235793 - fonts look weird in svg tetris sample

Status: RESOLVED WONTFIX

(Core :: SVG, defect)

Description

[basic]

cvs trunk debug build:
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7b) Gecko/20040226 Firefox/0.8.0+

The fonts on the tetris sample at croczilla looks weird on a recent svg enabled
libart backend Mozilla Firefox gtk2 xft build. Screenshots here:

http://basic.mozillanews.org/svg.png
http://basic.mozillanews.org/svg2.png

Comment 1

[Boris Zbarsky [:bzbarsky]]

I see about the same thing.  GTK1 build of seamonkey, no xft (but freetype
enabled to get text working in svg).

Comment 2

[basic]

here's another shot:
http://basic.mozillanews.org/svg-vera-sans.png

This time I've only Bitstream Vera Sans (Bold/Bold Italic/Italic/Normal) in my
font path.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Hmm... I have that font too....

Comment 4

[Robert Mohr]

WFM on 'Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7b) Gecko/20040302
Firefox/0.8.0+'.

Comment 5

[Alex Fritze]

The tetris title is a font of font-size="1px" that is scaled up to some value
depending on the size of the viewbox. So effectively we have something like this:
<svg:text font-size="1px" transform="scale(10)">Foo</svg:text>
which should result in a "Foo" of effective size "10px".
The way this is rendered by the libart/freetype backend is to get a FT glyph of
size 1px and then let FT transform it according to the current transform matrix.
This works fine for MS fonts, but the Vera fonts seem to have a problem when
being scaled up like that.
I'm inclined to mark this INVALID or WONTFIX, but possibly we could extract the
scale from the current transform matrix and request fonts at a more appropriate
size rather than relying on FT to scale correctly from subpixels. 
The relevant code is in layout/renderer/src/libart/nsSVGLibartGlyphMetrics.cpp &
nsSVGLibartGlyphGeometry.cpp.
Any takers?

Comment 6

[Boris Zbarsky [:bzbarsky]]

So the current approach is problematic if the font includes glyphs
custom-tailored for specific size ranges, right?

Comment 7

[basic]

update testcase url

Comment 8

[Jonathan Watt [:jwatt]]

Marking wontfix. We have no plans to continue development of the libart backend.