Current Chromium DRT does not run a spellcheck except for Mac and spellcheck-related tests are not usable. (See WebViewHost::spellCheck() for detail.) Mac DRT uses MockSpellCheker class to do that check. So using it would on Windows and Linux looks fine in general. But there are concerns: - A: If we simply enable MockSpellChecker on Windows/Linux, many pixel expectations under editing/* will get broken and massive rebaselinining will be required. - B: If we add an API to DRT that enables/disables spellchecking, such rebaselining could be avoided. But porting such API to other ports will need hassle. Any thoughts? At this time I think A is preferable but its landing need care. Your feedbacks and ideas are appreciated!
(In reply to comment #0) > - A: If we simply enable MockSpellChecker on Windows/Linux, many pixel expectations under editing/* will get broken and massive rebaselinining will be required. How many tests will be affected? Can we reduce the number by dumpAsText()/dump-as-markup.js conversion?
> How many tests will be affected? Can we reduce the number by dumpAsText()/dump-as-markup.js conversion? We have about 700 png files under LayoutTests/platform/chromium-win/editing/ now. I'm not sure which fraction of them contains a ill-spelled word... I can run a experiment though. On converting them to dump-as-markup, it will be definitely helpful. It is longer way than rebaselining, though. BTW, it might be worth to have something like a "textify task force" as a part of our editing improvement effort because most of editing test can be converted to the text-style, I guess.
I vote for A if it's not too hard. I think you can use the layout test try bots to see how many tests would need to be rebaselined.
Morita-san, Even though I'm not a WebKit guru, I think the option A may need rebaselines every time when we chage the spell-checking code of WebKit. I'm not sure whether it is acceptable. Regards, Hironori Bono
> Even though I'm not a WebKit guru, I think the option A may need rebaselines every time when we chage the spell-checking code of WebKit. I'm not sure whether it is acceptable. Good point. This is what I encountered when landing http://webkit.org/b/41423. as Kent-san mentioned, - C: turning editing tests into dump-as-markup becomes more appealing now.
I think this was resolved some months ago.