Issue metadata
Sign in to add a comment
|
Support permanent-consistency variations experiments in WebView (launch bug) |
||||||||||||||||||||||||||||||||||||||||
Issue descriptionFeature description: Enable permanent-consistency for webview. We have launched Finch in M68, but permanent-consistency support was missing. Eng owner: paulmiller@ Product owner: mardini@ PRD: Mocks: Design doc (4 parts): 1) Permanent-Consistency Finch Experiments in WebView (main): https://docs.google.com/document/d/1htxy8rj4qhvaHx3ssc67f8fSApO45JreTEcnRI0CvFo/edit 2) Persistent Preferences in WebView: https://docs.google.com/document/d/1X6Xoz9hl3qoAj3EhAKcvZ6hWLVcY_QpQgmUexpgDHKA/edit 3) An Efficient Low-Entropy Provider: https://docs.google.com/document/d/1LAGMRGWekGUIonRGl2PQWogXQmSCnP7g2HKwNvcljE8/edit 4) Reshuffling Low-Entropy Experiments: https://docs.google.com/document/d/1yB5xIc2OOUjQcZcXTGYjv-Mn9i3BvBaF4JoLJjVVTLQ/edit Metrics: Please list metrics that you will use to define success and the metrics that you will monitor to identify regressions. If available, please list histograms and user actions from UMA that you intend to analyze for this launch. See go/chrome-launch-metrics for Chrome examples and definitions. See go/CrOSlaunchMetrics for Chrome OS metrics template with examples and definitions. ################################################################# # Fill these surveys out as you are ready for various reviews. # ################################################################# Accessibility survey: N/A Privacy survey: When you flip Launch-Privacy to ReviewRequested, the privacy team will be notified. Once they've triaged your launch, a blocking privacy review bug will be filed. Fill out the privacy survey included in that bug. Email yitingc@ for any questions. Security: Security will review once Launch-Security is ReviewRequested. Please see https://goto.google.com/chrome-security-review-steps for Chrome browser, https://goto.google.com/chromeos-security-review-process for Chrome OS. Test survey (https://goto.google.com/chrome-test-questions): Fill in the survey at the go link https://goto.google.com/chrome-test-questions and paste the filled in survey as an update comment on this launch bug. Make sure to fill complete this when flipping to review-requested status. Email yihongg@ with any questions. UI survey: N/A
,
Dec 4
delaying to m73 because of reshuffle requirements: https://crbug.com/881517
,
Dec 18
,
Dec 19
,
Dec 19
,
Jan 8
,
Jan 8
Test survey:
1. Does the feature need manual testing?
Yes.
2. Feature implementation status
done
b. Is all the code landed in canary and ready for testing?
Yes.
3. Link to Design Doc:
Main doc:
Permanent-Consistency Finch Experiments in WebView:
https://docs.google.com/document/d/1htxy8rj4qhvaHx3ssc67f8fSApO45JreTEcnRI0CvFo
Supplemental docs:
Persistent Preferences in WebView:
https://docs.google.com/document/d/1X6Xoz9hl3qoAj3EhAKcvZ6hWLVcY_QpQgmUexpgDHKA
An Efficient Low-Entropy Provider:
https://docs.google.com/document/d/1LAGMRGWekGUIonRGl2PQWogXQmSCnP7g2HKwNvcljE8
Plan to Reshuffle Low-Entropy Experiments:
https://docs.google.com/document/d/1yB5xIc2OOUjQcZcXTGYjv-Mn9i3BvBaF4JoLJjVVTLQ
4. Link to demos (if any):
none
5. Link to Test Plan:
https://docs.google.com/document/d/1GOKNuW5MzXq-AUVKZZyBuh0RCotdvDgiQwLVixo5UlU
6. Are there flags or setup are needed to enable this feature? Is it on by default?
on by default
7. Are there any special points of interest from a testing perspective?
Everything is covered in the test plan.
8. Is the feature the same across all platforms?
Permanent-consistency is already supported on other platforms. This launch adds it to Android WebView.
a. Is the user flow same across all platforms? (Answer if applicable to both Desktop and Mobile)
N/A
b. Is the code flow/logic different on any specific platforms (i.e code branching or logic duplication)
Slightly. Android WebView saves prefs slightly differently than Android Chrome.
9. What's automated (Unit tests, Integration tests, API tests, Browser tests, E2E tests)? Please provide links to automated tests.
We added new tests under components/variations/variations_murmur_hash_unittest.cc and components/metrics/metrics_state_manager_unittest.cc. Finch itself already had many unit tests.
10. What’s not automated that needs manual testing?
See test plan.
11. What is the crbug label(s) and/or component(s) for the feature?
Mobile>WebView
12. Are there any other team members or mailing lists that should be notified if issues are found?
paulmiller@, changwan@, asvitkine@, mpearson@, desaic@
13. Rollout plan:
a. Will this roll out on by default for 100% or will it be rolled out incrementally behind a finch experiment?
On default for 100%. It's impractical to roll this out with a Finch experiment because this is a change to Finch itself.
b. Is it behind a finch flag even if it’s on by default?
No.
c. Is there a mechanism to turn off the feature if it breaks product functionality or performance?
https://crrev.com/c/1385849 is designed to be easy to revert. Aside from reverting, no.
d. How will you monitor the stability of the feature and/or experiment groups?
We’ll monitor the WebView crash dashboard, and the UMA.LowEntropySource(3)Value histograms.
14. Confirm that there is no negative meaningful performance impact and how that was checked:
Confirmed. See WebView tests results in the https://crrev.com/c/1385849 commit message. This is actually a performance improvement for Chrome; see the "Efficient Low-Entropy Provider" doc linked above.
,
Jan 8
|
|||||||||||||||||||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||||||||||||||||||
Comment 1 by paulmiller@chromium.org
, Oct 3