ci(macos-x64): measure WebKit in a docker-mac-x64 guest; keep the hosted render lane - #261
Merged
Merged
Conversation
…he guest Refs #251 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Refs #251 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…ment Refs #251 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Safari reads a command-line argument as a sandboxed file path and asks for consent to read it, which nothing in the guest can grant. A dependency-free helper calls LSOpenCFURLRef, as /usr/bin/open would. Also records the outcome in macos-arm-live-test.yml. Refs #251 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…r Safari Refs #251 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…ients The LaunchServices helper brought Safari forward without loading the URL, so it is removed. A home page preference written before Safari first runs loads the page. The probe now fails unless both the shipped viewer and Safari report. Refs #251 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…rint The slot-release clients sent their command on open. On a slow runner the login shell had printed nothing within the 15 s ready timeout: only the terminal's echo of the typed line came back (Linux x86 Terminal Test, run 35191433983). Wait for the shell's first output, as a user would, and allow 30 s. The Safari smoke used the same pattern. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
zackees
changed the title
zackees
marked this pull request as ready for review
September 17, 2026 07:17
2 of 10 tasks
zackees
added a commit
that referenced
this pull request
Sep 18, 2026
CLI: - `fastled <sketch>` no longer crashes with "Cannot drop a runtime" after starting the server (#269). - `--terminal-cmd` / FASTLED_TERMINAL_CMD runs an agent in the browser terminal, and a configured session survives a page reload through a single-use reattach token (#254, #267). Terminal: - The shell starts in the named directory and emits valid exit frames (#257). - A client that disconnects mid-write releases its session slot (#256, #258). Viewer: - Screen maps registered after setup() now reach the renderer, so sketches like Blink no longer show an empty canvas (#250, #264). - The page no longer overflows the window by 40 px, and the title scales with the window (#268, #270). CI: - Browser terminal suite on Linux, Windows and real Safari (#259, #260). - Linux render smoke that fails on a blank canvas (#247, #266). - Linux dependencies install before setup-soldr, so the dependency cook succeeds (#263); lint covers ci/ (#265); an Intel macOS guest WebKit probe (#251, #261). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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.
Fixes #251.
Result: a Linux-hosted macOS guest cannot render a sketch
The acceptance criteria allow recording why and keeping the hosted lane when the guest can't render. That's what this PR does. The reason was measured in a real guest, not assumed.
macos-x64-guest-webkit-probe.ymldoes the following:fastledbinary on Linux.ubuntu-latest.fastled --internal-viewer, the same WKWebView windowfastled <sketch>opens;Each client reports through the fastled server's
/viewer-logroute.Final run (35191433924):
Pages load and animate at about 30 fps, but neither WebKit client offers any WebGL. The guest has a plain QEMU VGA adapter and no GPU.
bootstrap.tsmakes WebGL2 a hard requirement, so no sketch can render there, and no screenshot of a render is possible. A newer macOS image wouldn't add a GPU.Measured duration: guest boot plus probe took 391–579 s across runs, and the whole job 9–16 min. The probe is not a per-PR gate. It runs on demand, and on PRs that touch the probe, so it can be repeated if the guest image or GPU situation changes.
Findings about the Recovery guest, kept in the probe
/var/rootis read-only, sofastledneedsHOMEunder/tmp.python3,open,screencapture,safaridriver,osascriptorsystem_profiler.LSOpenCFURLRef/LSOpenFromURLSpec, tried in this PR) brings Safari forward without loading the URL.Changes
.github/workflows/macos-x64-guest-webkit-probe.yml,ci/macos-x64/probe-guest.shandci/macos-x64/webgl.html. The job fails unless both the viewer and Safari report.macos-arm-live-test.yml: the comment now records why execution stays on the hosted runner.tests/frontend/test_terminal.pyandci/safari_terminal_smoke.py. The failing run was 35191433983: the login shell had printed nothing within 15 s.No JSPI flags or product code changes.
🤖 Generated with Claude Code