1634759 - Discard BrowsingContext subtrees in a consistent manner across processes

Status: RESOLVED WONTFIX

(Core :: DOM: Navigation, enhancement, P3)

Description

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

This is the complete fix for issues such as those found in bug 1634596, bug 1580565, etc.

It will involve adding support to BrowsingContext and WindowContext to trigger unloading and discarding of entire subtrees across processes, and will require centralizing the corresponding logic from other components, such as nsDocShell::Destroy.

Issues caused by destruction order of BrowsingContext instances loaded in remote subframes are covered here.

Comment 1

[Neha Kochar [:neha]]

Moving to M6c as this isn't known yet to cause serious problems. We might see some crashes because of this once we start enabling Fission.

Comment 2

[Chris Peterson [:cpeterson]]

M7

Comment 3

[Neha Kochar [:neha]]

No recent evidence of needing this. And if BFCache ends up ignoring unload event, then this will be even lower priority so moving to Fission-MVP.

Comment 4

[Steven MacLeod [:smacleod]]

The crash in Bug 1634596 started triggering for me pretty consistently in my local linux debug build with the following steps:

1. Visit https://developer.mozilla.org/en-US/docs/Web/HTML/Element/iframe
2. Enable RDM (Web Developer -> Responsive Design Mode)
3. Set device to iPhone 6/7/8
4. Refresh
5. Content process crashes

I don't think RDM is responsible, since I'm able to trigger it just by loading the page or loading and then refreshing, but that happens much less consistently. I was unable to reproduce on Mac OS, but Neha was able to once (see crash report: https://crash-stats.mozilla.org/report/index/b6259856-04a4-495f-bcb0-44c2b0210122 ). Barret also helpfully reproduced this for me on Windows just by loading and refreshing. I'm not sure what about that page or the iframe on it makes it so likely to trigger this.

I have an RR recording of the crash happening.

Comment 5

[Steven MacLeod [:smacleod]]

(In reply to Steven MacLeod [:smacleod] from comment #4)

I have an RR recording of the crash happening.

https://pernos.co/debug/pd2iRe0Jet9CRYDKFOOOZw/index.html

Comment 6

[Chris Peterson [:cpeterson]]

Moving to Fission Future because this bug doesn't need to block Fission MVP.

Nika says the fix will be rather invasive.

Comment 7

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

I don't think that this particular change is necessary anymore with the current situation. In the future we may want to change things about how we destroy these types, but where we're at right now is OK.