You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 18eb47c
Browse filesBrowse the repository at this point in the historyBrowse files
chore: build the driver and the generated files from a pinned microsoft/playwright commit
DRIVER_VERSION holds a microsoft/playwright commit. scripts/upstream.py clones
microsoft/playwright into .upstream/ (gitignored) at that commit; the wheel build
builds the driver from it, and scripts/update_api.sh generates the API, the
protocol files and the test registry from it. Nothing is downloaded from npm.
- CI, publish, the Docker test and the Azure Docker pipeline run the same steps:
get Playwright, build, test. The Lint job also checks that the generated files
are up to date.
- CONTRIBUTING.md, ROLLING.md and the playwright-roll skill follow these steps.
`python scripts/upstream.py changes` lists the upstream diff to port, with the
Python file for each change.
- The wheel build starts from an empty build/, so that a deleted module does not
go into the wheel.
- ChannelOwner types its channel as Channel: mypy reports a subclass that does not
declare its channel type, as Root did not.
- The Docker test runs the suites as CI does: the dot reporter, two workers, and
the async suite also after a failed sync suite.
- Remove scripts/example_sync.py and scripts/example_async.py, left over from the
conda recipe.
0 commit comments