1602115 - Make history.length Fission compatible

Status: RESOLVED FIXED

(Core :: DOM: Navigation, defect, P2)

Description

[Olli Pettay [:smaug][[e-mail]]]

history.length is from web page point of view synchronous, but once session history lives in the parent process, we can't get the information synchronously.

My current plan is to store length in (each) top level browsing context.
If a child process does an operation which would add something to the session history synchronously (for example pushState), child sends an async message with return value (== MozPromise) to parent process. The history.length becomes the cached length + pending MozPromises.
(This also shouldn't have the webcompat issues 1/2+pushState count and similar approaches would have)

This can be emulated even without session history in parent and assertions can be added whether the behavior of the new implementation matches the old one.

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

Attachment: Bug 1602115 - Make history.length Fission compatible

Comment 2

[Olli Pettay [:smaug][[e-mail]]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=c10b7971726e2c8770e059868a6a2adfe64c5143

https://treeherder.mozilla.org/#/jobs?repo=try&revision=a15a1ec5a553ecf542911a40c92d42e554f4f734

Comment 3

[Chris Peterson [:cpeterson]]

Deferring to Fission Nightly (M6) because this is related to session history.

Comment 4

[Chris Peterson [:cpeterson]]

M6a

Blocks session history in the parent

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

Attachment: Bug 1602115 - Make history.length Fission compatible, r=peterv

Depends on D78814

Comment 6

[Olli Pettay [:smaug][[e-mail]]]

Attachment: Bug 1602115, make it possible to test async history.length handling even when session history lives in the child process, r=peterv

Depends on D79197

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=27f8d9ac89698263d613b27b52e7f3090cb77212

The failure in [4] looks like https://bugzilla.mozilla.org/show_bug.cgi?id=1265637#c133

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/258d0ebd9e34
Make history.length Fission compatible, r=peterv
https://hg.mozilla.org/integration/autoland/rev/951c0fd65a00
make it possible to test async history.length handling even when session history lives in the child process, r=peterv

Comment 9

[Razvan Maries]

Backed out for perma failures.

Push with failure: https://treeherder.mozilla.org/#/jobs?repo=autoland&selectedTaskRun=AIuyI-j_TNC39zgDoF69qQ.0&resultStatus=testfailed%2Cbusted%2Cexception&revision=9fc2c601219a6bbb669d238887ce3165078bfb6f&failure_classification_id=2

Log: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=310051753&repo=autoland&lineNumber=2901

Backout: https://hg.mozilla.org/integration/autoland/rev/d391e3eebb4c233dddaa8f342ea5a27c32d2304f

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

Aha, the push just revealed a bug in https://hg.mozilla.org/mozilla-central/rev/15e2f6e7ea2f

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=8147a335849f377f9d48516507eafc08e08fc670

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/a55eebd5cb97
Make history.length Fission compatible, r=peterv
https://hg.mozilla.org/integration/autoland/rev/75f5953053de
make it possible to test async history.length handling even when session history lives in the child process, r=peterv

Comment 13

[Alexandru Michis [:malexandru]]

https://hg.mozilla.org/mozilla-central/rev/a55eebd5cb97
https://hg.mozilla.org/mozilla-central/rev/75f5953053de