158222 - Entities in XHTML files cause XML parsing error in Chimera

Status: VERIFIED FIXED

(Camino Graveyard :: Page Layout, defect)

Description

[Greg K.]

An HTML4-standard entity such as é is rejected by Chimera when its'
containing document (accessed via the filesystem) has an extension of [.xml] or
[.xhtml], but works when the document has an extension of [.html].

Comment 1

[Greg K.]

Attachment: XHTML testcase containing é entity

Comment 2

[Greg K.]

Note that the attached testcase, delivered with a MIME type of application/xml,
also fails.

Comment 3

[Greg K.]

Forgot to mention that it works fine in Mozilla.

Comment 4

[Henri Sivonen (:hsivonen)]

Authors should use UTF-8 or NCRs and not rely on stuff defined in the external
subset. How could they know that an user agent with a non-validating
non-external-subset-reading parser doesn't come along?

Anyway, the way this "works" in Mozilla is that Mozilla uses a catalog of
*pseudo*-DTDs. Enabling the same thing in Chimera is probably very easy.

Comment 5

[Greg K.]

Are you saying that XHTML-conforming user agents are not expected to handle the
entity sets listed in Appendix A.2 ([http://www.w3.org/TR/xhtml1/#h-A2])? (I'm
not accusing you of being wrong.)

Comment 6

[Henri Sivonen (:hsivonen)]

The XHTML specs doen't require the use of a validating XML parser and
non-validating parsers aren't required to process the external subset.
Therefore, it is unsafe to rely on stuff that is declared in the external subset.

Comment 7

[saari (gone)]

->heikki, how serious is this, and why would it break in Chimera?

Comment 8

[Heikki Toivonen (remove -bugzilla when emailing directly)]

As far as seriousness goes, there isn't that much XHTML content on the web yet.
But anyone making XHTML pages would be somewhat surpised if entities did not
work. Also, this could come up if you have XML files that use XSLT
transformations. I think you should fix this for Chimera unless you plan on
stripping out our XHTML, MathML, XSLT etc. support from Chimera (no need to have
the C++ code there if it doesn't work properly). I think it is far less work to
just fix this rather than strip out support that was not meant to be stripped out.

As Henri said, we have a hard coded "XML Catalog" (maps public identifier to
system identifier, e.g. URL) in nsExpatDriver.cpp
(http://lxr.mozilla.org/mozilla/source/htmlparser/src/nsExpatDriver.cpp#187).
The code should still be there in Chimera, so I believe it is a packaging
problem. For sample packaging, look in mozilla/embedding/config, and especially
file "basebrowser-win-sup". Since you have this problem with XHTML I am sure you
have it with MathML and SVG as well, so if you build with either of those you
must make sure that the resources they need are packaged as well. Where does
Chimera packaging happen?

I think this should be owned by someone who actually knows how to build Chimera
and is able to do it ;)

Comment 9

[Henri Sivonen (:hsivonen)]

Confirming that this is simply a packaging problem. The dtd folder is missing
from the res folder inside the app bundle. If I copy the dtd folder from
Mozilla's res to Chimera's res, character entities are supported in XHTML.

Comment 10

[saari (gone)]

okay, now the question is how big is the DTD folder? :-)

->bryner for owning

Comment 11

[Heikki Toivonen (remove -bugzilla when emailing directly)]

XHTML DTD is about 8kB, MathML 64kB. MathML and SVG also need special fonts and
CSS files. Does Chimera include MathML and/or SVG support?

Comment 12

[Greg K.]

I doubt it. MathML is just now showing up in Fizzilla 1.1 builds. SVG is only
available in experimental Win and Linux builds due to that libart licensing bug.

Comment 13

[rbs]

> XHTML DTD is about 8kB, MathML 64kB

And these are ASCII strings that probably compress well in the zip. So
correctness is worth the few K.

Re: fonts, no special packaging action needed at the moment; users are currently
prompted of missing math fonts (screenshot: attachment 71830 [details]) and an URL is
provided where they can install them.

Comment 14

[Henri Sivonen (:hsivonen)]

The Navigator and NavigatorStatic targets lack a "Copy Files" build phase that'd
copy ".../dist/Embed/res/dtd/xhtml11.dtd" to "res/dtd" using the "Executables"
style of copying.

Comment 15

[Henri Sivonen (:hsivonen)]

Attachment: Patch for making PB copy xhtml11.dtd to the right place

Comment 16

[Henri Sivonen (:hsivonen)]

bryner, could you please check in if you approve the patch?

Comment 17

[Henri Sivonen (:hsivonen)]

Attachment: Patch for making PB copy xhtml11.dtd to the right place (rot fixed)

The project file was changed in CVS. Attaching a new patch.

Comment 18

[Brian Ryner (not reading)]

Checked in.

Comment 19

[Sören 'Chucker' Kuklau (gone)]

V