Skip to content

test(computers): a multi env's computer follows its gateway's clock end to end on the local backend - #647

Closed
williamloo-scale wants to merge 1 commit into
wloo/computer-clock-syncfrom
wloo/computer-clock-e2e
Closed

williamloo-scale wants to merge 1 commit into
wloo/computer-clock-syncfrom
wloo/computer-clock-e2e

Conversation

@williamloo-scale

Copy link
Copy Markdown
Contributor

What and why

An end-to-end test of a multi env's computer on its gateway's clock on the local backend (tst/integration/computer/clock_follower_test.py, int_test_slow). It deploys a desktop (ubuntu-24.04, the reference driver and the clock app) beside an MCP server, then checks:

  • sync_env_clock sets the clock about 6 months back, and the arm takes under 30 s.
  • After the arm, after a one-day advance and after a re-arm at 60x, the session's date and Python agree with the gateway's clock to within 0.5 s.
  • The session user can't write the clock page or the clock file, read the follower's config, or kill the follower.
  • A follower killed as root is back for the next write.

How it was tested

pytest -m int_test_slow tst/integration/computer/clock_follower_test.py on local Docker (arm64) with the local registry: passed. The run is about 5 minutes with an image build. The amd64 image builds but wasn't run.

Stack: follower → sync → this → reset.

🤖 Generated with Claude Code

…nd to end on the local backend

The clock follower test bundle's desktop (ubuntu-24.04, the reference driver and the clock app) deployed beside an MCP
server. sync_env_clock arms the gateway about 6 months back, and the session's shell and Python read the gateway's time,
read between two reads of the gateway's clock; then after a day's advance and a re-arm at 60x, each write acknowledged
before it returns. The session user can't write the clock page or file, read the follower's config or stop the
follower, and a follower killed as root is back for the next write.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
@williamloo-scale
williamloo-scale deleted the branch wloo/computer-clock-sync November 12, 2026 23:38
@williamloo-scale

Copy link
Copy Markdown
Contributor Author

Closing: superseded by #649, which puts ubuntu-24.04 computers on the gateway's clock from the base image, with core configuring the follower at deploy and after a reset. FD-5477's remaining arm-time work (wait_followers_s and the refusals) stays with its assignee.

@williamloo-scale
williamloo-scale deleted the wloo/computer-clock-e2e branch November 12, 2026 23:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant