1320235 - test timer flood scenario

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Ben Kelly [:bkelly, not reviewing]]

We should add a timer flood test in order exercising the improvements from bug 1300659.  This code is too fragile and could easily regress in the future without a test running in automation.

Comment 1

[Ben Kelly [:bkelly, not reviewing]]

Attachment: Test that the browser continues to function during timer floods. r=bz

This tests that we can load iframes, perform network requests, and requestAnimationFrame() calls during a timer flood.

Try build with test on current tree:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=09b55c260f590845da700e3ebb3f664f087fe2c2

Try build with test on tree with bug 1300659 effectively backed out.  This should show that the test times out or crashes without our timer improvements.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=925d3869763b96eeccb35ed9e1b76e5df41185a9

Comment 2

[Ben Kelly [:bkelly, not reviewing]]

Comment on attachment 8814283 [details] [diff] [review]
Test that the browser continues to function during timer floods. r=bz

Boris, can you review this test?  It basically triggers an exponential timer flood and then tries to do various things:

1. Load some frames
2. Perform some network requests
3. Run some requestAnimationFrame() callbacks

The previous try runs show this passes with current tree, but times out without the anti-flood timer patches (that have already landed).

Comment 3

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8814283 [details] [diff] [review]
Test that the browser continues to function during timer floods. r=bz

r=me but maybe add a comment about running rAF for 5s in the line that does the 5*60 bit?

Comment 4

[Ben Kelly [:bkelly, not reviewing]]

Attachment: Test that the browser continues to function during timer floods. r=bz

Added comments.

Also, I switched the flood window to be a child iframe.  I don't want this test to suddenly work just because separate windows are in a different process, etc.  I discussed this with Boris in IRC and he was ok with it.

New try:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=db8b2cc548ab89eb9976059ab6df38e2b57f38fc

Try with timer flood protections removed:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=44ec12461b8229db1f620be95ba2e579b9c1c39e

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/9907f708f3fb
Test that the browser continues to function during timer floods. r=bz

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/9907f708f3fb

Comment 7

[Ben Kelly [:bkelly, not reviewing]]

Since bug 1300659 landed on FF52 I uplifted the test as well:

https://hg.mozilla.org/releases/mozilla-aurora/rev/3fa3896980f6726c3897ce1645a839996d5d5e8b