[00:51:21] 2024 is in sight! https://coverage.toolforge.org/charts/mediawiki.png [01:30:49] TimStarling: can you confirm that we still need a php74 to php81 step for Unicode? From what I can tell, there hasn't been a task like T292552 this time around yet. I suspect SRE may've mistaken the ICU-related improvements as obsoleting this step as well. [01:30:49] T292552: Rename articles and users to update our case mapping to PHP 7.4 and Unicode 11 - https://phabricator.wikimedia.org/T292552 [01:31:18] Although I do note that https://phabricator.wikimedia.org/T292552 is strangely attached to the php 8.1 upgraded task, so maybe I've got this backwards and we already prepped for php 8.1 last year? [01:32:16] php-src suggests php74->81 moves from Unicode 11 to Unicode 14 [01:32:19] https://github.com/php/php-src/commits/master/ext/mbstring/unicode_data.h [01:34:14] yes, every time we change the ICU library version we need to do this [01:34:32] although SRE can build against an old ICU library to decouple the steps [01:34:51] oh and yes case maps also [01:35:08] icu is dynamically linked, "unicode" is embedded. [01:35:11] I think? [01:35:50] yes [01:36:59] so they're doing 8.1 then? that's good news [01:37:57] well, not yet, still 3-6 months away it seems [01:38:15] but I think they're starting to upgrade libicu ahead of the debian upgrade [01:38:41] 8.2 is the latest stable now, we should switch to that [01:38:50] especially if it takes 3-6 months to upgrade [01:40:27] 8.3 is in the release candidate stage [01:40:28] We need to document this somewhere once and for all, in terms of collation, strtoupper and which steps need what functionally and constraints (sepearate from how we do the upgrade, which seems to keep changing). Like, which step requires maintenance scripts to run, or which needs shellbox to happen first and configured, at which step(s) we need to change OverrideUcfirstCharacters and when to run the maintenance script to rename page [01:40:28] titls and when presumably to change OverrideUcfirstCharacters again. And user names conflicts, does that apply to libicu or php/OverrideUcfirstCharacters change. [01:40:52] also whether changes in core are required for any of it [01:41:15] The scattered comments tthroughout tasks I can't reason about whether this upgrade is being done correctly right now. [01:41:38] some stuff happened at T329491 [01:41:39] T329491: ICU transition towards ICU 67 - https://phabricator.wikimedia.org/T329491 [01:41:47] and now more stuff at T345561 [01:41:47] T345561: Upgrade the MediaWiki servers to ICU 67 - https://phabricator.wikimedia.org/T345561 [01:42:47] some docs on how, regardless of what happened, to emperically verify that things are working "correctly" and that no steps were skipped (e.g. broken pages or user names) [01:46:15] the first step is always a minimally impactful PHP upgrade, so case map overrides need to be in place to present the old PHP case map to users, and we need either old ICU or shellbox compatibility for sort keys [01:46:37] the remaining steps are basically optional [01:47:07] and last time were fraught and complicated [02:08:10] shellbox compat, meaning, before the OS upgrade, add shellbox to continue using the current/previous ICU for a few weeks/months, called from upgraded MW servers? [02:08:19] I see the task currently provision shellbox with the next ICU version [02:10:43] hm.. ok, looks likek the "new" strategy was abandoned last wek in favour of the old way [02:11:37] but it's providing php74 with icu 67 on current debian first, should it not be php74 with icu 63 on next debian? [02:13:55] well, I guess if the php upgrade is happening later, there isn't a difference per se between updating collation on the old debian with new icu vs doing it afterward within the new debian.