[webkit-dev] Cross-platform fonts for Layout Tests

[Hao Zheng]

Unfortunately, even for SVG or images, different drawing
implementations will lead to different pixel results. Like this Skia
bug, http://code.google.com/p/skia/issues/detail?id=179 , which caused
most pages using SkFixed calculation, e.g. round-corner, gradient,
svg, etc., produce different rendered image on different platforms. So
I think there are many subtle, if possible, issues to make pixels
consistent across all platforms.

On Tue, Jun 7, 2011 at 1:23 PM, Eric Seidel <eric at webkit.org> wrote:
> So are we saying it's impossible to have matching results across all
> platforms if a test involves any text (in any font)?
>
> I know it's certainly possible to have pixel-results for tests which
> do not involve text match across all platforms (like SVG or images or
> css styling, etc.)
>
> Or is all this just theory?
>
> -eric
>
> On Mon, Jun 6, 2011 at 8:09 PM, Hao Zheng <zhenghao at chromium.org> wrote:
>> Yes, actually in Skia, Chromium/Linux uses a noop gamma implementation
>> in SkFontHost_gamma_none.cpp; however, if you use a substantial
>> implementation in SkFontHost_gamma.cpp, there will be much image
>> mismatch on Chromium/Linux for every font including Ahem. The slight
>> differences are on font fringe.
>> And I think if other WebKit is not using Skia at all, there will
>> surely be even more differences.
>>
>> As the gamma implementation on Clank is desired, we should not fix
>> this bug. Just file the bug to record this difference.
>>
>> On Fri, Jun 3, 2011 at 2:30 AM, Tony Chang <tony at chromium.org> wrote:
>>> Perhaps, but in practice, it's not enough.  Here's an ahem pixel test that
>>> is slightly different on Mac and Chromium Linux:
>>> http://trac.webkit.org/browser/trunk/LayoutTests/platform/mac/fast/block/basic/010-expected.png
>>> http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-linux/fast/block/basic/010-expected.png
>>>
>>> Also, I think it would be hard to tell by examining the HTML if a test uses
>>> another font.  For example, the line height of an empty block might depend
>>> on the default font that isn't specified (does <pre></pre> render the same
>>> height on all platforms?).
>>> On Thu, Jun 2, 2011 at 10:44 AM, Adam Barth <abarth at webkit.org> wrote:
>>>>
>>>> I thought the whole point of Ahem was to avoid those problems.
>>>>
>>>> Adam
>>>>
>>>>
>>>> On Thu, Jun 2, 2011 at 1:29 AM, Hao Zheng <zhenghao at chromium.org> wrote:
>>>> > Actually, even the same Ahem font will be rendered differently on
>>>> > different platform, depending on the font drawing library, the
>>>> > anti-aliasing algorithm, subpixel, tiny float-point calculation diff
>>>> > on different arch.
>>>> >
>>>> > On Thu, Jun 2, 2011 at 3:30 AM, Eric Seidel <eric at webkit.org> wrote:
>>>> >> I know that Ahem is safe to use across multiple platforms (the font
>>>> >> metrics
>>>> >> will be the same).  Do we know if there are any other fonts for which
>>>> >> this
>>>> >> is true?
>>>> >> I'd like to make the style-bot yell at people when they use pixel tests
>>>> >> with
>>>> >> non-safe fonts.  Right now that list would only include ahem.
>>>> >> -eric
>>>> >> _______________________________________________
>>>> >> webkit-dev mailing list
>>>> >> webkit-dev at lists.webkit.org
>>>> >> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>>>> >>
>>>> >>
>>>> > _______________________________________________
>>>> > webkit-dev mailing list
>>>> > webkit-dev at lists.webkit.org
>>>> > http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>>>> >
>>>> _______________________________________________
>>>> webkit-dev mailing list
>>>> webkit-dev at lists.webkit.org
>>>> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>>>
>>>
>> _______________________________________________
>> webkit-dev mailing list
>> webkit-dev at lists.webkit.org
>> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>>
>