The spec states that creating new windows "must be done without invoking
the focusing steps for the created browsing context". It also states we
should do so by "running the window open steps", but nowhere in those
steps is there an option to invoke or skip any focusing steps. So we do
so with a custom WebContent IPC parameter.
Previously it was possible to have following sequence of calls
while destroying a session:
1. `WebContentConnection::die()` calls `Client::close_session()`
2. `Client::close_session()` removes a session from active sessions
map which causes session destructor call.
3. Session destructor calls `Client::close_session()` to remove a
session from active sessions.
With `stop()` method inlined into destructor `close_session()` need
to be called just once while destroying a session.
Currently we have `m_should_block_pop_ups` set to true by default
which means `choose_a_browsing_context` will early return if new
top-level browsing context is requested and write `Pop-up blocked!`
in console. It is good but when WebDriver is connected we want it
to be able to actually open a new window if one is requested.
Currently, on Serenity, we connect to WebDriver from the browser-side of
the WebContent connection for both Browser and headless-browser.
On Lagom, we connect from within the WebContent process itself, signaled
by a command line flag.
This patch changes Lagom browsers to connect to WebDriver the same way
that Serenity browsers do. This will ensure we can do other initializers
in the same order across all platforms and browsers.
This can avoid getting into a situation where lots of MouseMove events
are queued up and they all trigger relayout (or something else that
takes a lot of time).
To make sure that we don't get out of sync with the input events queue
on the UI process side, we still send acknowledgements for coalesced
MouseMoves. There's room for improvement here.
My Discord friends list is now pleasantly responsive. :^)
Before this patch, we had an issue where the WebContent process could
get backed up with tons of pending input events (especially mouse moves)
and have to work through all of those before responding to a paint
request from the UI process.
This could lead to a situation where we went for a very long time
without seeing any visual updates.
The approach I've taken here is pretty simple, we basically make a queue
of all incoming input events on the WebContent process side, and then
process that queue one event at a time, using a zero timer. This is
basic, but it allows paint requests to come in between the input events
and we do now get more frequent visual updates even during heavy
pressure from input events.
LibGUI and WebDriver (read: JSON) API boundaries use DeprecatedString,
so that is as far as these changes can reach.
The one change which isn't just a DeprecatedString to String replacement
is handling the "null" prompt response. We previously checked for the
null DeprecatedString, whereas we now represent this as an empty
Optional<String>.
Similar to POSIX read, the basic read and write functions of AK::Stream
do not have a lower limit of how much data they read or write (apart
from "none at all").
Rename the functions to "read some [data]" and "write some [data]" (with
"data" being omitted, since everything here is reading and writing data)
to make them sufficiently distinct from the functions that ensure to
use the entire buffer (which should be the go-to function for most
usages).
No functional changes, just a lot of new FIXMEs.
The spec states to only try to close the session *if* it exists. This
situation can occur when closing a session after a Close Window command,
as the session will be closed automatically if it was the last window.
These steps now have more than one caller; specifically, they may be
called from the Delete Session and Close Window endpoints. The session
was only removed from the active session map for the former endpoint.
Instead, let's more accurately handle removing the session where the
spec tells us to, so that all callers properly perform this step.
When some WebDriver spec steps are implemented a bit more literally, we
will end up in a situation where we remove a session from its client's
active session map, but still have more steps to perform. Currently,
when we remove the session, it is immediately destroyed because it is
stored in an OwnPtr. Instead, we can store it as a RefPtr, which will
let the caller to such steps keep the session alive until the subsequent
steps are complete.
While here, this also changes the storage of active sessions to a
HashMap, as all lookups into it are currently a linear search.
WebDriver::Session::close_window may invoke Session::stop, which needs
the WebContent connection to still exist. Do not remove the window's
handle (thus destroying the connection) until it is no longer needed.
With current architecture every window has its own WebContent process
and there is one WebDriver process that is responsible for talking to
all opened windows. It thus make sense to manage open windows from
WebDriver process instead of WebContent process that is not supposed
to know about all other opened WebContent processes.
This mostly reverts 826d5f8f9a but also
adds `web_content_connection` to window structure and window id
generation (currently out of spec).
With these changes `get_window_handles`, `switch_to_window` and
`close_window` start to actually switch, close and returned handles
of currently opened windows.
