[00:19:38] I couldn't get a reproducer so far, but in the interim here's the change i needed to be able to run the tests: https://gerrit.wikimedia.org/r/c/mediawiki/services/parsoid/+/643782 [00:20:15] mszabo: Nice. Thank you. [20:34:18] Krinkle https://gerrit.wikimedia.org/r/c/mediawiki/core/+/643550 (a 9 patch revert) ends up fixing the issues on REL1_35 [20:36:06] addshore: do you know the root cause as to what goes wrong in those changes? [20:36:39] afaik we made some assumptions in later refactors that implicly expect these refactors so I'd rather fix in master and backport. [20:36:53] but good to have narrowed it down and confirmed to these commits [20:47:33] Krinkle: nope, i couldn't really get my head around it so far, but didnt have much time to invest really [20:48:12] Yeah, these reverts cover a 10 month timespan or something silly, but at least 1.35 seems to work for me now [20:48:28] I'd also rather just fix it in master and backport a fix :) [20:49:50] This is also just me trying to actually confirm my suspicions that it was this patch causing all of my issues, and gently applying more pressure ;) [21:07:21] I'm pretty sure that a few of those commits you can immediately rule out [21:07:39] "Fix typo: occured → occurred" did not cause the bug [21:43:57] Reedy: indeed, but they caused a bunch of conflicts I didn't feel like resolving by hand and making a mess with before seeing if it would actually work [21:44:21] seemingly that patch that landed on REL1_35 in the past 10 days fixes one part of the issue, with another bit seemingly still remaining [21:44:55] I'm unblocked now :) so will probably stop pursuing it, but it is still broken somehow :D