The first few lightning deploys could have been covered by an infrastructure which tested for CSS regressions. I started an e-mail thread which got some interesting replies so I think we should investigate that as a spike [1].
This latest one is kind of weird and I'm not sure how a test could accurately solve it, but in theory this was a result of sloppiness, shouldn't happen often and should have been caught in code review.
On the plus side, we are catching more bugs earlier with our browser automation tests and I feel confident that we will be needing less of these. :)
[1] https://wikimedia.mingle.thoughtworks.com/projects/mobile/cards/1779
On Tue, Mar 18, 2014 at 5:26 PM, Kenan Wang kwang@wikimedia.org wrote:
It seems that the past three issues we've seen haven't been something that we are able to write a regression test for...
Sent from my iPhone
On Mar 18, 2014, at 4:53 PM, Jon Robson jdlrobson@gmail.com wrote:
Due to human error we've pointed all of our wiki projects nearby pages to hit English Wikipedia. This is bad for many reasons
- Broken links on production on all our wikis other than enwiki
- enwiki now jsonps itself.
- enwiki will now get the sum of all traffic across all our projects
for the nearby page (can it cope?!) 4) projects like French Wikipedia will get English results on Nearby.
For these reasons I suggest we lightning deploy this asap: https://gerrit.wikimedia.org/r/#/c/119416/
In terms of acceptance tests to protect this from happening again - this is not really possible due to the longstanding issues we are having with being able to test nearby with browser tests.
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l