Original regression was by bug 1362858, and bug 1418629 wasn't enough to fix.
By bug 1362858, we use `CHAR_CLASS_SEPARATOR` in additional to DOM word separator. But some characters such as single quote, `@` and etc are `CHAR_CLASS_SEPARATOR`, so we may check spell by incomplete word.
We shouldn't separate word by characters that is email part, URL part or conditional punctuation.
And I also update test cases for this situation. `<textarea>` is better for spell checking since it can has multiple anonymous text nodes.
Differential Revision: https://phabricator.services.mozilla.com/D39829
--HG--
extra : moz-landing-system : lando
This seems to be regression by bug 1362858.
Actually, single quotation mark is always separator for spellchecker after
landing bug 1462858. When user tries to input "doesn't", "'" becomes separator
for spellchecker. Then "doesn" will be misspell word.
So we shouldn't mark single quotation mark as separator if user is inputting
word.
Differential Revision: https://phabricator.services.mozilla.com/D29153
--HG--
extra : moz-landing-system : lando
Mark Banner
Makoto Kato