Details
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | leila | T112326 [Epic] Reader segmentation research - Q2 | |||
Resolved | • jhobs | T113443 First QuickSurvey for reader segmentation research - external survey | |||
Resolved | Jdlrobson | T111445 GOAL: Run our first survey campaign | |||
Resolved | • jhobs | T110661 Deploy QuickSurveys extension to the English Wikipedia | |||
Declined | None | T96785 In-channel feedback with star rating / picklist (and on apps, simple text entry?) and backend de-spam | |||
Resolved | • Jhernandez | T104439 [GOAL]: Develop a mechanism for quick surveys on desktop + mobile | |||
Resolved | • rmoen | T110199 Get QuickSurveys enabled on beta cluster | |||
Duplicate | None | T110194 QuickSurveys code quality | |||
Resolved | Jdlrobson | T110198 Add basic browser tests | |||
Resolved | • bmansurov | T110225 Add basic browser tests: external surveys |
Event Timeline
Change 233933 had a related patch set uploaded (by Bmansurov):
Add browser tests for external surveys
Change 233930 had a related patch set uploaded (by Bmansurov):
Add an option to load a specific survey
Change 234710 had a related patch set uploaded (by Bmansurov):
Add browser tests for external surveys
Change 234710 abandoned by Bmansurov:
Add browser tests for external surveys
All tests pass locally when I disable the existing step "And the page has fully loaded". I'm not sure why that step doesn't work on my computer. And remotely, I get different errors. Does anyone know howto fix it? @phuedx, @rmoen, @Jdlrobson?
Judging on the artifact [1] attached to the report the survey seems to be loading twice. So I'd start with investigating why that might be.
[1] https://integration.wikimedia.org/ci/job/mwext-mw-selenium/669/artifact/log/External%20surveys%3A%20External%20survey%20is%20not%20present%20without%20flag.mp4
[2] https://integration.wikimedia.org/ci/job/mwext-mw-selenium/669/console
(I'm in a planning meeting all morning so will look at this later but it's possible you've just uncovered a new bug unrelated to your patch - I'd suggest you try adding the browser tests before your patch to see if the expected tests fail in those circumstance)
@Jdlrobson, thanks. Either the bucketing logic seems off or I'm not understanding it correctly. Say, my example survey has 0.1 coverage, then I'm always assigned to the control bucket, even though I should be assigned to the bucket A 10% of the time. No?
It's possible. Write some unit tests to test it more thoroughly. I believe in you Baha! You'll work it out :)
Either the bucketing logic seems off or I'm not understanding it correctly. Say, my example survey has 0.1 coverage, then I'm always assigned to the control bucket, even though I should be assigned to the bucket A 10% of the time. No?
The current bucketing mechanism operates like "10% of users see the survey all of the 100% of the time – until they interact with it".
See my review: https://gerrit.wikimedia.org/r/#/c/233933/ - issues with browser tests
Thanks, I've replied to your comment, @Jdlrobson.
Thanks for the clarification, @phuedx.
I can replicate the issue that Jenkins has locally. I will post a follow-up patch for your inspection when you wake up :)
@bmansurov see https://gerrit.wikimedia.org/r/235618 - primarily goal here is to make the feature files as readable as possible. I hope things become clearer with this patch. Feel free to squash it/take ideas from it/whatever :)