That diff points some HTML elements I've removed from a previous edit I've made (previous diff is bigger and less readable).
Process: While using VE to edit, I've opened a template and copied some wikitext inside of it. I've pasting it where I needed to. It has been converted and I've then saved the page.
May be related to T145211: Headings occasionally serialised as <h2> elements rather than \n\n== Foo ==\n in Parsoid. Config: Firefox 48 on Xubuntu 16.04 LTS.
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Strip RESTBase IDs from HTML generated when wikitext is pasted | mediawiki/extensions/VisualEditor | master | +32 -17 |
This looks like a VE issue with element ids not being cleared after copy-paste - I've added those projects.
@Trizek-WMF can you add info about which template you opened on the page? That can help with reproduction and debugging.
https://fr.wikipedia.org/wiki/Mod%C3%A8le:Colonnes
I've done the following:
I dumped the edited DOM in VE by following these directions.
I looked at the HTML for the snippet in question:
<ul id="mwAQ"><li id="mwAg"><a title="Infrastructure ferroviaire" id="mwAw" rel="mw:WikiLink" href="./Infrastructure_ferroviaire">Infrastructure ferroviaire</a></li><li id="mwBA"><a title="Chemin de fer" id="mwBQ" rel="mw:WikiLink" href="./Chemin_de_fer">Chemin de fer</a></li><li id="mwBg"><a title="Gare" id="mwBw" rel="mw:WikiLink" href="./Gare">Gare</a></li></ul>
And, the bug is there in front of us. VE is preserving the ids generated by parsing the pasted wikitext. But, this is a problem since those ids could conflict with ids on existing elements on the DOM => you would cause Parsoid to reuse data-parsoid from unrelated elements.
The fix is for VE to strip element ids from HTML generated when wikitext is copy-pasted.
Change 312170 had a related patch set uploaded (by Alex Monk):
Strip element IDs from HTML generated when wikitext is pasted
Change 312170 abandoned by Alex Monk:
Strip element IDs from HTML generated when wikitext is pasted
Reason:
Yeah, this should actually be handled by someone who knows what they're doing in this part of the code. That's not me.
If we strip IDs then we break references. We need to be able to imported the generated HTML from "<ref>Foo</ref>".
Change 312170 restored by Esanders:
Strip element IDs from HTML generated when wikitext is pasted
Change 312170 merged by jenkins-bot:
Strip RESTBase IDs from HTML generated when wikitext is pasted