[00:05:18] thanks Reedy [07:44:18] Amir1: there's quite a lot more under https://en.wikipedia.org/wiki/Special:LinkSearch?target=wikimedia.org than just wikis. I imagine we'll see a noticalbe drop in db size once this is fully internalised by refresh links over time. (Assuming no optimize for most hosts, I expect this will only show up visually in your graphs by lakc of growth rather than actual shrinkage) [07:44:45] e.g upload.wikimedia.org and mail.wikimedia.org as well [10:16:23] https://grafana-rw.wikimedia.org/d/QLtC93rMz/backend-pageview-timing?orgId=1&from=now-90d&to=now&timezone=utc&var-platform=$__all&forceLogin=true&viewPanel=panel-43 [10:17:02] mobile latency regressed from ~300ms to ~600ms at p75 [10:17:09] desktop elevated as well but much less so [10:17:37] cscott: ^ may be of interest, re: parsoid and output transform changes [10:17:53] https://grafana-rw.wikimedia.org/d/QLtC93rMz/backend-pageview-timing?orgId=1&from=now-90d&to=now&timezone=utc&var-platform=$__all&forceLogin=true [10:40:36] ori: see T413958 for enforcing this via phpcs [10:40:37] T413958: Enforce `use function` import for array_key_exists() (and other common PHP functions?) in namespaced files via PHPCS - https://phabricator.wikimedia.org/T413958 [11:45:20] Reedy: at long last, i managed to avoid getting sidetracked for long enough to push an update to the parsoid patch :p [12:42:06] Krinkle: yup, I estimate it'll be dropping 200M rows in commons \o/ [16:37:21] Hi! I've submitted a patch for T413938 (diff pages crashing in Chrome 65-66 due to timezone RangeError). Fixed DateFormatter to use 'Etc/GMT' instead of 'Etc/GMT-0' for zero offset. Patch: https://gerrit.wikimedia.org/r/c/mediawiki/core/+/1224121 - would appreciate a review! [16:37:22] T413938: Diff pages do not render on older browsers: Last modified bar throws exception meaning diff popup overlay doesn't render - https://phabricator.wikimedia.org/T413938 [16:43:00] aarohi: Generally you don't need to go around asking for CR after ~20 minutes of uploading a patch. You've correctly attached it to a bug, so the reporter will see that "soon" and usually would respond during their working day [16:58:54] Reedy: just to 100% clarify, in https://gerrit.wikimedia.org/r/c/mediawiki/vendor/+/1221225/comment/ba34be93_a05cd687/ are you saying i should rebase the parsoid patch (i.e. now, before it gets merged)? [16:59:40] There's no harm in doing so... Except when you start rebasing everytime another patch is merged into the repo (unless they conflict!) [17:00:25] fair enough :) i guess i'll find out if gerrit thinks the rebase is 'trivial' enough to keep the previous +1s :p [17:00:37] I think based on what has been merged.. it should be [17:04:36] you were right! [17:05:07] Like I say, I suspect we can get a +1 (or +2) from cscott in the next couple of days, which will get it merged [17:08:33] on a slight side note, it'd be nice if the CI infra could automagically load a parsoid patch into vendor/ without needing to also have an additional 3 DNM patches to get it to work. i admit that i don't know how feasible that would be, though [17:12:46] If the patch was merged, you could use dev-master#hash in the vendor repo [17:12:53] without needing to fiddle with the repo [17:13:20] But this is a very uncommon workflow to be wanting to test vendorised library patches in mw core before merge like this [17:14:06] composer uses github, and I guess the branches for patches pushed for review won't be replicated over there.. if we could point it at gerrit, there might be a way [17:22:06] i mean, to be fair i can easily believe you that this is a fairly uncommon scenario re testing. so maybe something like that wouldn't be worth the implementation effort :) [17:37:31] Thanks for letting me know, I'm a new contributor.. [21:22:12] Krinkle: so, regarding fesh, AFAIK, install-freshnode needs a bump to FRESH_VERSION, the SHA256 for fresh-node24 should be updated to match the current content, and the tag of the bumped version should be made pointing to the current master HEAD. Then the https://gerrit.wikimedia.org/g/fresh/+/%s/bin/%s?format=TEXT download and hash check should succeed. Manually putting fresh-node24 in ~/bin seems to work for me. [21:23:18] Krinkle: I've been using it for running candidate take-home tests that use node (some need node 24).