[16:54:36] Pchelolo: do you want me to work on T288639? [16:54:36] T288639: SpamBlacklistHooks::onEditFilterMergedContent causes every edit to be rendered twice - https://phabricator.wikimedia.org/T288639 [19:03:38] I suspect links to global user pages in edit summaries were blue instead of red until recently. Anyone else have a better memory on this? https://www.mediawiki.org/w/index.php?title=Project:Sandbox&diff=4902855&oldid=4902286 [19:04:06] could be fall out from comment batch formatting refactor perhaps not calling isKnown() anymore after preloading the local data. [19:11:59] Krinkle: https://phabricator.wikimedia.org/T293665 [19:15:00] ah, ok, yeah, there's definitely a common pattern there of the red links all being for things that don't have a "simple" local page, so batch lookup would fail there [20:47:17] Krinkle: GeneralizedSql goes to great lengths to generalize and normalize the sql so it would be aggergateable in logstash. Right? [20:47:48] if that's the case, then why it appends trxId to the sql? [20:48:02] * Amir1 headdesks [20:48:14] Amir1: yes, yes, and yes. [20:48:21] Amir1: but cause and effect are not connected here [20:48:43] it is generalized and normalized mainly for security/privacy, and to make debugging easier by removing irrelevant information. [20:48:51] not for ease of aggregation [20:49:01] but yeah, seems feasible to get there [20:49:19] Amir1: The primary way in which I aggregate queries and reports about them, though, is the "Function:" line, not the "Query" line. [20:49:30] That is 99.9% 1:1 with a logical query in the code. [20:49:40] noted, I think if we want to replace tendril's slow queries with this, we might have to still send the values to logstash [20:49:56] including across layers of abstraction. whichever function is conceptually in charge of (most) of the query, is the fname we pass. [20:50:52] I suggest sending them all [20:50:54] so I'd aggregate on that if possible, but given kibana limitations on aggreate===displayed value, maybe normalized sql would be better, or maybe it's possible to group by fname, and then display a random sqline (like how SQL itself does grouping when you select multipe columns) [20:50:58] sending where? [20:51:11] logstash [20:51:19] oh you mean new logging, not existing. [20:51:21] right, yeah. [20:58:52] I hope you're fine with me making some patches to it [22:47:37] TimStarling: hi, for when you have a bit of time. This turns djvu to use json for metadata https://gerrit.wikimedia.org/r/c/mediawiki/core/+/738280 [22:47:57] tested locally and CI tests are also happy [22:52:33] I'm looking [22:53:30] thanks