896538 - Autophone - s1s2 - Regression in "time to throbber start" - 2013-07-16

Status: RESOLVED WONTFIX

(Firefox for Android Graveyard :: General, defect, P5)

Description

[Geoff Brown [:gbrown]]

http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstart/local-blank/2013-07-15/2013-07-22/notcached/noerrorbars/standarderror shows a regression in "Time to throbber start" for Local Blank Page, on July 16, for (almost?) all phones.

For example, for 44_a7_cf_ea_f9_73_lg-revolution, compare:

Tue Jul 16 2013 (5976b9c673f8): 1495±18
Tue Jul 16 2013 (41ed26826acb): 1753±21

Comment 1

[Brad Lassey [:blassey] (use needinfo?)]

Wes, Finkle suggested you may have done this sort of bisection before. Is that true?

Comment 2

[Aaron Train [:aaronmt]]

Has this regression resolved itself over time or is still being looked at for 25?

Comment 3

[Brad Lassey [:blassey] (use needinfo?)]

regression range http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=5976b9c673f8&tochange=41ed26826acb

Comment 4

[Lukas Blakk [:lsblakk] use ?needinfo]

This has been shipped in FF24 & FF25 so we wouldn't block FF26 on it but a low risk uplift nomination would be considered if ready.

Comment 5

[Ioana Chiorean]

(In reply to Brad Lassey [:blassey] (use needinfo?) from comment #3)
> regression range
> http://hg.mozilla.org/mozilla-central/
> pushloghtml?fromchange=5976b9c673f8&tochange=41ed26826acb

Removing the keyword as this was found already.

Comment 6

[Brad Lassey [:blassey] (use needinfo?)]

we need that regression window narrowed down to a single changeset

Comment 7

[Brad Lassey [:blassey] (use needinfo?)]

filter on [mass-p5]

Comment 8

[Richard Newman [:rnewman]]

Are we still planning to chase down a regression from three months ago, across an autophone measurement change?

Comment 9

[Brad Lassey [:blassey] (use needinfo?)]

(In reply to Richard Newman [:rnewman] from comment #8)
> Are we still planning to chase down a regression from three months ago,
> across an autophone measurement change?

This regression is from a year and 3 months ago. That said, identifying the changeset that caused the regression can still potentially give us a relatively easy perf win, which is why it tracks.

Comment 10

[Aaron Train [:aaronmt]]

I know nothing about autophone.

Comment 11

[Joel Maher ( :jmaher ) (UTC -8)]

and another year has gone by: I would think this should be closed as a wontfix