[02:51:38] <Krinkle>	 ref T108434, https://phabricator.wikimedia.org/T108434#10672231
[02:51:40] <stashbot>	 T108434: Some rows (from the year 2004) in SQL databases have text in latin1 encoding - https://phabricator.wikimedia.org/T108434
[02:51:57] <Krinkle>	 wrote `mwscript ~/findLegacyEncodingRows.php --wiki frwiki --table actor` to find bad entries. 
[02:52:06] <Krinkle>	 https://gerrit.wikimedia.org/r/1130770
[23:47:49] <Krinkle>	 Amir1: Ran on testwiki incl comment table this time, found non-zero, which I did not expect on testwiki
[23:48:01] <Krinkle>	 they're all cases where we seem to have trimmed comments incorrectly, I think?
[23:48:49] <Krinkle>	 Afaik we don't trim comments directly, since that would be surprising I think. If too much data is submitted via API or via UI workaround (or lack of maxlength support, in theory), we presumably reject the edit 
[23:49:03] <Krinkle>	 but these are auto-generated summaries were we include an excerpt of the page content
[23:49:13] <Krinkle>	 and that excerpt is of course trimmed to the first 100 chars or whatever
[23:58:21] <Amir1>	 Everything in mediawiki is a rabbit hole