Repository navigation
Complete recording setup and isolate terminal project recordings - #335
Merged
Merged
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configuration
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
This was referenced Oct 14, 2026
guan4tou2
marked this pull request as ready for review
October 14, 2026 17:48
This was referenced Oct 15, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Capture setup can be reopened after skipping or recording, without resetting evidence. Prepare → Prove keeps missing tools, errors, manual remedies and retry visible; installation and skipping never verify capture. Project rotation preserves live panes while closing and hashing separate recording segments. Per-pane stop also stops raw casts, and cross-boundary completions are explicitly partial.
Managed HTTP requires a private per-start response from its owned addon. Process liveness, foreign listeners, buffered output or stale responses cannot report running. Readiness probes stay outside engagement evidence and user HTTP/HTTPS proof. Startup is bounded and stops only its own child on failure.
The change also repairs spool collisions and receiving-profile identity, transient PTY EOF/EIO loss, signing-key replacement, frozen filtered exports, changed attachments and projection verification. Installed drivers use the selected executable, isolate Electron user data and assert the packaged runtime.
Dependency installation now has a finite fifteen-minute bound, elapsed waiting time, keyboard cancellation and a distinct cancelled/retry state. Leaving, reloading or losing the requesting renderer, or quitting the app, stops its owned install tree. One install per dependency prevents duplicate jobs. A real installed-app test checks both child and grandchild termination while an unrelated process survives. The reload case was first RED and is preserved. Strict smoke-report gates reject missing tests, skips, failures, flakes and not-exercised steps.
Current source 80cbfea: the five ordered local gates pass; 3,605 unit tests pass / 24 explicit skips. The rebuilt ad-hoc-sealed macOS ARM64 candidate passes thirteen installed journeys, with a strict JSON gate confirming zero skips/not-exercised. This includes the immediate first Bash command's actual end/exit/cast output, bundle/verifier/restart, controlled-process cancel/leave/reload, ten setup/recovery cases using controlled IPC, and a fresh private tool/cache/Python install (42,381ms) followed by real HTTP/HTTPS nonce proof using only the test-profile CA. Network timing is not attributed to the timer fix. Original timeout/reload/first-command failures remain retained; the eight-minute healthy installer regression survives and a hung installer still stops at fifteen minutes.
Final DMG SHA-256: 39acaebdb1b8c85f1be53f2d87a6320752f59216d21f6c1b9b15e5b7cf13a9a6. Strict resource sealing passes. The old unsealed DMG (9d964e…) is superseded; ad-hoc signing is neither publisher authentication nor notarization.
An isolated native Kali 2026.3 ARM64 desktop VM previously passed five installed journeys on the bae0415 product: 100 unique commands with stdout/stderr/exits, native nc job control, PTY helper/profile removal, actual HTTP/HTTPS → projection bundle/shipped verifier → restart, and pane rotation/privacy. Final command interval 83,956ms; enrollment plus verification 84,047ms excludes DEB installation and human setup. Tools were prepared separately; guest external networking stayed disabled. The operator also manually confirmed Ctrl-Z → raw/fg with resumed peer data; kernel polling confirms stop/resume/tty descriptors and local chain verification. The controller's late acknowledgement remains raw completed:false; this is not a full automated quickstart/first-install stopwatch.
New native jobs build/install/remove the Linux x64 DEB, relocate and launch the actual AppImage, and launch the actual Windows portable self-extracting wrapper. Portable checks builtin output/cast, native PowerShell metadata/exits, bundle/verifier and process-termination/restart; it does not claim graceful temporary-directory cleanup or WSL. ddea520 CI passed (Ubuntu unit 3,608/21 skips, Windows unit 3,293/340 skips, Electron 132/7 skips), and NSIS installed/captured/exported/uninstalled with zero smoke skips. Linux DEB: five passed, one failed; AppImage was not reached. Portable: failed before command proof because its driver did not leave missing-tool preparation. The DEB first-command loss reproduces in the installed macOS Bash: its cast executes the canary at 0.224s, while the hook is injected at 0.604s. The startup hook is now queued before the pane accepts input, without adding a test sleep. The portable driver explicitly continues to independent command proof. 80cbfea all three workflows PASS: CI (Ubuntu 3,608/21 explicit skips, Windows 3,293/340 skips, Electron 132/7 skips); Windows NSIS + actual portable (both complete, no smoke skips/not-exercised); Linux x64 DEB + AppImage (six DEB journeys, uninstall, two relocated AppImage journeys, both strict gates pass). AppImage uses APPIMAGE_EXTRACT_AND_RUN=1, not a FUSE/default desktop launch claim. Linux's 100-command interval is 40,105ms (40,137ms including enrollment/verification), excluding installation/human setup. Binaries, raw reports and GitHub artifact digests are preserved. ddea520's earlier results remain historical. Historical 1fd2879 CI and NSIS passed; those results do not certify ddea520.
One local full run additionally observed an external Zsh output downgraded to metadata-only; its unchanged six-test focus and subsequent full suite pass. Raw diagnostics are retained without assigning an unproven retrospective cause.
Still open: native WSL, full manual Kali quickstart/first-install stopwatch, actual quarantined Mac download/Dock launch and Developer ID/notarization. Specs 047/055/056 remain Implemented; Spec 052 remains Draft. No merge or release.
Original #333 single-occurrence syscall traces are unavailable. The authorized six-file free OpenCode review timed out without a final review; no paid fallback or extra upload was used. See docs/PACKAGED-SMOKE.md and spec verification records for evidence boundaries.
Related: #325, #326, #327, #329, #331, #332, #333, #226, #317. Timeline redesign remains deferred.
Artifact-retention limit: the later AppImage Playwright run cleaned the earlier DEB JSON from test-results before upload. The raw Linux job log preserves all six DEB passes and its strict JSON gate PASS; four other copied JSON reports were revalidated after download. No missing DEB JSON was reconstructed. A retained multi-format workflow should use separate output directories; an installed-only Linux workflow also removes that overlap.