1589517 - Make it possible for the parent process to choose a content process to assign a remote <xul:browser> to

Status: RESOLVED FIXED

(Core :: DOM: Content Processes, task, P2)

Description

[Mike Conley (:mconley) (:⚙️)]

This is going to be needed for Picture-in-Picture, which opens up an always-on-top window, and inserts a <xul:browser> that is supposed to run in the same content process as the originating video.

Right now, we use the old sameProcessAsFrameLoader mechanism to pull this off with the originating tab's frameLoader: https://searchfox.org/mozilla-central/rev/ebe492edacc75bb122a2b380e4cafcca3470864c/toolkit/components/pictureinpicture/content/player.js#91

But that's not going to fly with Fission. Any suggestions, Nika?

Comment 1

[Nika Layzell [:nika] (ni? for response)]

This might work if we establish an opener relationship between the PiP window & the source frame's window. If you do that & the principal matches, it should load into the same process.

Not sure if we have any super handy facilities for this from the parent process, but it shouldn't be too hard to add one if we need it.

Altenatively we could have something like sameProcessAsWindow which takes an inner window id?

Comment 2

[Chris Peterson [:cpeterson]]

mconley says this bug is needed for PiP Fission. PiP blocks enabling Fission in Nightly (M6), so we'll need this bug for M6, too.

Comment 3

[Chris Peterson [:cpeterson]]

Tracking for Fission M6b

Where possible, we'd like DocumentChannel to own this process logic, not <xul:browser>.

kmag says he'll need to do something similar for WebExtension BCG.

Comment 4

[Neha Kochar [:neha]]

Leaving NI for Nika here as she has an idea for implementing this.

Comment 5

[Nika Layzell [:nika] (ni? for response)]

Attachment: Bug 1589517 - Part 1: Add initialBrowsingContextGroupId attribute,

This attribute will subsume the existing sameProcessAsFrameLoader attribute. It
works by specifying the BrowsingContextGroup which the initial BrowsingContext
in a <browser> should be created within.

Due to bug 1652144, all documents within the same BrowsingContextGroup with the
same remote type will be loaded in the same process, meaning that specifying
both "initialBrowsingContextGroupId" and "remoteType" will cause the initial
about:blank document to be loaded in a specific content process.

Comment 6

[Nika Layzell [:nika] (ni? for response)]

Attachment: Bug 1589517 - Part 2: Add initialBrowsingContextGroupId to sameProcessAsFrameLoader callers,

This attribute, when combined with the remoteType attribute, should subsume all
callers of sameProcessAsFrameLoader, and also support selecting the same process
as an out-of-process iframe.

Comment 7

[Nika Layzell [:nika] (ni? for response)]

Attachment: Bug 1589517 - Part 3: Add initialBrowsingContextGroupId to extension browsers,

This attribute can be used to force non-tab extension browsers to be loaded in
the correct BrowsingContextGroup, and also subsumes the existing
sameProcessAsFrameLoader uses in extension code.

Comment 8

[Nika Layzell [:nika] (ni? for response)]

Attachment: Bug 1589517 - Part 4: Remove sameProcessAsFrameLoader,

The functionality has been fully subsumed by the new
initialBrowsingContextGroupId attribute, so it is no longer necessary.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/7f5deeecabd4
Part 1: Add initialBrowsingContextGroupId attribute, r=farre
https://hg.mozilla.org/integration/autoland/rev/bed38dcb3c26
Part 2: Add initialBrowsingContextGroupId to sameProcessAsFrameLoader callers, r=mconley,smacleod
https://hg.mozilla.org/integration/autoland/rev/c6bac228345d
Part 3: Add initialBrowsingContextGroupId to extension browsers, r=zombie
https://hg.mozilla.org/integration/autoland/rev/5cdeca6b1feb
Part 4: Remove sameProcessAsFrameLoader, r=zombie,mconley,farre,smacleod

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/91dcf4dce9e6
Remove unnecessary aSameProcessAsFrameLoader declaration from tabbrowser.js for causing lint failures. a=lint-fix

Comment 11

[Cristian Brindusan [:cbrindusan]]

https://hg.mozilla.org/mozilla-central/rev/7f5deeecabd4
https://hg.mozilla.org/mozilla-central/rev/bed38dcb3c26
https://hg.mozilla.org/mozilla-central/rev/c6bac228345d
https://hg.mozilla.org/mozilla-central/rev/5cdeca6b1feb
https://hg.mozilla.org/mozilla-central/rev/91dcf4dce9e6