Closed Bug 1665440 Opened 4 years ago Closed 4 years ago

Remove conditioned profiles from Android startup tests

Categories

(Testing :: mozperftest, task)

Default
task

Tracking

(firefox83 fixed)

RESOLVED FIXED
83 Branch
Tracking Status
firefox83 --- fixed

People

(Reporter: acreskey, Assigned: acreskey)

Details

Attachments

(1 file)

We enabled conditioned profiles for the Fenix startup tests.

However when verifying the performance of the tests, :mcomella noticed that they have a significant impact, slowing down startup.

As an example:

With conditioned profiles: 3162.93ms on G5
https://treeherder.mozilla.org/#/jobs?repo=try&selectedTaskRun=Csgr2Of7SuSV8gp8QLpZpg.0&tier=1%2C2%2C3&revision=c5f2fba98fb8b3b22668d14097217f7ae114957e

Without: 2617.71ms on G5
https://treeherder.mozilla.org/#/jobs?repo=try&tier=1%2C2%2C3&revision=8fd36f665d47c2cc52b6e84dd7cfa321792fe84a&selectedTaskRun=XKgKndA9TdGMSF2pEtYLOA.0

One of the causes was found in Bug 1664025.
This is problematic for the multi-commit test since the gecko version can change for a given commit, but one conditioned profile is used for all the tests.

Since removing the conditioned profiles helps us match the performance of the simpler FNPRMS stack, we will remove them for now.

Summary: Remove conditioned profiles from Fenix startup tests → Remove conditioned profiles from Android startup tests

Remove the use of conditioned profiles for Fenix, GVE, since they create a user scenario that we are not seeing in local testing.

Pushed by csabou@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/a4210548b707
Remove conditioned profiles from Android startup tests r=sparky
Status: NEW → RESOLVED
Closed: 4 years ago
Resolution: --- → FIXED
Target Milestone: --- → 83 Branch
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: