1603734 - browser_frame_history.js is strange and should probably be updated for e10s

Status: RESOLVED FIXED

(Firefox :: Session Restore, defect, P3)

Description

[Johann Hofmann [:johannh]]

I saw that this test was perma-failing on TV only in one of my try runs and investigated. It turns out that this test doesn't pass with e10s enabled at all locally, even on a fresh mozilla-central. I originally thought this was due to its use of contentDocument etc. but it seems like what the test is testing (frame history on session restore) is actually slightly broken with e10s on. I'm really confused.

• Is this test broken locally for everyone or just on my Mac?
• Why is this test not failing/running on e10s in infra?
• Why did the failure show up on TV?

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

The priority flag is not set for this bug.
:mikedeboer, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 2

[Mike de Boer [:mikedeboer]]

No clue! Patches welcome ;-)

Comment 3

[Neha Kochar [:neha]]

Nika looked at this and found out that it's failing at "iframe: Go back in time, 1 load should take place" so it's probably happening because of bfcache. The theory is that an earlier test might be disabling bfcache and that's why this test passes when run with the lot, but fails when run in isolation for TV because bfcache is enabled in those runs.

Comment 4

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

Explicitly disabling bfcache doesn't help. The test still doesn't pass locally. Investigating.

Comment 5

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

SessionStore has (bad) assumptions on non-existing or 0 value entry.ID.

Comment 6

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

Attachment: Bug 1603734 - browser_frame_history.js triggers a bug in sessionstore when run in e10s, r=jesup

SHIP starts IDs from 1, so non-SHIP could too.
I decided to change nsSHEntry and not SessionStore since that seems to be safer approach in case
one passes some explicit json data to the current SessionStore implementation and relies on the old
behavior.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=104f98e50e82359abc367f34551fa2f5100c578d

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/2ce4bb768681
browser_frame_history.js triggers a bug in sessionstore when run in e10s, r=farre

Comment 8

[Alexandru Michis [:malexandru]]

https://hg.mozilla.org/mozilla-central/rev/2ce4bb768681