879357 - Autophone - s1s2 - Regression in "time to throbber stop" on May 31 -- Nexus One and Droid Pro only

Status: RESOLVED WORKSFORME

(Firefox for Android Graveyard :: General, defect)

Description

[Geoff Brown [:gbrown]]

http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstop/local-blank/2013-05-05/2013-06-04/notcached/noerrorbars/standarderror shows a  regression in "Time to throbber stop" for "Local blank page" on May 31, but only on the Droid Pro and both Nexus One devices.

For example, for 00_23_76_b1_92_39_nexus-one compare 

Fri May 31 2013 (4c7a395fadcc): 5562±38
Fri May 31 2013 (2222b07ab207): 6633±97

Comment 1

[Aaron Train [:aaronmt]]

Kevin has more experience looking into these particular issues

Comment 2

[Kevin Brosnan [Ex-Mozilla]]

What is the best way to go about running this test? Can you push revisions to try?

Comment 3

[Mark Finkle (:mfinkle) (use needinfo?)]

(In reply to Kevin Brosnan [:kbrosnan] from comment #2)
> What is the best way to go about running this test? Can you push revisions
> to try?

Naoki used to have a script that would start Fennec via an Intent (over adb) which could be used to look for the "Throbber Stop" adb logcat output. Since we are only looking for a relative change, not the exact change, you could even just run Fennec via an Intent and load about:blank. Look in the logcat for the last "Throbber Stop" time.

Oh, I just found this:
https://wiki.mozilla.org/Mobile/Fennec/Android#S1.2FS2_Automation

Comment 4

[Geoff Brown [:gbrown]]

Also, :mcote may be able to help get you set up.

Comment 5

[Mark Côté [:mcote]]

These tests are run through autophone: https://github.com/mozilla/autophone

There are some installation instructions in there.  I am out for a couple days, but I can help you out next week.  :bc also works on this system (more so than I do these days, in fact).

Comment 6

[Kevin Brosnan [Ex-Mozilla]]

Picked up a Droid Pro from Brian. Will look at this next week.

Comment 7

[Kevin Brosnan [Ex-Mozilla]]

URL in comment 0 is dead now http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstop/local-blank/2013-05-30/2013-06-04/notcached/noerrorbars/standarderror

Comment 8

[Kevin Brosnan [Ex-Mozilla]]

Attachment: Startup adb log

adb shell am start -a android.intent.action.VIEW -n org.mozilla.fennec_kbrosnan/.App -d file://mnt/sdcard/download/startup6.html

Assumption time to throbber stop = throbber stop time - throbber start. Is this correct?

Attached is a log of Firefox starting there are 3 throbber start log entries before a throbber stop. Which of the 3 do you use for calculating the above?

Comment 9

[Bob Clary [:bc] (inactive)]

(In reply to Kevin Brosnan [:kbrosnan] from comment #8)

> Which of the 3 do you use for calculating the above?

The first.

Comment 10

[Kevin Brosnan [Ex-Mozilla]]

I built three revisions of Fennec, one of the good and bad change sets from comment 0 and the first level of the hg bisect. I am not seeing a difference in startup times for the three builds.

my mozconfig

# Build Fennec
ac_add_options --enable-application=mobile/android

# # Android
ac_add_options --target=arm-linux-androideabi

# Add the correct paths here:
ac_add_options --with-android-ndk="/opt/android-ndk-r8d"
ac_add_options --with-android-sdk="/opt/android-sdk-linux-v21/platforms/android-17"

ac_add_options --with-android-gnu-compiler-version=4.6
ac_add_options --with-android-version=9
ac_add_options --with-system-zlib
ac_add_options --enable-profiling
ac_add_options --with-ccache

mk_add_options MOZ_OBJDIR=./objdir-droid

and data from running the command in comment 8 is at https://docs.google.com/a/mozilla.com/spreadsheet/ccc?key=0Ao22mdF2xZhudFZqMjJ2em1RUHJJREtMQVRoc0puZFE#gid=2

Comment 11

[Kevin Brosnan [Ex-Mozilla]]

Works for me after consulting with bc. Recording the data from the Google doc for completeness.

good changeset - 4c7a395fadcc - 2747.33 ms over 24 runs
bad changeset  - 2222b07ab207 - 2695.92 ms over 24 runs
first bisect   - 1e52181d41b6 - 2710.63 ms over 24 runs