The unreachable code after return statement warning no longer appears for me in Firefox at https://en.wikipedia.beta.wmflabs.org/wiki/Sandbox. Marking as resolved.
Yeah, let's assume that for now. Anyone re-open if it pops up again.
Here's that same error but as reported to the exception log, which is a bit easier to read (no json).
Can you try it on https://test.wikipedia.org? I've just granted you sysop there for 1 year.
Leaving open until the RESTBase change is deployed and we've confirmed at https://logstash.wikimedia.org/app/kibana#/dashboard/api-feature-usage that the !rvslots pattern no longer has WMF user-agent hits.
Thu, Aug 16
I suspect this may've been due to T199416 which is now fixed.
Yes! Thank you.
I'd recommend considering Message::text() instead of Message::parse(). This would disallow templates and parser functions, which aren't currently used in those messages.
@Tbayer If I understand correctly, you're stating that the current token is unique enough for our purposes. That's good to hear and obsoletes most of my other questions.
Seems to work on currently supported versions.
Closing as this isn't actionable. The report is from 2011, imported from BugZilla for a user who does not have a Phabricator account, and the original report does not contain enough details to act on.
Note that the "search" feature of Special:ListFiles is disabled on wikis hosted by the Wikimedia Foundation, via the wgMiserMode setting. As such, this problem does not apply to Wikimedia Commons, Wikipedia, their sister projects, including test wikis and beta cluster.
@Daimona @Huji Could you re-review https://gerrit.wikimedia.org/r/#/c/mediawiki/extensions/AbuseFilter/+/444442/ to see if there's anything that stands out that could explain the test sometimes failing? If not, it is also possible that the failure is caused by a change in core or another extension.
Wed, Aug 15
@Tbayer Is it important for the consumption of the Popups schema and other event data that the timestamp is included in the token? It might be easier, as well as simpler and more secure, if this is done during the consuming of the data.
Two quick notes:
Most benchmarkers I've used do not have this option. It might be worth considering whether the iterated code can be changed in a way to produce the expected results "within" the current model. I don't have many ideas off-hand, but one thing might be to have a way to disable the (static?) caches. Would that suffice?
- https://grafana.wikimedia.org/dashboard/db/ci-docker-jobs – It seems that, most hours, we're 100% on Docker jobs already.
- https://grafana.wikimedia.org/dashboard/db/nodepool – It seems, most hours, the 12 OpenStack instances allocated for Nodepool are "ready" (idle).