[17:24:50] mdholloway: Should I merge https://gerrit.wikimedia.org/r/410073 or wait a bit? [17:25:02] bearND: go for it [17:25:09] i'll add tests in a follow-up [17:25:09] k [17:27:35] mdholloway: I think if we brought something like https://gerrit.wikimedia.org/r/#/c/409176/2/lib/summary.js back then it would fix http://localhost:6928/de.wikipedia.org/v1/page/summary/Audi (https://de.wikipedia.org/api/rest_v1/page/html/Audi/173398256) [17:43:37] bearND: i guess that might be all right if we keep it isolated to the summary endpoint (as it was in that patch). i'd be hesitant about removing 'display:none' elements from lead intros generally since templates can use those in non-intuitive ways. [17:44:31] it would definitely need thorough testing with compare-extracts again, though. were you thinking of postponing the switchover again, or of adding this as a possible improvment afterward? [17:47:28] *improvement [17:48:07] mdholloway: Yes, we only want this for summaries. I think this could come later and would not be a blocker from my POV. [17:48:19] bearND: ok, that sounds good. [17:49:02] mdholloway: But I haven't heard yet from Pchelolo if they can do it today. [17:55:33] bearND: does https://gerrit.wikimedia.org/r/#/c/410216/ look ok? or should i prep today's deployment without it? [17:55:37] do what bearND ? [17:56:39] Pchelolo: summary switchover (after today's MCS deployment, happening in the next hour) [17:57:37] *next half hour, more likely [17:58:36] bearND: so here's the deal with the summary stuff - tomorrow and the day after tomorrow there's gonna be nobody from us around - I'm taking 2 days off, Marko as well.. [17:59:03] given the number of crashes and bugs we've seen with the new summary I'm worried our abcense might be a problem [18:52:52] Pchelolo: bummer, completely forgot that you're going to be out. Does that mean we should delay until Tuesday then? //cc: coreyfloyd [18:54:59] I would guess so, the history of attempts to get this out was really not encouraging. [20:18:13] bearND: it looks like https://de.wikipedia.org/api/rest_v1/page/html/Audi/173398256 has had this bad extract throughout [20:18:52] cherry-picking https://gerrit.wikimedia.org/r/#/c/409176/ doesn't fix it because the bad text is outside the 'display:none' span [20:19:02] https://phabricator.wikimedia.org/P6690 [20:19:25] the original relocateFirstParagraph transform skips it because it applies a minimum textContent.length requirement [20:20:10] enforcing a minimum textContent length was dropped in the updated extractLeadIntroduction per discussion on T149852 [20:20:10] T149852: Lead introduction edge case in MobileFrontend where paragraph precedes list element - https://phabricator.wikimedia.org/T149852 [20:26:41] (actually, it looks like even with relocateFirstParagraph it's only doing what we want arguably because of a bug working in our favor, but i digress...) [20:39:21] mdholloway: https://gerrit.wikimedia.org/r/410256 works for me [20:47:40] bearND: where do we use Parsoid HTML? iOS only or mobile web as well? [20:47:52] apart from specialized things like the summary [20:48:14] tgr: only in the Android app for now [20:48:25] thx [20:48:35] I always mix up which app does what [20:48:46] tgr you mean for page content, right? [20:48:51] yeah [20:49:13] ok [20:51:15] I was looking for things affected by T187142 [20:51:15] T187142: Deduplicate template styles in Parsoid - https://phabricator.wikimedia.org/T187142 [20:57:15] bearND: that works but i'm not exactly sure why. i have to run out for a few minutes but i'll look more when i get back. [20:57:39] mdholloway: I put it in the unit test [20:58:53] The span with display:none comprises most of the paragraph. After that part is removed the paragraph is empty. [22:03:33] bearND: oh, i see, the #coordinates span is getting stripped in stripunneededmarkup. [22:04:37] this will still leave the 'intro' property in bad shape in formatted-lead for this article. do we care about that if we're not exposing formatted-lead and not using the 'intro' property for anything but summaries? [22:04:56] this looks fine to me otherwise, as long as it doesn't cause regressions anywhere else. [22:26:22] mdholloway: I guess we don't need to sweat the formatted-lead. [22:29:46] bearND: ok then, is https://gerrit.wikimedia.org/r/#/c/410256/ ready to merge or do you want to do a pilot run with compare-extracts first? [22:30:33] mdholloway: going to run one later