Last modified: 2013-08-29 23:54:55 UTC
In this series of edits Parsoid exhibited strange behaviour: 1. Parsed everything correctly, no problems at all. https://en.wikipedia.org/w/index.php?title=2013_US_Open_%28tennis%29&diff=570482557&oldid=570478151 2. removed linebreaks in the table, changing: |{{n/a}} |{{n/a}} |{{n/a}} to: |{{n/a}}|{{n/a}}|{{n/a}} https://en.wikipedia.org/w/index.php?title=2013_US_Open_%28tennis%29&diff=570483444&oldid=570482557 3: nowikied the pipes it moved in the previous edit https://en.wikipedia.org/w/index.php?title=2013_US_Open_%28tennis%29&diff=570484857&oldid=570484805 As far as I can tell, nothing that should make any difference at all happened between 1 and 2 (they were consecutive edits to the page.
The real error here is the loss of line breaks in step #2. The nowiki-ing in Step #3 is expected. So, if #2 is fixed, this will be fine.
This does not seem to be a simple Parsoid issue: echo -ne '{|\n|{{n/a}}\n|{{n/a}}\n|}' | node parse --wt2wt {| |{{n/a}} |{{n/a}} |} Maybe VE dropped the newlines in HTML?
This does actually seem to be a selser issue, probably in the separator handling: in.txt: {| |{{n/a}} |{{n/a}} |} node parse < in.txt > out.html cat out.html | sed s/a/b/ | node parse --oldtextfile in.txt --oldhtmlfile /tmp/.html --selser --html2wt {| |{{n/a}}|{{n/a}} |}
Change 81611 had a related patch set uploaded by GWicke: Bug 53468: Don't consume trailing IEW when skipping about siblings https://gerrit.wikimedia.org/r/81611
Change 81611 merged by jenkins-bot: Bug 53468: Don't consume trailing IEW when skipping about siblings https://gerrit.wikimedia.org/r/81611
Now deployed in production.
