Krinkle added a comment.
@aude A week is fine. This isn't a regression from current week's branch.TASK DETAILhttps://phabricator.wikimedia.org/T177311EMAIL PREFERENCEShttps://phabricator.wikimedia.org/settings/panel/emailpreferences/To: aude, KrinkleCc: Lydia_Pintscher, gerritbot, hoo, aude,
aude added a comment.
@Krinkle is it okay for this to wait to go out w/ the train next week or should we consider a backport?
(while I'm sure the patch works good, I would prefer the train, so the code has time on test wikis etc.)TASK DETAILhttps://phabricator.wikimedia.org/T177311EMAIL
Krinkle added a comment.
In T177311#3659950, @gerritbot wrote:
Change 382383 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] Split Echo hook handlers based on services needed
https://gerrit.wikimedia.org/r/382383
@aude Thanks for the quick patch!
public static function
gerritbot added a comment.
Change 382383 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] Split Echo hook handlers based on services needed
https://gerrit.wikimedia.org/r/382383TASK DETAILhttps://phabricator.wikimedia.org/T177311EMAIL
gerritbot added a comment.
Change 382386 abandoned by Aude:
Split hook handler that runs on Echo extension setup
https://gerrit.wikimedia.org/r/382386TASK DETAILhttps://phabricator.wikimedia.org/T177311EMAIL PREFERENCEShttps://phabricator.wikimedia.org/settings/panel/emailpreferences/To:
gerritbot added a comment.
Change 382386 had a related patch set uploaded (by Aude; owner: Aude):
[mediawiki/extensions/Wikibase@master] Split hook handler that runs on Echo extension setup
https://gerrit.wikimedia.org/r/382386TASK DETAILhttps://phabricator.wikimedia.org/T177311EMAIL
gerritbot added a comment.
Change 382383 had a related patch set uploaded (by Aude; owner: Aude):
[mediawiki/extensions/Wikibase@master] Split Echo hook handlers based on services needed
https://gerrit.wikimedia.org/r/382383TASK DETAILhttps://phabricator.wikimedia.org/T177311EMAIL
hoo added a comment.
This should probably indeed be split up into different handlers, which might solve some of the problems here.
Also we could probably make WikibaseClient faster by lazy-creating the SiteLookup.
Regarding the NamespaceChecker construction: This is in parts so expensive due to
Krinkle added a comment.
The above flame graph is from HHVM profiling samples over 24 hours from production traffic. While useful to see the overall impact on average, it does not provide timings (due to being sampled). Here two profiles from XHGui for an individual request (including flamegraph).
