[Webkit-unassigned] [Bug 118053] Parsing of MathML length

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=118053





--- Comment #7 from Frédéric Wang <fred.wang at free.fr>  2013-06-26 14:23:14 PST ---
> I see. 
> 
> I don't know if this is the best way to test these features. For example, the veryverythinspace would pass as long as it's different from the default thickness. 
> 
> But if the parsing logic changed and this value became different somehow, it could still pass as long as it was different.
> 
> I feel like having expected results for these will result in more strict tests

Here veryverythinspace is a == test so that should not happen, although what you say is true with != tess. What can happen is if the linethickness parsing is if completely broken (e.g. always use the default value) then == tests will always pass. However, the idea is to combine several != and == reftests and to rely on other features that are expected to be stable (e.g. CSS or HTML). This generally gives a good test coverage and using image reference is not needed. The tests for mspace would be important here as they test against CSS/HTML.

Strict tests are sometimes painful, especially for MathML where the exact pixel rendering is not defined as opposed to e.g. SVG or CSS. I've followed the WebKit MathML developments in the last few years and basically each time some tweaking is done on the rendering (e.g. modify the spacing in formulas), it breaks all the tests that have an image reference. In that situation, reftests are much more robust. The Mozilla and MathJax projects really only use these formats for layout tests.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.