1541640 - Pageload regression for youtube.com since FF58

Status: RESOLVED DUPLICATE of bug 1540777

(Core :: Storage: localStorage & sessionStorage, defect)

Description

[Denis Palmeiro [:denispal]]

Firefox 58:
[2019-04-03 18:26:08] INFO: https://www.youtube.com 0, backEndTime: 353ms (±9.67ms), firstPaint: 454ms (±13.76ms), firstVisualChange: 2.04s (±33.89ms), DOMContentLoaded: 2.82s (±40.61ms), Load: 2.94s (±43.72ms), speedIndex: 2384 (±36.44), visualComplete85: 3.18s (±52.56ms), lastVisualChange: 4.27s (±91.04ms), rumSpeedIndex: 2392 (±35.46) (10 runs)

Firefox 67:
[2019-04-03 18:42:06] INFO: https://www.youtube.com 0, backEndTime: 340ms (±4.60ms), firstPaint: 434ms (±7.31ms), firstVisualChange: 2.80s (±104.14ms), DOMContentLoaded: 3.58s (±103.24ms), Load: 3.70s (±99.68ms), speedIndex: 3143 (±102.48), visualComplete85: 3.93s (±107.09ms), lastVisualChange: 4.82s (±97.08ms), rumSpeedIndex: 3107 (±89.58) (10 runs)

Comment 1

[Denis Palmeiro [:denispal]]

Caused by the dom.storage.next_gen pref flip:

Tested with April 4th Nightly:

dom.storage.next_gen = true
[2019-04-05 21:06:38] INFO: https://www.youtube.com 0, backEndTime: 306ms (±6.24ms), firstPaint: 388ms (±11.46ms), firstVisualChange: 2.82s (±77.83ms), DOMContentLoaded: 3.68s (±81.90ms), Load: 4.70s (±109.61ms), speedIndex: 3522 (±79.96), visualComplete85: 3.79s (±83.18ms), lastVisualChange: 15.06s (±122.60ms), rumSpeedIndex: 2316 (±56.70) (5 runs)

dom.storage.next_gen = false
[2019-04-05 21:16:19] INFO: https://www.youtube.com 0, backEndTime: 337ms (±10.03ms), firstPaint: 433ms (±15.85ms), firstVisualChange: 1.94s (±32.56ms), DOMContentLoaded: 2.68s (±52.69ms), Load: 3.87s (±34.57ms), speedIndex: 2675 (±38.04), visualComplete85: 2.93s (±34.94ms), lastVisualChange: 14.32s (±59.69ms), rumSpeedIndex: 1747 (±29.79) (5 runs)

Comment 2

[Jan Varga [:janv]]

pageload of youtube.com is measured by our tp6 talos tests and maybe some new raptor tests.
How come we don't see a regression there ?
Actually I paid special attention to youtube.com, using tp6 to measure performance and I didn't see any regressions in the end.

Comment 3

[Denis Palmeiro [:denispal]]

(In reply to Jan Varga [:janv] from comment #2)

pageload of youtube.com is measured by our tp6 talos tests and maybe some new raptor tests.
How come we don't see a regression there ?
Actually I paid special attention to youtube.com, using tp6 to measure performance and I didn't see any regressions in the end.

I'm not really sure why tp6 didn't catch this. It could be any of :

a) tp6 tests are using recordings and I'm testing live sites here
b) different testing frameworks and methodologies
c) the tp6 recordings are too old?

Most likely b.

Comment 4

[Jan Varga [:janv]]

I think the summary should be changed because the regression is probably caused by enabling LSNG on Mar 8th.