Double escaping on widgets ?downcast?

[xmo]

When introducing non-ascii characters in widget content, after destroying the editor the characters end up escaped to entities twice (and thus one ends up with the character's HTML entity displaying raw: first the character is escaped to an entity, then the entity's opening & is itself escaped to & e.g. the character é ends up saved as é).

The main text (outside of widgets) has no such issue.

Demonstration:

• go to the ckeditor widgets demo page
• In the simple box's content, insert a non-ascii character e.g. é or ñ
• Switch to source, the character is visibly double-escaped
• Switch back to interpreted, the character has been replaced by its HTML entity

Oddly enough, characters with no named entities don't seem affected e.g. 中国 ("China" in chinese), 한국 ("Korea" in korean) or 日本 ("Japan" in japanese) are unaffected, and so are Россия ("Russia" in russian), المملكة العربية السعودية ("Kingdom of Saudi Arabia" in arabic) or ราชอาณาจักรไทย ("Kingdom of Thailand" in thai), but Lýðveldið Ísland ("Republic of Iceland" in icelandic) breaks. And Cộng hòa Xã hội chủ nghĩa Việt Nam ("Social ist Republic of Vietnam" in vietnamese) is partially broken and partially unbroken.

(nb: had to breakup social ist in 2 because the spam filter is dumb)

[Piotrek Koszuliński]

DUP of #11372. We are going to fix it in CKEditor 4.3.2.

[xmo]

Whoops sorry, tried a search but did not match anything.

Thanks.

[Piotrek Koszuliński]

No problem. It's never easy :D