Closed
Description
I use Japanese and Taiwanese Mandarin input methods, which have font variants which are sometimes very different. Currently, the candidate panel always shows me the Japanese variant of any 漢字 candidate, so in my case the candidate looks very wrong when I type "直" in Chinese, because of how different these variants are.
In Japanese it's 直, while in Chinese fonts it's 直. And there are subtler differences, things like like the styling on the top stroke of 言 (言{ja} vs 言{zh-TW}).
I think it may be possible to automatically select the correct variant by setting the pango language attribute for candidate text based on the engine's advertised language.