|expectedGeneration mismatch| I
H -->|All checks pass| J[ACCEPT - Bind surface to window portal]
D -.->|Blocks any new binds| I
F -.->|Blocks any new binds| I
style B fill:#90EE90
style D fill:#FFD700
style F fill:#FF6B6B
style I fill:#FF6B6B
style J fill:#90EE90
Last reviewed commit: 56b5df0 |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
Sources/TabManager.swift (1)
1869-1876: Deduplicate split direction label mapping.The same
SplitDirection→ string switch is repeated in three places (Lines 1869-1876, 1901-1908, 2038-2044). Extracting a small helper/computed property (e.g.,direction.debugLabel) would reduce drift risk in future log updates.Also applies to: 1901-1908, 2038-2044
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@Sources/TabManager.swift` around lines 1869 - 1876, The repeated switch-to-string mapping for SplitDirection should be extracted into a single computed property on the SplitDirection type (e.g., add var debugLabel: String { switch self { case .left: return "left" ... } }) and replace the three inline switches that assign directionLabel with a call to direction.debugLabel; update all usages where a debug label is needed (the places that currently switch on direction) to use this new property to avoid duplication and drift.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py`:
- Around line 131-142: The _portal_integrity_error function currently coerces
missing totals keys to 0 which can false-pass; instead, first assert that totals
contains the required keys ("orphan_terminal_subview_count",
"visible_orphan_terminal_subview_count", "stale_entry_count") and if any are
missing return a non-nil error stating which keys are absent, otherwise parse
the integer values and perform the existing orphan/stale checks (do not silently
default missing keys to 0); update the logic in _portal_integrity_error to
detect malformed stats payloads and fail closed by returning an error when
required counters are not present.
- Around line 181-214: The rebind detector incorrectly suppresses visible events
while deinit is in-progress because the RE_VISIBLE_ON handling checks "sid in
close_pending and sid not in deinit_started"; remove the deinit_started
exclusion so any visible event for a sid in close_pending is treated as a
violation (the RE_DEINIT_END handler already discards sids from close_pending on
completion). Update the RE_VISIBLE_ON branch in _find_close_rebind_violations to
append the line when sid is in close_pending (remove the "and sid not in
deinit_started" condition), keeping the existing RE_DEINIT_BEGIN/RE_DEINIT_END
logic and sets (deinit_started, close_pending) intact.
---
Nitpick comments:
In `@Sources/TabManager.swift`:
- Around line 1869-1876: The repeated switch-to-string mapping for
SplitDirection should be extracted into a single computed property on the
SplitDirection type (e.g., add var debugLabel: String { switch self { case
.left: return "left" ... } }) and replace the three inline switches that assign
directionLabel with a call to direction.debugLabel; update all usages where a
debug label is needed (the places that currently switch on direction) to use
this new property to avoid duplication and drift.
ℹ️ Review info
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (8)
Sources/AppDelegate.swiftSources/GhosttyTerminalView.swiftSources/Panels/TerminalPanel.swiftSources/TabManager.swiftSources/TerminalController.swiftSources/TerminalWindowPortal.swiftSources/Workspace.swifttests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
1 issue found across 8 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py">
<violation number="1" location="tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py:211">
P1: The regression detector stops flagging close→visible rebinds after `deinit.begin`, which can miss visibility rebinds that occur before `deinit.end` (still before deinit completion).</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
Contributor
Author
|
Addressed the review feedback in commit 7a0fb42:
@cubic-dev-ai please re-review. |
@lawrencecchen I have started the AI code review. It will take a few minutes to complete. |
|
✅ Actions performedReview triggered.
|
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py`:
- Around line 124-125: The RPC call in _portal_stats currently uses
c._call("debug.portal.stats") with the default 20s timeout and should be bounded
by the integrity loop deadline; update _portal_stats to accept a timeout
argument (e.g., timeout_s) and call c._call("debug.portal.stats",
timeout_s=timeout_s) (or similar param name used by cmux._call), then modify
callers such as _wait_for_portal_integrity(timeout_s=...) to compute the
remaining deadline and pass that remaining timeout into _portal_stats so each
RPC is never allowed to block longer than the integrity loop budget.
ℹ️ Review info
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
Sources/TabManager.swifttests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py
🚧 Files skipped from review as they are similar to previous changes (1)
- Sources/TabManager.swift
There was a problem hiding this comment.
♻️ Duplicate comments (1)
tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py (1)
124-125:⚠️ Potential issue | 🟠 MajorBound
debug.portal.statsRPC timeout to the integrity-loop deadline.Line 125 still uses
c._call("debug.portal.stats")with the default 20s timeout (seetests_v2/cmux.py), so_wait_for_portal_integrity(timeout_s=...)can overrun its budget and hang/flap under RPC latency.💡 Suggested fix
-def _portal_stats(c: cmux) -> dict: - stats = c._call("debug.portal.stats") or {} +def _portal_stats(c: cmux, *, timeout_s: float) -> dict: + stats = c._call("debug.portal.stats", timeout_s=timeout_s) or {} if not isinstance(stats, dict): raise cmuxError(f"debug.portal.stats returned non-dict payload: {stats!r}") return stats @@ def _wait_for_portal_integrity(c: cmux, *, timeout_s: float, context: str) -> None: deadline = time.time() + timeout_s last = None error = None while time.time() < deadline: - last = _portal_stats(c) + remaining = deadline - time.time() + if remaining <= 0: + break + last = _portal_stats(c, timeout_s=min(remaining, 0.5)) error = _portal_integrity_error(last) if error is None: return time.sleep(POLL_S)Also applies to: 165-171
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py` around lines 124 - 125, _portal_stats currently calls c._call("debug.portal.stats") with the default RPC timeout which can exceed the overall _wait_for_portal_integrity deadline; change _portal_stats to accept a timeout/deadline parameter (e.g., timeout_s or remaining_s) and pass that through to c._call as the RPC timeout, and update callers (including _wait_for_portal_integrity and the other occurrences around lines 165-171) to compute and forward the remaining time so the RPC is bounded by the integrity-loop deadline.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Duplicate comments:
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py`:
- Around line 124-125: _portal_stats currently calls
c._call("debug.portal.stats") with the default RPC timeout which can exceed the
overall _wait_for_portal_integrity deadline; change _portal_stats to accept a
timeout/deadline parameter (e.g., timeout_s or remaining_s) and pass that
through to c._call as the RPC timeout, and update callers (including
_wait_for_portal_integrity and the other occurrences around lines 165-171) to
compute and forward the remaining time so the RPC is bounded by the
integrity-loop deadline.
There was a problem hiding this comment.
2 issues found across 8 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="Sources/TerminalController.swift">
<violation number="1" location="Sources/TerminalController.swift:8534">
P2: Avoid `DispatchQueue.main.sync` here; it can deadlock if the debug command is invoked on the main thread and it bypasses the existing `v2MainSync` guard. Use `v2MainSync` (or a `Thread.isMainThread` guard) to keep it safe and consistent.</violation>
</file>
<file name="tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py">
<violation number="1" location="tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py:125">
P2: The portal integrity poll is using an unbounded RPC timeout relative to the loop deadline. A slow `debug.portal.stats` call can block far longer than the integrity budget and make this test hang or become flaky. Pass an explicit short timeout for this RPC.</violation>
</file>
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
Contributor
Author
|
Addressed follow-up review items in commit e536b86:
@cubic-dev-ai please re-review. |
|
I'll kick off a fresh incremental review now covering those changes. ✅ Actions performedReview triggered.
|
@lawrencecchen I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py (1)
247-279: Guarantee workspace cleanup withfinallyto avoid test state leakage.At Line 278,
c.close_workspace(workspace_id)only runs on the success path. If any earlier step raises, the workspace can remain open and affect subsequent runs.♻️ Suggested refactor
with cmux(SOCKET_PATH) as c: c.activate_app() workspace_id = c.new_workspace() - c.select_workspace(workspace_id) - time.sleep(0.2) - - c.new_split("right") - layout = _wait_for_panes(c, 2, timeout_s=PANE_TIMEOUT_S, context="initial right split") - _wait_for_portal_integrity(c, timeout_s=INTEGRITY_TIMEOUT_S, context="after initial right split") - - for iteration in range(1, ITERATIONS + 1): - right_panel_id = _rightmost_panel(layout) - c.focus_surface_by_panel(right_panel_id) - c.new_split("down") - layout = _wait_for_panes( - c, - 3, - timeout_s=PANE_TIMEOUT_S, - context=f"iter={iteration} after split down", - ) - - bottom_right_panel_id = _bottom_right_panel(layout) - layout = _close_bottom_right_via_ctrl_d( - c, - bottom_right_panel_id=bottom_right_panel_id, - context=f"iter={iteration}", - ) - _wait_for_portal_integrity(c, timeout_s=INTEGRITY_TIMEOUT_S, context=f"iter={iteration} integrity") - if POST_CLOSE_SETTLE_S > 0: - time.sleep(POST_CLOSE_SETTLE_S) - - c.close_workspace(workspace_id) + try: + c.select_workspace(workspace_id) + time.sleep(0.2) + + c.new_split("right") + layout = _wait_for_panes(c, 2, timeout_s=PANE_TIMEOUT_S, context="initial right split") + _wait_for_portal_integrity(c, timeout_s=INTEGRITY_TIMEOUT_S, context="after initial right split") + + for iteration in range(1, ITERATIONS + 1): + right_panel_id = _rightmost_panel(layout) + c.focus_surface_by_panel(right_panel_id) + c.new_split("down") + layout = _wait_for_panes( + c, + 3, + timeout_s=PANE_TIMEOUT_S, + context=f"iter={iteration} after split down", + ) + + bottom_right_panel_id = _bottom_right_panel(layout) + layout = _close_bottom_right_via_ctrl_d( + c, + bottom_right_panel_id=bottom_right_panel_id, + context=f"iter={iteration}", + ) + _wait_for_portal_integrity(c, timeout_s=INTEGRITY_TIMEOUT_S, context=f"iter={iteration} integrity") + if POST_CLOSE_SETTLE_S > 0: + time.sleep(POST_CLOSE_SETTLE_S) + finally: + c.close_workspace(workspace_id)🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py` around lines 247 - 279, Wrap the test body that uses cmux and the created workspace in a try/finally so workspace cleanup always runs: after calling c.new_workspace() (workspace_id) and c.select_workspace(workspace_id) execute the split/iteration logic inside try and call c.close_workspace(workspace_id) in the finally block (guarding that workspace_id is set) to ensure the workspace is closed even on exceptions; reference cmux, c.new_workspace, c.select_workspace, and c.close_workspace to locate where to add the try/finally.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py`:
- Around line 145-149: The test currently coerces
totals["orphan_terminal_subview_count"],
totals["visible_orphan_terminal_subview_count"], and totals["stale_entry_count"]
via int(...), which masks malformed values; change this to strict integer
validation instead of coercion: read the raw values from totals (the keys
referenced above), assert their types are exact integers (e.g., type(value) is
int or equivalently isinstance(value, int) and not isinstance(value, bool)), and
raise/propagate a TypeError when a value is not an exact int so the test fails
closed on non-integer counters.
---
Nitpick comments:
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py`:
- Around line 247-279: Wrap the test body that uses cmux and the created
workspace in a try/finally so workspace cleanup always runs: after calling
c.new_workspace() (workspace_id) and c.select_workspace(workspace_id) execute
the split/iteration logic inside try and call c.close_workspace(workspace_id) in
the finally block (guarding that workspace_id is set) to ensure the workspace is
closed even on exceptions; reference cmux, c.new_workspace, c.select_workspace,
and c.close_workspace to locate where to add the try/finally.
ℹ️ Review info
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
Sources/TerminalController.swifttests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py
🚧 Files skipped from review as they are similar to previous changes (1)
- Sources/TerminalController.swift
Comment on lines
+145
to
+149
| try: | ||
| orphan = int(totals["orphan_terminal_subview_count"]) | ||
| visible_orphan = int(totals["visible_orphan_terminal_subview_count"]) | ||
| stale = int(totals["stale_entry_count"]) | ||
| except (TypeError, ValueError): |
There was a problem hiding this comment.
Fail closed on counter types; int(...) coercion can mask malformed portal stats.
At Line 146–148, using int(...) accepts/truncates non-integer values (0.7, "0", False), which can false-pass integrity checks that are supposed to reject non-integer counters.
💡 Suggested hardening
- try:
- orphan = int(totals["orphan_terminal_subview_count"])
- visible_orphan = int(totals["visible_orphan_terminal_subview_count"])
- stale = int(totals["stale_entry_count"])
- except (TypeError, ValueError):
+ orphan_raw = totals["orphan_terminal_subview_count"]
+ visible_orphan_raw = totals["visible_orphan_terminal_subview_count"]
+ stale_raw = totals["stale_entry_count"]
+ if type(orphan_raw) is not int or type(visible_orphan_raw) is not int or type(stale_raw) is not int:
return (
"portal totals contains non-integer counters "
- f"(orphan={totals.get('orphan_terminal_subview_count')!r}, "
- f"visible_orphan={totals.get('visible_orphan_terminal_subview_count')!r}, "
- f"stale={totals.get('stale_entry_count')!r})"
+ f"(orphan={orphan_raw!r}, visible_orphan={visible_orphan_raw!r}, stale={stale_raw!r})"
)
+ orphan = orphan_raw
+ visible_orphan = visible_orphan_raw
+ stale = stale_raw🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py` around lines 145
- 149, The test currently coerces totals["orphan_terminal_subview_count"],
totals["visible_orphan_terminal_subview_count"], and totals["stale_entry_count"]
via int(...), which masks malformed values; change this to strict integer
validation instead of coercion: read the raw values from totals (the keys
referenced above), assert their types are exact integers (e.g., type(value) is
int or equivalently isinstance(value, int) and not isinstance(value, bool)), and
raise/propagate a TypeError when a value is not an exact int so the test fails
closed on non-integer counters.
There was a problem hiding this comment.
♻️ Duplicate comments (1)
tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py (1)
145-149:⚠️ Potential issue | 🟠 MajorUse strict integer type checks for portal counters.
Line 145–149 still coerces with
int(...), which accepts values like"0",False, and0.7; that can false-pass malformeddebug.portal.statspayloads instead of failing closed.💡 Proposed hardening
- try: - orphan = int(totals["orphan_terminal_subview_count"]) - visible_orphan = int(totals["visible_orphan_terminal_subview_count"]) - stale = int(totals["stale_entry_count"]) - except (TypeError, ValueError): + orphan_raw = totals["orphan_terminal_subview_count"] + visible_orphan_raw = totals["visible_orphan_terminal_subview_count"] + stale_raw = totals["stale_entry_count"] + if ( + type(orphan_raw) is not int + or type(visible_orphan_raw) is not int + or type(stale_raw) is not int + ): return ( "portal totals contains non-integer counters " - f"(orphan={totals.get('orphan_terminal_subview_count')!r}, " - f"visible_orphan={totals.get('visible_orphan_terminal_subview_count')!r}, " - f"stale={totals.get('stale_entry_count')!r})" + f"(orphan={orphan_raw!r}, " + f"visible_orphan={visible_orphan_raw!r}, " + f"stale={stale_raw!r})" ) + orphan = orphan_raw + visible_orphan = visible_orphan_raw + stale = stale_raw#!/bin/bash set -euo pipefail # 1) Confirm current implementation still uses int(...) coercion. sed -n '131,160p' tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py # 2) Verify Python coercion behavior that can mask malformed counters. python - <<'PY' samples = ["0", False, 0.7, 0] for v in samples: try: print(f"value={v!r:<6} type={type(v).__name__:<7} int(v)={int(v)} strict_int={type(v) is int}") except Exception as e: print(f"value={v!r:<6} type={type(v).__name__:<7} int(v)=<error {type(e).__name__}> strict_int={type(v) is int}") PY🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py` around lines 145 - 149, The test currently coerces portal counters with int(...) which accepts strings, bools and floats; instead, fetch each value from totals (totals["orphan_terminal_subview_count"], totals["visible_orphan_terminal_subview_count"], totals["stale_entry_count"]) and perform a strict type check (require type(value) is int — not isinstance to exclude bool) before assigning to orphan, visible_orphan, and stale; if the check fails, raise or assert (e.g., ValueError/AssertionError) so malformed debug.portal.stats payloads fail the test rather than silently coerced.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Duplicate comments:
In `@tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py`:
- Around line 145-149: The test currently coerces portal counters with int(...)
which accepts strings, bools and floats; instead, fetch each value from totals
(totals["orphan_terminal_subview_count"],
totals["visible_orphan_terminal_subview_count"], totals["stale_entry_count"])
and perform a strict type check (require type(value) is int — not isinstance to
exclude bool) before assigning to orphan, visible_orphan, and stale; if the
check fails, raise or assert (e.g., ValueError/AssertionError) so malformed
debug.portal.stats payloads fail the test rather than silently coerced.
ℹ️ Review info
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
Sources/TerminalController.swifttests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.py
This was referenced Mar 4, 2026
Closed
This was referenced Mar 6, 2026
This was referenced Mar 17, 2026
This was referenced Mar 25, 2026
bn-l
pushed a commit
to bn-l/cmux
that referenced
this pull request
Apr 3, 2026
* Fix ghost terminal lifecycle rebind race * Address review feedback on portal regression checks * Address follow-up review feedback
This branch was successfully deployed
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
1 participant
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.
Summary
Fixes the ghost terminal behavior where a Ctrl-D closed terminal can briefly reappear in a split (half-present/orphaned surface state).
Root Cause
A stale portal bind path could reattach a terminal surface after close had already started, so visibility/active flags could flip back before deinit completed.
Changes
TerminalSurfaceportal lifecycle transitions.debug.portal.statsCLI debug command and portal stats plumbing.tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.pythat catches close->visible rebind races during split churn.Validation
CMUX_SOCKET=/tmp/cmux-debug-ghost-surface-orphan-regression.sock python3 tests_v2/test_split_cmd_shift_d_ctrl_d_no_portal_orphans.pyPASS: no close->visible rebind races during split-down + ctrl-d churn (iters=16, closes=16)Summary by cubic
Prevents “ghost” terminals from briefly reappearing after Ctrl-D by guarding portal binds with surface lifecycle/generation and sealing the surface on close/deinit. Adds portal integrity stats and a regression test that fails on any close→visible rebind or portal orphan/stale state.
Bug Fixes
New Features
Written for commit e536b86. Summary will update on new commits.
Summary by CodeRabbit
New Features
Bug Fixes
Chores
Tests