Support Edge instances in different displays #1013

[HeikoKlare]

The current implementation of the Edge browser creates a WebView2Environment on first creation of an WebView2 / Edge browser instance. On every further creation of a WebView2 instance, this environment is used. In case the WebView2 instance is created for a different display, i.e., within a different thread, the instantiation fails as the WebView2Environment has been created in a different thread.

To support the creation of WebView2 instances in different displays, this change replaces the static WebView2Environment with one environment for every display. The WebView2 instances are created for the environment belonging to the display for which the current instantiation is requested. An according regression test is added.

Contributes to #1013

A documentation for using multiple WebView2 environments can be found here: https://learn.microsoft.com/en-us/microsoft-edge/webview2/concepts/process-model?tabs=csharp#multiple-environment-objects

[github-actions]

Test Results

   299 files  ±0     299 suites  ±0   6m 28s ⏱️ -48s
 4 100 tests +1   4 092 ✅ +1   8 💤 ±0  0 ❌ ±0 
12 212 runs  +3  12 137 ✅ +1  75 💤 +2  0 ❌ ±0 

Results for commit 82d080b. ± Comparison against base commit 44432b2.

♻️ This comment has been updated with latest results.

[vogella]

Thanks @HeikoKlare

I suggest to merge this also today, so that we have test all your fixes in tomorrows I-build

[HeikoKlare]

I suggest to merge this also today, so that we have test all your fixes in tomorrows I-build

Sure, just note that "test" precisely means "manual test" because there are currently only three tests executed against the Edge browser (since I did not manage to make #672 running successfully yet).
At least I applied the commit of #672 to this PR to execute all the browser tests against the Edge browser as well, and only test_setUrl_local() fails (but also does without this PR).

I made slight changes after your review, in 52a73be. They concern the following:

• Store the environment an Edge instance belongs to within that instance, so that race conditions when accessing the global map on browser disposal are avoided
• Use waitForPassCondition() in tests instead of doing busy waiting without a timeout
• Dispose the additional displays in the test

There is one thing to consider left: the logic for clearing cookies (provided statically via Browser.clearSessions()) will clear the cookies of the environment for which the current thread is the UI thread. This is bit strange, as one might expect a static method to clear all cookies of all environments. And if you call the method on an object rather than the class, it is even more misleading as you might have browser1 and browser2 with different displays and call browser1.clearSessions() from the UI thread of browser2, resulting in the coockies of only browser2 to be cleared.
In my opinion, this is rather an existing API problem (clearSessions() being a static method), which we cannot fix easily. But I still wanted to point out that behavior.

[vogella]

Thank you @HeikoKlare Looking forward to test your changes tomorrow, but with the edge data dir set it is already very stable for us.