The reason for this change is that in Playwright Python we would need the related `protocol.yml` and `api.md` for the installed NPM package. For that we could either add the Git hash to the released package e.g. as a file (and go over the GitHub repo to get the file content) but Pavel proposed that it might be better to include the two files in the NPM package.
Tested locally by adding to the `utils/publish_all_packages.sh` script `--dry` to the NPM publish commands.
cc @aslushnikov @pavelfeldman
Related issues: https://github.com/microsoft/playwright-python/pull/101 and https://github.com/microsoft/playwright-python/pull/96
This adds one more protocol message __dispose__
to dispose a scope and all child objects.
Now, client side does not need a notion of scope anymore -
it just disposes the whole object subtree upon __dispose__.
Server, on the other hand, marks some objects as scopes
and disposes them manually, also asserting that all parents
are proper scopes.
This makes it easier to reason about our packages.
The only difference is what each package downloads.
When the browser is not downloaded, it will fail to launch.
Each browser gets a 'download' attribute in the browser.json file.
The original plan was to rnu some checks against libc version the
binary is compiled with, but these turn out to be a little complicated:
parsing out libc version from both static binary and host system
requires text processing, and it's hard to make sure it works reliably
across distributions.
Instead, let's start with a very particular check against running
Firefox on Ubuntu 16.04.
References #2745
Before typing/pressing, we focus the target element. WebKit
sometimes selects the value in this case. To unify the behavior
between the browsers we behave similar to human:
- when the input is already focused, we just type;
- when the input is not focused, we focus it, move caret
to the start (like if user clicked at the start to focus the input)
and then type.
Note this only affects inputs with non-empty value.
This establishes a single naming for all our blobs with browser
builds that we upload to CDN: `<browser-name>-<os-version>`
- `<browser-name>` is either `firefox` or `webkit`.
- `os-version` is the OS that was used to produce the build.
References #2745
Originally there was no way to build both webkit-gtk and webkit-wpe
in the same checkout. As a result, we were:
- building webkit-gtk & uploading it to CDN
- building webkit-wpe & uploading it to CDN
- downloading webkit-gtk & webkit-wpe from CDN and compiling a single
webkit-gtk-wpe folder that we uploaded to CDN
As of today, however, we can build WPE and GTK ports together. This
patch starts using this to drastically simplify build process:
- build webkit-gtk and webkit-wpe
- use `archive.sh` script that compiles a single folder with both ports
- upload archive to the cloud
This should also fix currently failing webkit build that fails to
assemble the `webkit-gtk-wpe` folder (I missed these codepaths while preparing
for Ubuntu-20.04 bot).
References #2745
This patch:
- specializes "linux" scripts into "Ubuntu 18.04" scripts
- renames all future linux blobs on CDN:
* `firefox-linux.zip => firefox-ubuntu-18.04.zip`
* `minibrowser-gtk.zip => minibrowser-gtk-ubuntu-18.04.zip`
* `minibrowser-wpe.zip => minibrowser-wpe-ubuntu-18.04.zip`
* `minibrowser-gtk-wpe.zip => minibrowser-gtk-wpe-ubuntu-18.04.zip`
- updates downloader to deal with the new names
References #2745
