602854 - Bad rendering SVG via XSLT over JavaScript API

Status: RESOLVED WORKSFORME

(Core :: SVG, defect)

Description

[Hynek Musil]

User-Agent:       Mozilla/5.0 (Windows NT 5.1; rv:2.0b8pre) Gecko/20101007 Firefox/4.0b8pre
Build Identifier: Mozilla/5.0 (Windows NT 5.1; rv:2.0b8pre) Gecko/20101007 Firefox/4.0b8pre

The test is on the page: http://test.formax.cz/xslt-svg-bug/

In the onload XSLT transformation starts:
...
<xsl:template match="/*">
  <xsl:variable name="x1" select="10"/>
  <xsl:variable name="y1" select="10"/>
  <svg height="100" width="100" xmlns="http://www.w3.org/2000/svg">
    <polyline points="0,0 {$x1},{$y1}"/>
  </svg>
</xsl:template>

Javascript:
...
var fragment = stylesheetProc.transformToFragment(xmlDoc, document);
document.body.appendChild(fragment);
...

Reproducible: Always

Steps to Reproduce:
1. Test is on the page: http://test.formax.cz/xslt-svg-bug/

Actual Results:  
...
<svg xmlns="http://www.w3.org/2000/svg" width="100" height="100"><style type="text/css">
	<polyline points="0,0"/>
...

Expected Results:  
...
<svg xmlns="http://www.w3.org/2000/svg" width="100" height="100"><style type="text/css">
	<polyline points="0,0 10,10"/>
        ...

Comment 1

[(mostly gone) XtC4UaLL [:xtc4uall]]

Attachment: Reporter's Testcase

Comment 2

[Boris Zbarsky [:bzbarsky]]

Presumably because SVG normalizes attribute values, right?  Testcase:

data:text/html,<svg height="100" width="100" xmlns="http://www.w3.org/2000/svg"><polyline points="0,0 {$x1},{$y1}"/></svg><script>alert(document.getElementsByTagName("polyline")[0].getAttribute("points"))</script>

I believe the XSLT folks want to start using generic XML for the templates for this reason, but should we really lose the original value of @points there?

Comment 3

[Boris Zbarsky [:bzbarsky]]

Oh, the XSLT thing wouldn't help because getSource parses the SVG via the normal XML parser, so will in fact create SVG elements.

Comment 4

[Boris Zbarsky [:bzbarsky]]

nsSVGNumberList::SetValueString will happily parse whatever and return NS_OK.  So we never end up using the proxy value stuff.

Seems like we should be able to remove the proxy value business and just use the same setup as style attributes, right?  And make SetValueString pass the string to DidModify even on error conditions so the SetAttr call that gets made there will actually save the right value?

Comment 5

[Boris Zbarsky [:bzbarsky]]

Attachment: Start of a patch in that direction; enough to fix this bug

Maybe we should just take this as-is, even....

Comment 6

[Jonathan Watt [:jwatt]]

The combination of bug 610990 followed by bug 630083 should have fixed this, I think. Maybe we should still be adding a test before closing this though.

Comment 7

[Hynek Musil]

It's ok, thanks :-)

Comment 8

[Loic]

Is it fixed?

Comment 9

[Robert Longson [:longsonr]]

comment 2 works so I'm assuming this is fixed.