505783 - Setting dynamically xml:base for XHTML elements has no effect in HTML documents

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[nanto_vi (TOYAMA Nao)]

Attachment: Testcase: setting xml:base on an XHTML "a" element

Setting the xml:base attribute on an XML (including XHTML) element changes the base URI of the element, but Mozilla now ignores xml:base if the elements is an XHTML element and the document is an HTML document (served as text/html).

See the attachment to reproduce this.

Expected:
  http://example.org/
  http://example.org/foo

Actual:
  https://bugzilla.mozilla.org/attachment.cgi?id=nnn
  https://bugzilla.mozilla.org/foo

This is a regression.  Firefox 3.5 and older works as expected.  This breaks resolving-relative-URIs techniques [1] and some Greasemonkey scripts that use the techniques [2] [3].

[1] http://www.google.com/codesearch?q=createElementNS+setAttributeNS+xml%3Abase+file%3A\.js%24
[2] http://userscripts.org/scripts/show/8551
[3] http://userscripts.org/scripts/show/22702

Comment 1

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

I think this is a regression from Bug 468708.

Comment 2

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

I suppose we could enable xml:base in HTML too :( I'm not a big fan of xml:base in general since I think it's somewhat over complicated, but if people really use it that seems like a reasonable solution.

Comment 3

[Henri Sivonen (:hsivonen)]

IIRC, HTML5/Hixie says that script-inserted xml:base in HTML documents should work.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Isn't this just a matter of this code in nsGenericHTMLElement::GetBaseURI() bailing out early?

Do we need to make sure to fix this in 1.9.2, if this is breaking real things out there?

We could just take out the early-bail and see what happens, perf-wise, I suppose...

Maybe we should flag nodes that have an ancestor with xml:base set, so that we can fast-path all the other (common!) cases?

Comment 5

[Henri Sivonen (:hsivonen)]

Attachment: Don't make xml:base special in HTML

Need to get some perf numbers for this.

Comment 6

[Henri Sivonen (:hsivonen)]

Comment on attachment 394261 [details] [diff] [review]
Don't make xml:base special in HTML

Looks OK on tryserver talos.

Comment 7

[Henri Sivonen (:hsivonen)]

http://hg.mozilla.org/mozilla-central/rev/50c82c606fcd

Are the flags on this bug now right for 3.6 approval consideration?

Comment 8

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Is this really important enough for 1.9.2? Seems very much like an edgecase. If you think we need it you need to set the approval 1.9.2? flag on the patch.

Comment 9

[Henri Sivonen (:hsivonen)]

Comment on attachment 394261 [details] [diff] [review]
Don't make xml:base special in HTML

This is unlikely to matter for Web compat. However, not having this patch in 1.9.2 could break Greasemonkey scripts or extensions as stated in the bug description.

Comment 10

[Benjamin Smedberg]

Comment on attachment 394261 [details] [diff] [review]
Don't make xml:base special in HTML

I can't imagine taking this on branch without a test (and it should have a test anyway, no?)

Comment 11

[Henri Sivonen (:hsivonen)]

Attachment: Mochitest

Could this bug get branch approval with this test case, please?

Comment 12

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Comment on attachment 404235 [details] [diff] [review]
Mochitest

Could you *also* test that things work if the <a> is inserted into the document.

r=me with that

Comment 13

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Comment on attachment 394261 [details] [diff] [review]
Don't make xml:base special in HTML

a=sicking with the testcase once fixed per above comment.

Comment 14

[Henri Sivonen (:hsivonen)]

Test pushed with additional in-document check:
http://hg.mozilla.org/mozilla-central/rev/1470fa9c9f01

Pushed to branch:
http://hg.mozilla.org/releases/mozilla-1.9.2/rev/aa636984cd78

Thanks!