512501 - (svg11tests) We should pass all the W3C SVG 1.1 Full tests

Status: RESOLVED WONTFIX

(Core :: SVG, defect)

Description

[Jonathan Watt [:jwatt]]

We should work our way through the W3C testsuite for SVG 1.1 Full:

  http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/full-index.html

and fix all the tests we fail on. Relatively up to date status can be found at:

  http://www.codedread.com/svg-support-table.html

Currently we're passing roughly 67% of the tests with our current SMIL support turned on.

Comment 1

[Brian Birtles (:birtles)]

Marking dependent on bug 275196 as some tests (e.g. animate-elem-201-t) appear to rely on xml:id support.

Comment 2

[Henri Sivonen (:hsivonen)]

(In reply to comment #1)
> Marking dependent on bug 275196 as some tests (e.g. animate-elem-201-t) appear
> to rely on xml:id support.

It seems to me the right fix is for the W3C to make tests not depend on xml:id. IIRC, the SVG WG itself was of the opinion that xml:id isn't worth the trouble after the SVG 1.2 Tiny experience.

Also, bug 275196 didn't get into Gecko for perf issues, and WebKit devs have decided not to support xml:id.

Comment 3

[Olli Pettay [:smaug][[e-mail]]]

The minor perf problems bug 275196 had could be fixed (and probably are in the
latest patch).
But sure, if xml:id won't be supported everywhere, perhaps it isn't worth to
implement it.

Jonathan, do you know if SVG WG could change the spec so that id attribute
could be used as an ID?

Comment 4

[Jonathan Watt [:jwatt]]

Brian, I suggest for the moment you "fix up" the tests for your testing using something like:

  find . \( -name '*.svg' -o -name '*.html' \) -exec sed -i -e 's/xml:id/id/' {} \;

Smaug, the "id" attribute is already used as an ID on elements in the SVG namespace. If SVG tests use xml:id gratuitously, they should probably be changed to just use id.

> (In reply to comment #1)
> IIRC, the SVG WG itself was of the opinion that xml:id isn't worth the trouble
> after the SVG 1.2 Tiny experience.

I don't remember there being any final decision on xml:id, and I believe there are members in favor of it. Personally I'd rather we just treat "id" in the null namespace as ID for all content, and preferably have a new version of Namespaces in XML that requires that, and has other fixes to make namespaces user friendly. E.g.

http://groups.google.com/group/mozilla.dev.tech.xml/browse_thread/thread/e7b034b21af81242#

Comment 5

[Jonathan Watt [:jwatt]]

Removing bug 275196 from the dep list since this bug is specifically for the SVG 1.1 test suite, and there appear to be no tests in that suite that use xml:id. (The test Brian mentions is in the 1.2 Tiny suite.)

Comment 6

[Brian Birtles (:birtles)]

(In reply to comment #5)
> (The test Brian mentions is in the 1.2 Tiny suite.)

Oops, my bad. I thought I checked that. Sorry for creating a fuss about nothing.

Comment 8

[Lars Sonchocky-Helldorf]

(In reply to Jonathan Watt [:jwatt] from comment #0)
> We should work our way through the W3C testsuite for SVG 1.1 Full:
> 
>  
> http://www.w3.org/Graphics/SVG/Test/20061213/htmlObjectHarness/full-index.
> html

There's new version of the test suite in the works, which fixes some bugs of the above one:

http://dev.w3.org/SVG/profiles/1.1F2/test/harness/htmlObject/index.html

Comment 9

[Robert Longson [:longsonr]]

These tests are now obsolete, a conformant SVG 2 implementation would not pass them all.