Re: Use of non-Latin graphemes in tests

Le Dim 1 juillet 2012 15:13, Geoffrey Sneddon a écrit :
> Hey,
> Every few months I've got asked questions by colleagues as to what
> construes a pass for various bidi tests with a pass condition such as,
> "Test passes if characters are in the same order".
> These questions normally resolve around whether two glyphs are
> realizations of the same grapheme (i.e., they are allographs). This has
> been especially problematic in cases where we've failed to render
> ligatures correctly, and hence the deviation from the reference image is
> even greater.
> Attached are a couple of examples of the sort of result that causes
> issues (direction-unicode-bidi-017 and direction-unicode-bidi-026
> respectively).



> I believe both screenshots represent a pass, though
> Arabic is not a language I'm immensely familiar with. :)
> I wonder how many of the bidi tests could be rewritten to be reftests,
> and have a (very) small set of manually inspected pages (I wonder if
> there's any way to get a better reference text in cases where a reftest
> isn't possible?)… Any ideas? (I starred at the bidi algorithm for a bit,
> then realized I'd need to spend far more time than it's really worth now
> to get that all in my head again.)

The tested sentence specifically request a serif font but the comparing
image uses a non-serif font. This can only contribute to create a
differential rendering.

Ideally, the list of fonts should at least start with the same font with
which the reference image was created. And then the generic sans-serif
font should end the declaration.

There are other minor issues with those 2 tests.

+CC: Richard Ishida

Contributions to the CSS 2.1 test suite:

CSS 2.1 Test suite RC6, March 23rd 2011:

CSS 2.1 test suite harness:

Contributing to to CSS 2.1 test suite:

Received on Sunday, 1 July 2012 21:42:30 UTC