[Bug]: Using context.setOffline(true) before loading causes Webkit error

[kristojorg]

Version

1.49.1

Steps to reproduce

1. Clone repo: https://github.com/kristojorg/pw-repro
2. npm install
3. npm run test

Expected behavior

All tests pass

Actual behavior

The webkit and Firefox tests for "works offline when immediately setting browser to offline" fail, while the chrome version succeeds.

Additional context

I am not sure which behavior is correct. I ran in to this while trying to test an app in "offline mode" where I only allowed localhost requests and everything else failed. I also set context.setOffline(true) in order to trigger the navigator.online flag which my app uses internally. However, both webkit and firefox fail when setting them to offline before loading a page.

In addition, if you use the "show browsers" option, any subsequent tests will fail for Webkit after the "works offline when immediately setting browser to offline" test. This might be because the webkit browser is actually running into a bug itself.

In any case, I would expect these to behave the same across browsers. In addition, it might be useful to update the docs with instructions about how to test this scenario. I have included a working example using page.addlistener("load") to set the network offline only after the page loads.

Environment

System:
    OS: macOS 15.3
    CPU: (10) arm64 Apple M1 Pro
    Memory: 200.11 MB / 32.00 GB
  Binaries:
    Node: 20.17.0 - ~/.nvm/versions/node/v20.17.0/bin/node
    npm: 10.9.0 - ~/.nvm/versions/node/v20.17.0/bin/npm
    bun: 1.1.34 - /opt/homebrew/bin/bun
  IDEs:
    VSCode: 1.96.3 - /usr/local/bin/code
  Languages:
    Bash: 3.2.57 - /bin/bash
  npmPackages:
    @playwright/test: ^1.49.1 => 1.49.1

[Schweinepriester]

I believe we are facing the same issue at my company.

I tried https://github.com/kristojorg/pw-repro with @playwright/[email protected] instead of 1.49.1, but behaves the same.

[pavelfeldman]

"offline mode" where I only allowed localhost requests and everything else failed.

What is the user scenario that you are testing? Could you describe your objecting, expectations and the actual result?