Compare
https://pt.wikipedia.org/wiki/?oldid=36341945&action=edit&preview=yes
and
https://pt.wikipedia.org/wiki/?oldid=36341945&veaction=edit
The problem is even more noticiable in a real article:
https://pt.wikipedia.org/wiki/Scapanulus_oweni?oldid=34667801&veaction=edit
It seems parsoid is eating a </span> tag:
http://parsoid.wmflabs.org/_rt/pt/?oldid=36341945
Version: unspecified
Severity: normal
URL: https://pt.wikipedia.org/wiki/Scapanulus_oweni?oldid=34667801&veaction=edit
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | cscott | T56844 Image / media handling (tracking) | |||
| Resolved | cscott | T52976 Image inside a table incorrectly removes a </span> tag |
Can we get a simpler explanation of the bug? It's hard for me to see exactly what's going on here.
I think the original issue has been fixed (?), but note the insertion of ~ 10,000px worth of height of empty HTML in the infobox table:
<tr data-parsoid="{"startTagSrc":"|-"}">
<td data-parsoid="{"autoInsertedStart":true,"autoInsertedEnd":true}">
<p data-parsoid="{}"><br data-parsoid="{}"></p>
[…]
<p data-parsoid="{}"><br data-parsoid="{}"></p></td></tr>
– see http://parsoid-lb.eqiad.wikimedia.org/ptwiki/Scapanulus_oweni?oldid=34667801
I suspect that the empty HTML is in fact a symptom of the fact that the PHP parser omits 'empty' table rows, while parsoid doesn't. nell-wikipedia had a CSS row to suppress the empty rows.
...although it seems that the rows in this case aren't, strictly speaking, empty? This merits some investigation.
This was fixed as part of T69452 where empty rows emitted by templates are stripped by parsoid as well.