This edit removed some div tags, messing up the rendering because one of the <div>s has an unclosed quote! ==> <div style="clear:both; class="NavFrame">.
A user at de.wiki argues that these unintended changes shouldn't happen.
Description
Details
Project | Branch | Lines +/- | Subject | |
---|---|---|---|---|
mediawiki/services/parsoid | master | +33 -13 | More permissive attribute name parsing |
Related Objects
- Mentioned In
- T168675: Param in template argument gets parsed incorrectly
T168404: BUG: Snippets of styling code for certain East-Asian text styles are shown in the app
T168900: Notice: Undefined index: dsr in /extensions/Linter/includes/ApiRecordLint.php on line 65 - Mentioned Here
- rGPARb4187f183211: More permissive attribute name parsing
T168404: BUG: Snippets of styling code for certain East-Asian text styles are shown in the app
T168675: Param in template argument gets parsed incorrectly
T168900: Notice: Undefined index: dsr in /extensions/Linter/includes/ApiRecordLint.php on line 65
Event Timeline
Parsoid does what it can to match PHP parser's tokenizing and also handle different error scenarios. However, Parsoid will never be able to catch all possible erroneous markup and recover from them. The error recovery we add depends on how common those errors are in wikis and how much complexity it adds to the codebase. I think for low profile errors, it is reasonable to expect the markup to be fixed instead.
Change 361916 had a related patch set uploaded (by Arlolra; owner: Arlolra):
[mediawiki/services/parsoid@master] More permissive attribute name parsing
Change 361916 merged by jenkins-bot:
[mediawiki/services/parsoid@master] More permissive attribute name parsing
The Parsoid bug that prevented tokenizing as an html element is fixed, but I'm leaving this open because maybe a linter category for these types of attribute syntax errors would be useful?