Even though background page startup is delayed, we still instantiate a BackgroundPage object during startup which instantiates a hidden window. We don't need to be doing that during startup. Kris suggested we could also share a single window for all hidden extension pages.

Also, use a single hidden window to hold <browser> elements for multiple extensions.

Pushed by aswan@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/4db0d3e10cb1 Defer window/browser creation for extension pages r=kmag

Backed out changeset 4db0d3e10cb1 (bug 1498027) for a11y failures Backout: https://hg.mozilla.org/integration/autoland/rev/aee6c9dbb430598724cab97d8e8c601cfe54e4b9 Failure push: https://treeherder.mozilla.org/#/jobs?repo=autoland&revision=4db0d3e10cb162b850e90948015a7ed99e18b0e5 Failure log: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=216408054&repo=autoland&lineNumber=11213 [task 2018-12-11T18:27:00.528Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXPCComponents_Exception [task 2018-12-11T18:27:00.529Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXPCComponents_ID [task 2018-12-11T18:27:00.529Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 6 nsXPCComponents_Interfaces [task 2018-12-11T18:27:00.530Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 6 nsXPCComponents_Results [task 2018-12-11T18:27:00.530Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 6 nsXPCComponents_Utils [task 2018-12-11T18:27:00.531Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXPCComponents_utils_Sandbox [task 2018-12-11T18:27:00.531Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 38 nsXPCWrappedJS [task 2018-12-11T18:27:00.532Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 11 nsXPCWrappedJSClass [task 2018-12-11T18:27:00.532Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXPConnect [task 2018-12-11T18:27:00.533Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXULCommandDispatcher [task 2018-12-11T18:27:00.534Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 2 nsXULControllers [task 2018-12-11T18:27:00.534Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1941 nsXULElement [task 2018-12-11T18:27:00.535Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 26 nsXULPopupListener [task 2018-12-11T18:27:00.535Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXULPopupManager [task 2018-12-11T18:27:00.536Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 4145 nsXULPrototypeAttribute [task 2018-12-11T18:27:00.536Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXULPrototypeCache [task 2018-12-11T18:27:00.537Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 12 nsXULPrototypeDocument [task 2018-12-11T18:27:00.538Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1503 nsXULPrototypeNode [task 2018-12-11T18:27:00.538Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsXULTooltipListener [task 2018-12-11T18:27:00.539Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 2 nsZipArchive [task 2018-12-11T18:27:00.539Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 1 nsZipReaderCache [task 2018-12-11T18:27:00.540Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 8 xpc::CompartmentPrivate [task 2018-12-11T18:27:00.540Z] 18:27:00 INFO - TEST-INFO | leakcheck | default leaked 2 xpcJSWeakReference [task 2018-12-11T18:27:00.541Z] 18:27:00 INFO - TEST-UNEXPECTED-FAIL | leakcheck | default 1819396 bytes leaked (AbstractThread, AbstractWatcher, AnimationTimeline, AsyncFreeSnowWhite, AtomSet, ...) [task 2018-12-11T18:27:00.542Z] 18:27:00 INFO - [task 2018-12-11T18:27:00.542Z] 18:27:00 INFO - runtests.py | Running tests: end. [task 2018-12-11T18:27:00.543Z] 18:27:00 INFO - Buffered messages finished [task 2018-12-11T18:27:00.543Z] 18:27:00 INFO - SUITE-END | took 363s [task 2018-12-11T18:27:00.592Z] 18:27:00 INFO - Return code: 0

Pushed by aswan@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/66a1ed62e2d8 Defer window/browser creation for extension pages r=kmag

This generated a performance improvement == Change summary for alert #18256 (as of Mon, 17 Dec 2018 09:59:05 GMT) == Improvements: 6% sessionrestore opt e10s stylo 326.92 -> 307.92 6% sessionrestore_no_auto_restore opt e10s stylo 344.92 -> 325.92 5% sessionrestore pgo e10s stylo 317.83 -> 300.50 5% sessionrestore opt e10s stylo 334.42 -> 317.33 5% sessionrestore_no_auto_restore opt e10s stylo 354.71 -> 336.75 5% sessionrestore_no_auto_restore pgo e10s stylo 338.50 -> 321.50 5% sessionrestore opt e10s stylo 385.79 -> 366.50 5% sessionrestore_no_auto_restore pgo e10s stylo 332.67 -> 317.00 4% sessionrestore pgo e10s stylo 318.04 -> 304.08 4% sessionrestore opt e10s stylo 351.75 -> 338.58 For up to date results, see: https://treeherder.mozilla.org/perf.html#/alerts?id=18256

Hey Gijs, this seemed relevant to the about:blank stuff you were doing, so putting it on your radar.