Last modified: 2013-12-04 02:12:21 UTC
Currently, DSR computation attempts to infer opening/closing wikitext tag widths for generated HTML nodes based on existing tsr value and knowledge of wikitext. While this works reasonably well currently, we occasionally run into bugs when there are changes made (ex: https://gerrit.wikimedia.org/r/#/c/95696/). We could possibly pass along information from tokenizer where wikitext tag widths are available and use that in DSR computation. This will not be a trivial fix because: * a lot of tokenizer productions will need fixing * we need to make sure not to break the carefully tuned code in list handler that adjusts tsr values based on how bullets are parsed to lists * dsr computation might need some tweaking based on this