Convert talos tart addon to a webextension

RESOLVED FIXED in Firefox 64

Status

enhancement
RESOLVED FIXED
8 months ago
7 months ago

People

(Reporter: aswan, Assigned: aswan)

Tracking

Version 3
mozilla64
Points:
---
Dependency tree / graph

Firefox Tracking Flags

(firefox64 fixed)

Details

Attachments

(1 attachment)

Assignee

Description

8 months ago
No description provided.
Assignee

Comment 1

8 months ago
Mike/Jared, I think you are the two most familiar with this test.  From a relatively quick read of the code (oy!) it looks like this uses two urls, about:blank and chrome://tart/content/blank.icon.html
If possible, I'd like to get rid of all use of chrome: in the test addon to smooth the webextension transition, its pretty easy to make this test use either resource: or http: but I'm wondering if that would add undesirable noise to the test.  http would be a bigger change, just switching to resource: means tab contents would load in a content process instead of the parent process where the chrome: page loads right now.
Any opinions?
Flags: needinfo?(mconley)
Flags: needinfo?(jaws)
I'm fine with switching to resource: or http:. The test should be covering normal usage of the browser. I think previously we wanted to leave off http to limit the code paths, but since we're migrating to webextension the test results might not be comparable before/after anyways.
Flags: needinfo?(jaws)
Assignee

Comment 3

7 months ago
(In reply to Jared Wein [:jaws] (please needinfo? me) from comment #2)
> I'm fine with switching to resource: or http:. The test should be covering
> normal usage of the browser. I think previously we wanted to leave off http
> to limit the code paths, but since we're migrating to webextension the test
> results might not be comparable before/after anyways.

Well, I can continue using chrome: urls in which case the conversion to a webextension should not change the measurements as that's all just about how the extension is loaded, not about what happens during the test and how the results are gathered.  I just asked since it was unclear to me exactly where this test was meant to reside on the spectrum from a microbenchmark that doesn't necessarily represent a real world scenario to a more realistic but hence more complex and noisier real setup.  Assuming that "limit the code paths" means you'd like to keep the test focused on the browser chrome bits and avoid an extra noise, I can keep it as it currently works...
Assignee

Comment 5

7 months ago
Huh, a single talos linux64 run resulted in a noticeable speedup:
https://treeherder.mozilla.org/perf.html#/compare?originalProject=mozilla-central&newProject=try&newRevision=ccfcbde411aa8f5b8a899ed63cf3618f85fc913e&framework=1&showOnlyComparable=1&selectedTimeRange=172800

I'll trigger a few more runs in case that one run is just a big outlier.
Assignee

Updated

7 months ago
Flags: needinfo?(mconley)

Comment 6

7 months ago
Pushed by aswan@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/3133f6ab1bd8
Convert talos tart extension to a webextension r=mconley
Backed out changeset 3133f6ab1bd8 (Bug 1498420) for causing merge conflicts with Bug 1486741. Feel free to rebase and re-land whenever.
Flags: needinfo?(aswan)
https://hg.mozilla.org/mozilla-central/rev/fb6f3dea0bea
Status: NEW → RESOLVED
Last Resolved: 7 months ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla64

Comment 10

7 months ago
Backout by dvarga@mozilla.com:
https://hg.mozilla.org/mozilla-central/rev/f3bb47ac5f01
Backed out changeset 3133f6ab1bd8 for causing merge conflicts with bug 1486741
Pulsebot was gone for some hours on the weekend, don't get confused. This landed, pulsebot's comment had been delayed.
You need to log in before you can comment on or make changes to this bug.