Skip to content

fix(release): the browser base's smoke starts it with init, as its image does - #741

Merged
earakely-scale merged 1 commit into
mainfrom
edgararakelyan/browser-base-smoke-starts-init
Nov 17, 2026
Merged

earakely-scale merged 1 commit into
mainfrom
edgararakelyan/browser-base-smoke-starts-init

Conversation

@earakely-scale

@earakely-scale earakely-scale commented Nov 17, 2026 •

Copy link
Copy Markdown
Collaborator

Why

v0.9.1552's publish run passed main-ci and published the gateway image and the protocol. Then its agent-build (agentenv-ubuntu-24.04-browser) failed at the smoke check, so agent-publish was skipped and agentenv-framework stayed off PyPI, which still shows 0.9.1549:

##[error]agentenv-ubuntu-24.04-browser: `docker run --rm --entrypoint sh …-browser:src-d07a…-arm64 -c start-computer & for _ in $(seq 60); …
unix:///run/agentenv/supervisor.sock no such file
usage: start-computer init|xvfb

#708 gave the browser base's start-computer two subcommands: init (the root start, which starts the session's supervisord) and xvfb (which supervisord runs). The image's CMD became start-computer init, but the release's smoke in .github/scripts/infra_build.py still ran start-computer with no argument. That printed the usage and exited, so supervisord never started. Only the release runs that smoke, so no pull request's CI caught it.

What changes

  • The browser base's smoke runs start-computer init &, as its image's CMD does. The desktop base's smoke already runs start-desktop init &.
  • New unit test: each base computer's smoke starts the base with its image's own CMD, read from its Dockerfile. It fails on main for the browser base and passes here, so a later change to a base's start can't leave the release's smoke behind.

Testing

  • tst/unit/test_infra_build_script.py: 79 passed. The new test fails without the fix.
  • This rebuilds nothing new. The browser base's image starts with start-computer init in every deploy, which the computer integration tests run in CI.

This PR carries bump-version, so merging it cuts the release that publishes #725, #731, #732, #708 and #733.

🤖 Generated with Claude Code

RetriggerView in GreptileConfidence Score: 5/5

The PR appears safe to merge.

Summary

Fixes the browser base’s release smoke check by adding the required init argument.

  • The browser base smoke starts the supervisor with its image command.

Reviews (1) · Last reviewed commit: "fix(release): the browser base's smoke s..." · Reviewed by Greptile

…age does

#708 gave start-computer the subcommands init and xvfb, and the image's CMD runs start-computer init. The release's
smoke of agentenv-ubuntu-24.04-browser still ran start-computer with no argument, which printed its usage and
exited, so v0.9.1552's agent-build failed and agentenv-framework stayed off PyPI. A unit test now ties each base's
smoke to its image's CMD.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@earakely-scale
earakely-scale requested a review from a team as a code owner November 17, 2026 17:27
@earakely-scale
earakely-scale merged commit ff44021 into main Nov 17, 2026
35 checks passed
@earakely-scale
earakely-scale deleted the edgararakelyan/browser-base-smoke-starts-init branch November 17, 2026 17:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant