[Webkit-unassigned] [Bug 25645] SVG - overflow for big sized elements

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=25645


David Yonge-Mallo <davinci at chromium.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |davinci at chromium.org




--- Comment #5 from David Yonge-Mallo <davinci at chromium.org>  2010-03-24 05:41:12 PST ---
(In reply to comment #4)
> Hi David,
> 
> did you notice the -expected.txt results don't make much sense? Can you
> investigate the cause?
> 123456789 is correct, and afterwards it's broken...
> 
>  5     RenderPath {path} at (0,0) size 0x0 [stroke={[type=SOLID]
> [color=#0000FF]}] [data="M-1000.00,12345679395506094080.00 L200.00,200.00"]
>  6     RenderPath {path} at (0,0) size 0x0 [stroke={[type=SOLID]
> [color=#00FF00]}] [data="M600.00,400.00 L1000.00,-98765435164048752640.00"]
> 
> CC'ing Darin, who might have more comments here!

The results are like that because of FloatType's (lack of) precision.  The
numbers are 1.234568e+19 and 9.876543e+19.  When these are formatted for
printing (using a format string of "%.2f"), that is how they appear.  I checked
the output on Mac and Linux, and perhaps surprisingly they are the same.  

If we want to prettify these up, we'll have to change the dump render tree
code, which will involve different changes on Mac and Linux (since the Linux
one uses Skia).

On the other hand, we could go to a higher precision representation of floats. 
Changing the definition of FloatType, however, will have repercussions beyond
just this bug.  The question is then whether around 7 decimal digits of
precision (float) is enough, and if switching to double (around 16 digits) is
worth it.  It's probable that we will rarely even need 7 significant digits on
numbers sufficiently large that precision becomes a problem.  But that's a
separate (though related) bug from overflow.

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